Insulin responsiveness of glucose transporter 4 in 3T3-L1 cells depends on the presence of sortilin

Insulin-dependent translocation of glucose transporter 4 (Glut4) to the plasma membrane of fat and skeletal muscle cells plays the key role in postprandial clearance of blood glucose. Glut4 represents the major cell-specific component of the insulin-responsive vesicles (IRVs). It is not clear, however, whether the presence of Glut4 in the IRVs is essential for their ability to respond to insulin stimulation. We prepared two lines of 3T3-L1 cells with low and high expression of myc₇-Glut4 and studied its translocation to the plasma membrane upon insulin stimulation, using fluorescence-assisted cell sorting and cell surface biotinylation. In undifferentiated 3T3-L1 preadipocytes, translocation of myc₇-Glut4 was low regardless of its expression levels. Coexpression of sortilin increased targeting of myc₇-Glut4 to the IRVs, and its insulin responsiveness rose to the maximal levels observed in fully differentiated adipocytes. Sortilin ectopically expressed in undifferentiated cells was translocated to the plasma membrane regardless of the presence or absence of myc₇-Glut4. AS160/TBC1D4 is expressed at low levels in preadipocytes but is induced in differentiation and provides an additional mechanism for the intracellular retention and insulin-stimulated release of Glut4.Adipocytes, skeletal muscle cells, and some neurons respond to insulin stimulation by translocating intracellular glucose transporter 4 (Glut4) to the plasma membrane. In all these cells, the insulin-responsive pool of Glut4 is localized in small membrane vesicles, the insulin-responsive vesicles (IRVs; Kandror and Pilch, 2011 ; Bogan, 2012 ). The protein composition of these vesicles has been largely characterized (Kandror and Pilch, 2011 ; Bogan, 2012 ). The IRVs consist predominantly of Glut4, insulin-responsive aminopeptidase (IRAP), sortilin, low-density-lipoprotein receptor–related protein 1 (LRP1), SCAMPs, and VAMP2. Glut4, IRAP, and sortilin physically interact with each other, which might be important for the biogenesis of the IRVs (Shi and Kandror, 2007 ; Shi et al., 2008 ). In addition, the IRVs compartmentalize recycling receptors, such as the transferrin receptor and the IGF2/mannose 6-phosphate receptor, although it is not clear whether these receptors represent obligatory vesicular components or their presence in the IRVs is explained by mass action (Pilch, 2008 ), inefficient sorting, or other reasons.Deciphering of the protein composition of the IRVs is important because it is likely to explain their unique functional property: translocation to the plasma membrane in response to insulin stimulation. Even if we presume that IRV trafficking is controlled by loosely associated peripheral membrane proteins, the latter should still somehow recognize the core vesicular components that create the “biochemical individuality” of this compartment. In spite of our knowledge of the IRV protein composition, however, the identity of the protein(s) that confer insulin sensitivity to these vesicles is unknown.Insulin responsiveness of the IRVs was associated with either IRAP or Glut4. Thus it was shown that Glut4 interacted with the intracellular anchor TUG (Bogan et al., 2003 , 2012 ), whereas IRAP associated with other proteins implemented in the regulation of Glut4 translocation, such as AS160 (Larance et al., 2005 ; Peck et al., 2006 ), p115 (Hosaka et al., 2005 ), tankyrase (Yeh et al., 2007 ), and several others (reviewed in Bogan, 2012 ). Results of these studies, or at least their interpretations, are not necessarily consistent with each other, as the existence of multiple independent anchors for the IRVs is, although possible, unlikely.Ablation of the individual IRV proteins has also led to controversial data. Thus knockout of IRAP decreases total protein levels of Glut4 but does not affect its translocation in the mouse model (Keller et al., 2002 ). On the contrary, knockdown of IRAP in 3T3-L1 adipocytes has a strong inhibitory effect on translocation of Glut4 (Yeh et al., 2007 ). In yet another study, knockdown of IRAP in 3T3-L1 adipocytes did not affect insulin-stimulated translocation of Glut4 but increased its plasma membrane content under basal conditions (Jordens et al., 2010 ). By the same token, total or partial ablation of Glut4 had various effects on expression levels, intracellular localization, and translocation of IRAP (Jiang et al., 2001 ; Abel et al., 2004 ; Carvalho et al., 2004 ; Gross et al., 2004 ; Yeh et al., 2007 ). Knockdown of either sortilin or LRP1 decreased protein levels of Glut4 (Shi and Kandror, 2005 ; Jedrychowski et al., 2010 ).One model that might explain these complicated and somewhat inconsistent results is that depletion of either major integral protein of the IRVs disrupts the network of interactions between vesicular proteins and thus decreases the efficiency of protein sorting into the IRVs (Kandror and Pilch, 2011 ). Correspondingly, the remaining IRV components that cannot be faithfully compartmentalized in the vesicles are either degraded (Jiang et al., 2001 ; Keller et al., 2002 ; Abel et al., 2004 ; Carvalho et al., 2004 ; Shi and Kandror, 2005 ; Yeh et al., 2007 ; Jedrychowski et al., 2010 ) or mistargeted (Jiang et al., 2001 ; Jordens et al., 2010 ), depending on experimental conditions and types of cells used in these studies. In other words, knockdown of any major IRV component may decrease vesicle formation along with insulin responsiveness. Thus, in spite of a large body of literature, the identity of protein(s) that confer insulin responsiveness to the IRVs is unknown.Here we used a gain-of-function approach to address this question. Specifically, we attempted to “build” functional IRVs in undifferentiated 3T3-L1 preadipocytes by forced expression of the relevant proteins. Undifferentiated preadipocytes do not express Glut4 or sortilin and lack IRVs (ElJack et al., 1999 ; Shi and Kandror, 2005 ; Shi et al., 2008 ). Correspondingly, IRAP, which is expressed in these cells, shows low insulin response (Ross et al., 1998 ; Shi et al., 2008 ). We found that ectopic expression of increasing amounts of Glut4 in undifferentiated preadipocytes does not lead to its marked translocation to the plasma membrane upon insulin stimulation. On the contrary, sortilin expressed in undifferentiated preadipocytes was localized in the IRVs and was translocated to the plasma membrane in response to insulin stimulation. Moreover, upon coexpression with Glut4, sortilin dramatically increased its insulin responsiveness to the levels observed in fully differentiated adipocytes. Thus sortilin may represent the key component of the IRVs, which is responsible not only for the formation of vesicles (Shi and Kandror, 2005 ; Ariga et al., 2008 ; Hatakeyama and Kanzaki, 2011 ), but also for their insulin responsiveness. It is worth noting that sortilin levels are significantly decreased in obese and diabetic humans and mice (Kaddai et al., 2009 ). We thus suggest that sortilin may be a novel and important target in the fight against insulin resistance and diabetes.Our experiments also demonstrate that undifferentiated preadipocytes lack a mechanism for the full intracellular retention of Glut4 that can be achieved by ectopic expression of AS160/TBC1D4.     

Secretory production of an FAD cofactor-containing cytosolic enzyme (sorbitol–xylitol oxidase from Streptomyces coelicolor) using the twin-arginine translocation (Tat) pathway of Corynebacterium glutamicum

Carbohydrate oxidases are biotechnologically interesting enzymes that require a tightly or covalently bound cofactor for activity. Using the industrial workhorse Corynebacterium glutamicum as the expression host, successful secretion of a normally cytosolic FAD cofactor-containing sorbitol–xylitol oxidase from Streptomyces coelicolor was achieved by using the twin-arginine translocation (Tat) protein export machinery for protein translocation across the cytoplasmic membrane. Our results demonstrate for the first time that, also for cofactor-containing proteins, a secretory production strategy is a feasible and promising alternative to conventional intracellular expression strategies.The secretory expression of recombinant proteins can offer significant process advantages over cytosolic production strategies, since secretion into the growth medium greatly facilitates downstream processing and therefore can significantly reduce the costs of producing a desired target protein (Quax, 1997). And, in fact, the enormous secretion capacity of certain Gram-positive bacteria (e.g. various Bacillus species) has been used since many years in industry for the production of mainly host-derived secretory proteins such as proteases and amylases, resulting in amounts of more than 20 g l⁻¹ culture medium (Harwood and Cranenburg, 2008). In contrast, attempts to use Bacillus species for the secretory production of heterologous proteins have often failed or led to disappointing results, a fact that, among other reasons, could in many cases be attributed to the presence of multiple cell wall-associated and secreted proteases that rapidly degraded the heterologous target proteins (Li et al., 2004; Sarvas et al., 2004; Westers et al., 2011). Therefore, an increasing need exists to explore alternative host systems with respect to their ability to express and secrete problematic and/or complex heterologous proteins of biotechnological interest.So far, the Gram-positive bacterium Corynebacterium glutamicum has been used in industry mainly for the production of amino acids and other low-molecular weight compounds (Leuchtenberger et al., 2005; Becker and Wittmann, 2011; Litsanov et al., 2012). However, various recent reports have indicated that C. glutamicum might likewise possess a great potential as an alternative host system for the secretory expression of foreign proteins. Corynebacterium glutamicum belongs to a class of diderm Gram-positive bacteria that, besides the cytoplasmic membrane, possess an additional mycolic acid-containing outer membrane-like structure that acts as an extremely efficient permeability barrier for hydrophilic compounds (Hoffmann et al., 2008; Zuber et al., 2008). Despite this fact, an efficient secretion of various heterologous proteins into the growth medium of this microorganism has been observed (e.g. Billman-Jacobe et al., 1995; Meissner et al., 2007; Kikuchi et al., 2009; Tateno et al., 2009; Tsuchidate et al., 2011).In bacteria, two major export pathways exist for the transport of proteins across the cytoplasmic membrane that fundamentally differ with respect to the folding status of their respective substrate proteins during the actual translocation step. The general secretion (Sec) system transports its substrates in a more or less unfolded state and folding takes places on the trans side of the membrane after the actual transport event (Yuan et al., 2010; du Plessis et al., 2011). In contrast, the alternative twin-arginine translocation (Tat) system translocates its substrates in a fully folded form and therefore provides an attractive alternative for the secretory production of proteins that cannot be produced in a functional form via the Sec route (Brüser, 2007). Carbohydrate oxidases are biotechnologically interesting enzymes (van Hellemond et al., 2006) that are excluded from Sec-dependent secretion since they depend on a tightly or covalently bound cofactor for their activity and, for this reason, require that their folding and cofactor insertion has to take place in the cytosol. Because C. glutamicum has shown to be an excellent host for the Tat-dependent secretion of the cofactor-less model protein GFP (Meissner et al., 2007; Teramoto et al., 2011), we now asked whether it is likewise possible to secrete a cofactor-containing enzyme into the supernatant of C. glutamicum using the same protein export route.As a model protein, we chose the sorbitol–xylitol oxidase (SoXy) from Streptomyces coelicolor, a normally cytosolic enzyme that possesses a covalently bound FAD molecule as cofactor (Heuts et al., 2007; Forneris et al., 2008). FAD is incorporated into the apoprotein in a post-translational and self-catalytic process that only occurs if the polypeptide chain has adopted a correctly folded structure (Heuts et al., 2007; 2009). To direct SoXy into the Tat export pathway of C. glutamicum, we constructed a gene encoding a TorA–SoXy hybrid precursor in which SoXy is fused to the strictly Tat-specific signal peptide of the periplasmic Escherichia coli Tat substrate trimethylamine N-oxide reductase (TorA) (Fig. 1) which, in our previous study, has been proven to be a functional and strictly Tat-specific signal peptide also in C. glutamicum (Meissner et al., 2007). The corresponding torA–soxy gene was cloned into the expression vector pEKEx2 (Eikmanns et al., 1991) under the control of an IPTG-inducible P_tac promotor. After transformation of the resulting plasmid pTorA–SoXy into the C. glutamicum ATCC13032 wild-type strain, two independent colonies of the resulting recombinant C. glutamicum (pTorA–SoXy) strain and, as a control, a colony of a strain that contained the empty expression vector without insert [C. glutamicum (pEKEx2)] were grown in CGXII medium (Keilhauer et al., 1993) at 30°C for 16 h in the presence of 1 mM IPTG. Subsequently, the proteins present in the culture supernatants were analysed by SDS-PAGE followed by staining with Coomassie blue. As shown in Fig. 2, in the supernatants of the pTorA–SoXy-containing cells (lanes 3 and 4), a prominent protein band of approximately 44 kDa can be detected, the size of which is very similar to the calculated molecular mass (44.4 kDa) of SoXy. Since this band is completely lacking in the supernatant of the control strain (lane 2), this strongly suggests that this band corresponds to SoXy that has been secreted into the culture supernatant of C. glutamicum (pTorA–SoXy). And, in fact, this suggestion was subsequently confirmed in a direct way by MALDI-TOF mass spectrometry after extraction of the protein out of the gel followed by tryptic digestion (Schaffer et al., 2001) (data not shown).Open in a separate windowFigure 1The TorA–SoXy hybrid precursor protein. Upper part: Schematic drawing of the relevant part of the pTorA–SoXy expression vector. P_tac, IPTG-inducible tac promotor. RBS, ribosome binding site. To maintain the authentic TorA signal peptidase cleavage site, the first four amino acids of the mature TorA protein (black bar) were retained in the TorA–SoXy fusion protein. White bar: TorA signal peptide (TorA^SP); grey bar: SoXy (amino acids 2–418). Lower part: Amino acid sequence of the signal peptide and early mature region of the TorA–SoXy hybrid precursor. The twin-arginine consensus motif of the TorA signal peptide is underlined. The four amino acids derived from mature TorA are shown in italics. The signal peptidase cleavage site is indicated by an arrowhead.Open in a separate windowFigure 2Secretion of SoXy into the growth medium of C. glutamicum. Cells of C. glutamicum ATCC13032 containing the empty vector pEKEx2 and two independently transformed colonies of C. glutamicum (pTorA–SoXy) were grown overnight in 5 ml of BHI medium (Difco) at 30°C. The cells were washed once with CGXII medium (Keilhauer et al., 1993) and inoculated to an OD₆₀₀ of 0.5 in 5 ml of fresh CGXII medium containing 1 mM IPTG. After 16 h of further growth at 30°C, the supernatant fractions were prepared as described previously (Meissner et al., 2007). Samples corresponding to an equal number of cells were subjected to SDS-PAGE followed by staining with Coomassie blue. Lane 1, molecular mass marker (kDa). Lane 2, C. glutamicum (pEKEx2); lanes 3 and 4, C. glutamicum (pTorA–SoXy). The position of the secreted SoXy protein is indicated by an arrow.Next, the supernatant of C. glutamicum (pTorA–SoXy) was analysed for SoXy enzyme activity by measuring the production of H₂O₂ that is formed during the enzymatic conversion of sorbitol to fructose (Meiattini, 1983). Six hours after induction of gene expression by 1 mM IPTG, an enzymatic activity of 10.3 ± 1.6 nmol min⁻¹ ml⁻¹ could be determined in the supernatant of C. glutamicum (pTorA–SoXy). In contrast, no such activity was found in the supernatant of the control strain C. glutamicum (pEKEx2). From these results we conclude that we have succeeded in the secretion of enzymatically active and therefore FAD cofactor-containing SoXy into the culture supernatant of C. glutamicum.Finally, we examined whether the secretion of SoXy had in fact occurred via the Tat pathway of C. glutamicum. Plasmid pTorA–SoXy was used to transform C. glutamcium ATCC13032 wild type and a C. glutamicum ΔTatAC mutant strain that lacks two essential components of the Tat transport machinery and therefore does not possess a functional Tat translocase (Meissner et al., 2007). The corresponding cells were grown in BHI medium (Difco) at 30°C in the presence of 1 mM IPTG for 6 h. Subsequently, the proteins present in the cellular and the supernatant fractions of the corresponding cells were analysed by SDS-PAGE followed by Western blotting using SoXy-specific antibodies. As shown in Fig. 3, polypeptides corresponding to the unprocessed TorA–SoXy precursor and some minor smaller degradation products of it can be detected in the cellular fractions of both the wild-type and the ΔTatAC deletion strains (lanes 3 and 5). In the supernatant fraction of the Tat⁺ wild-type strain (lane 4), but not that of the ΔTatAC strain (lane 6), a polypeptide corresponding to mature SoXy is present, clearly showing that export of SoXy in the wild-type strain had occurred in a strictly Tat-dependent manner. Another noteworthy finding is the observation that hardly any mature SoXy protein accumulated in the cellular fraction of the Tat⁺ wild-type strain (lane 3), indicating that SoXy is, after its Tat-dependent translocation across the cytoplasmic membrane and processing by signal peptidase, rapidly transported out of the intermembrane space across the mycolic acid-containing outer membrane into the supernatant. However, the mechanism of how proteins cross this additional permeability barrier is completely unknown so far (Bitter et al., 2009).Open in a separate windowFigure 3Transport of TorA–SoXy occurs in a strictly Tat-dependent manner. Plasmid pTorA–SoXy was transformed into C. glutamcium ATCC13032 (Tat⁺) and a C. glutamicum ΔTatAC mutant that lacks a functional Tat translocase (Meissner et al., 2007). As a control, the empty pEKEx2 expression vector was transformed into C. glutamicum ATCC13032 (Tat⁺). The respective strains were grown overnight in 5 ml of BHI medium (Difco) at 30°C. The cells were washed once with BHI and resuspended in 20 ml of fresh BHI medium containing 1 mM IPTG. After 6 h of further growth at 30°C, the cellular (C) and supernatant (S) fractions were prepared as described previously (Meissner et al., 2007). Samples of the C and S fractions were subjected to SDS-PAGE followed by immunoblotting using anti-SoXy antibodies as indicated at the top of the figure. Lanes 1 and 2: C. glutamicum ATCC13032 (pEKEx2); lanes 3 and 4: C. glutamicum ATCC13032 (pTorA–SoXy); lanes 5 and 6: C. glutamicum ΔTatAC (pTorA–SoXy). Asterisk: TorA–SoXy precursor; arrow: secreted SoXy protein. The positions of molecular mass markers (kDa) are indicated at the left margin of the figure.To the best of our knowledge, our results represent the first documented example of the successful secretion of a normally cytosolic, cofactor-containing protein via the Tat pathway in an active form into the culture supernatant of a recombinant expression host. Our results clearly show that, also for this biotechnologically very interesting class of proteins, a secretory production strategy can be a promising alternative to conventional intracellular expression strategies. Besides for SoXy and other FAD-containing carbohydrate oxidases, for which various applications are perceived by industry such as the in situ generation of hydrogen peroxide for bleaching and disinfection performance in technical applications, their use in the food and drink industry, as well as their use in diagnostic applications and carbohydrate biosynthesis processes (Oda and Hiraga, 1998; Murooka and Yamashita, 2001; van Hellemond et al., 2006; Heuts et al., 2007), a secretory production strategy might now be an attractive option also for biotechnologically relevant enzymes that are used as biocatalysts in chemo-enzymatic syntheses and that possess cofactors other than FAD, such as pyridoxal-5′-phosphate (PLP)-dependent ω-transaminases (Mathew and Yun, 2012) or various thiamin diphosphate (TDP)-dependent enzymes (Müller et al., 2009).     

Lipids in cell biology: how can we understand them better?

Lipids are a major class of biological molecules and play many key roles in different processes. The diversity of lipids is on the same order of magnitude as that of proteins: cells express tens of thousands of different lipids and hundreds of proteins to regulate their metabolism and transport. Despite their clear importance and essential functions, lipids have not been as well studied as proteins. We discuss here some of the reasons why it has been challenging to study lipids and outline technological developments that are allowing us to begin lifting lipids out of their “Cinderella” status. We focus on recent advances in lipid identification, visualization, and investigation of their biophysics and perturbations and suggest that the field has sufficiently advanced to encourage broader investigation into these intriguing molecules.Lipids are fundamental building blocks of all cells and play many important and varied roles. They are key components of the plasma membrane and other cellular compartments, including the nuclear membrane, the endoplasmic reticulum, the Golgi apparatus, and trafficking vesicles such as endosomes and lysosomes. The lipid composition of different organelles, cell types, and ultimately tissues can vary substantially, suggesting that different lipids are required for different functions (Saghatelian et al., 2006 ; Klose et al., 2013 ). Mammalian cells express tens of thousands of different lipid species and use hundreds of proteins to synthesize, metabolize, and transport them. Although the complexity and diversity of lipids approach those of proteins, we have a much poorer understanding of their functions, making lipids in many ways the “Cinderellas” of cell biology. We discuss here recent advances, and challenges, in investigating how these molecules contribute to the many biological processes in which they participate.Like proteins, lipids can have structural (e.g., by stabilizing different membrane curvatures) or signaling roles. Posttranslational lipidation of proteins (e.g., palmitoylation or farnesylation) and carbohydrate-linked lipids (glycolipids) are also important examples of cellular lipid pools. It is clear that higher-order organization is key to most lipid functions, and they are believed to assemble into signaling platforms that contain both lipids and proteins (Kusumi et al., 2012 ). Some lipid microdomains are termed lipid rafts, and much ongoing effort is being focused on defining their parameters (Simons and Sampaio, 2011 ; Suzuki et al., 2012 ; Klotzsch and Schutz, 2013 ). Because of this focus on lipid rafts, we have a better understanding of the properties of proposed raft lipids (e.g., sphingomyelins and cholesterol) than of many other lipid species. Much of what we know about lipids has come from studying synthetic membranes with specific lipid compositions. Although model membranes usually consist of very few (<10) lipid species, they have been useful for understanding the biophysical properties of lipids. At the other end of the spectrum, lipids have been implicated in various diseases, with cholesterol metabolism being a prominent example. We are now getting to a point at which we can address the roles of lipids in the middle of the spectrum—in cells.     

Molecular basis of cross‐species ACE2 interactions with SARS‐CoV‐2‐like viruses of pangolin origin

Correction to: The EMBO Journal (2021) 40: e107786. DOI 10.15252/embj.2021107786 | Published online 8 June 2021The authors would like to add three references to the paper: Starr et al and Zahradník et al also reported that the Q498H or Q498R mutation has enhanced binding affinity to ACE2; and Liu et al reported on the binding of bat coronavirus to ACE2.Starr et al and Zahradník et al have now been cited in the Discussion section, and the following sentence has been corrected from:“According to our data, the SARS‐CoV‐2 RBD with Q498H increases the binding strength to hACE2 by 5‐fold, suggesting the Q498H mutant is more ready to interact with human receptor than the wildtype and highlighting the necessity for more strict control of virus and virus‐infected animals”.to“Here, according to our data and two recently published papers, the SARS‐CoV‐2 RBD with Q498H or Q498R increases the binding strength to hACE2 (Starr et al, 2020; Zahradník et al, 2021), suggesting the mutant with Q498H or Q498R is more ready to interact with human receptor than the wild type and highlighting the necessity for more strict control of virus and virus‐infected animals”.The Liu et al citation has been added to the following sentence:“In another paper published by our group recently, RaTG13 RBD was found to bind to hACE2 with much lower binding affinity than SARS‐CoV‐2 though RaTG13 displays the highest whole‐genome sequence identity (96.2%) with the SARS‐CoV‐2 (Liu et al, 2021)”.Additionally, the authors have added the GISAID accession IDs to the sequence names of the SARS‐CoV‐2 in two human samples (Discussion section). To make identification unambiguous, the sequence names have been updated from “SA‐lsf‐27 and SA‐lsf‐37” to “GISAID accession ID: EPI_ISL_672581 and EPI_ISL_672589”.Lastly, the authors declare in the Materials and Methods section that all experiments employed SARS‐CoV‐2 pseudovirus in cultured cells. These experiments were performed in a BSL‐2‐level laboratory and approved by Science and Technology Conditions Platform Office, Institute of Microbiology, Chinese Academy of Sciences.These changes are herewith incorporated into the paper.     

Genome-Wide Association Studies and the Problem of Relatedness Among Advanced Intercross Lines and Other Highly Recombinant Populations

Model organisms offer many advantages for the genetic analysis of complex traits. However, identification of specific genes is often hampered by a lack of recombination between the genomes of inbred progenitors. Recently, genome-wide association studies (GWAS) in humans have offered gene-level mapping resolution that is possible because of the large number of accumulated recombinations among unrelated human subjects. To obtain analogous improvements in mapping resolution in mice, we used a 34th generation advanced intercross line (AIL) derived from two inbred strains (SM/J and LG/J). We used simulations to show that familial relationships among subjects must be accounted for when analyzing these data; we then used a mixed model that included polygenic effects to address this problem in our own analysis. Using a combination of F₂ and AIL mice derived from the same inbred progenitors, we identified genome-wide significant, subcentimorgan loci that were associated with methamphetamine sensitivity, (e.g., chromosome 18; LOD = 10.5) and non-drug-induced locomotor activity (e.g., chromosome 8; LOD = 18.9). The 2-LOD support interval for the former locus contains no known genes while the latter contains only one gene (Csmd1). This approach is broadly applicable in terms of phenotypes and model organisms and allows GWAS to be performed in multigenerational crosses between and among inbred strains where familial relatedness is often unavoidable.SUSCEPTIBILITY to diseases such as drug abuse is partially determined by genetic factors. The identification of the alleles that underlie disease susceptibility is an immensely important goal that promises to revolutionize both the diagnosis and the treatment of human disease. Genome-wide association studies (GWAS) in humans can locate common alleles with great precision. However, GWAS may be unable to identify the bulk of the heritable variability for common genetic diseases; some of this “missing heritability” is thought to be due to rare alleles (Manolio et al. 2009). Model organisms are complementary to human genetic studies and offer unique advantages including the ability to control the environment, perform dangerous or invasive procedures, and test hypotheses by manipulating genes via genetic engineering; a final advantage is that crosses between two inbred strains avoid many of the difficulties associated with rare alleles.Studies in model organisms have frequently employed intercrosses (F₂''s) to identify quantitative trait loci (QTL) that underlie phenotypic variability. F₂ crosses are easy to produce and easy to analyze; however, due to a lack of recombination they can identify only larger genomic regions and are thus unsuitable for identifying the genes that cause QTL (Flint et al. 2005; Peters et al. 2007). This is a serious limitation that can be addressed by using populations with greater numbers of accumulated recombinations. Darvasi and Soller (1995) suggested the creation of advanced intercross lines (AILs) by successive generations of random mating after the F₂ generation to produce additional recombinations. An AIL offers vastly improved mapping resolution while maintaining the desirable property that all polymorphic alleles are common.We used an AIL to study sensitivity to methamphetamine, which is a genetically complex trait that may be useful for identifying genetic factors influencing the subjectively euphoric response to stimulant drugs and susceptibility to drug abuse (Palmer et al. 2005; Phillips et al. 2008; Bryant et al. 2009). For example, a prior study suggested that the gene Casein Kinase 1 Epsilon (Csnk1e) might influence sensitivity to the acute locomotor response to methamphetamine in mice (Palmer et al. 2005). This conclusion has been bolstered by additional pharmacological (Bryant et al. 2009) and genetic studies. In addition, we have shown that polymorphisms in this gene are associated with sensitivity to the euphoric effects of amphetamine in humans (Veenstra-Vanderweele et al. 2006). Another group has subsequently reported that this same gene is associated with heroine addiction (Levran et al. 2008). Thus, genes that modulate the acute locomotor response to a drug in mice may also be important for sensitivity to similar drugs in humans as well as the risk for developing drug abuse.The purpose of this study was to develop a framework for rapid identification of high precision QTL and ideally specific genes that influence sensitivity to methamphetamine in mice by employing an AIL. We produced an F₂ cross (n = 490) and a corresponding 34th generation AIL (n = 688) derived from the inbred strains SM/J and LG/J. This allowed us to compare and integrate the results from F₂ and AIL mice. We examined the locomotor stimulant response to a 2-mg/kg dose of methamphetamine, which is extremely disparate in the two progenitor strains. We performed a GWAS using either simple regression, which ignored relatedness, or a mixed model that accounted for relatedness by using identity coefficients that were calculated from the pedigree. We also explored two methods to estimate significance: simple permutation and gene dropping. We discuss the performance of a mixed model that includes polygenic effects vs. simple regression and the performance of permutation vs. gene dropping. The methods used in this study are applicable to a variety of other phenotypes and populations.     

How a first research experience had an impact on my scientific journey

As I look back to my scientific trajectory on the occasion of being the recipient of the E. B. Wilson Medal of the American Society for Cell Biology, I realize how much an early scientific experience had an impact on my research many years later. The major influence that the first scientific encounters can have in defining a scientist’s path makes the choice of the training environment so important for a future career.

During my scientific career I have worked on different topics. However, my scientific journey represents a progression of steps that build onto each other. Each experience, no matter how successful, had an impact on how I interpreted subsequent results, on how I assessed their implications, and on how I made subsequent programmatic decisions. My research at the interface of cell biology and neuroscience has spanned from membrane traffic and membrane remodeling, to phosphoinositide signaling, and more recently also to the role of membrane contact sites in the control of membrane lipid homeostasis in physiology and disease. While only partially related, all these topics are interconnected by an intellectual thread. As an example—which shows how defining our training years can be—I will summarize here how an early scientific experience in medical school contributed to shape in unexpected ways my research directions many years later.From the day I enrolled in medical school at the University of Milano I knew I wanted to be a scientist. While a medical student, I explored several research environments that would expose me to leading-edge science. Eventually I encountered a young scientist, Jacopo Meldolesi, who had just returned to Italy after studying the secretory pathway as a postdoc with Jim Jamieson and George Palade at Rockefeller University, and I decided to join his lab to carry out the required short medical school thesis project. Jacopo was interested in the regulation of secretion and proposed that I explore mechanisms underlying the so-called “PI (Phosphatidyl Inositol) effect” described by Hokin and Hokin in the 1950s (Hokin and Hokin, 1953). These investigators had found that stimulation of secretory cells resulted in the rapid cleavage of PI immediately followed by its resynthesis. It remained unclear whether these changes reflected a signaling reaction elicited by secretagogues or metabolic changes associated with membrane traffic reactions triggered by the stimulus. So, the idea was to determine in which membrane compartment this turnover occurred: selectively at the plasma membrane (thus supporting a role in signaling) or in membranes of the secretory pathway. We were puzzled by finding that at least under the rather poor time resolution of our experiments—incubation of pancreatic lobules for tens of minutes with radioactive inositol following stimulation with acetylcholine—newly synthesized PI was found at similar concentrations in all membranes. But then we became aware of the recently described proteins that transport lipids between membranes through the cytosol and independent of membrane traffic (Wirtz and Zilversmit, 1969). We thought that perhaps these proteins could contribute to rapidly equilibrating PI pools in different compartments, questioning the validity of our experimental approach to identify the site of PI resynthesis. I moved on to other projects, and I thought I would never again work with lipids.In the following years the work of several labs established that the PI effect discovered by Hokin and Hokin mainly reflected agonist-stimulated phospholipase C (PLC)-dependent cleavage of the phosphoinositide PI(4,5)P₂ at the plasma membrane to generate the second messengers diacylglycerol (DAG) and IP3 (phosphoinositides are the phosphorylated products of PI) (Lapetina and Michell, 1973; Inoue et al., 1977; Berridge et al., 1983). These findings became textbook knowledge, and the idea that PI(4,5)P₂ had primarily a function in signal transduction at the cell surface became mainstream. The additional discovery of the PI(3,4)P₂ and PI(3,4,5)P₃ as a mediators of growth factor actions added a new twist to the field by showing that not only metabolites generated by PI(4,5)P₂ cleavage, but also phosphoinositides themselves, could have a signaling role (Traynor-Kaplan et al., 1988; Whitman et al., 1988; Auger et al., 1989). However, once again, this signaling appeared to be implicated in classical signal transduction, not in membrane traffic.In the meantime, I had moved to Yale, first as a postdoc with Paul Greengard and then as a faculty in the Section of Cell Biology (which subsequently became the Department of Cell Biology). Following my postdoc with Greengard, I had become interested in the cell biology of synapses and in the molecular machinery underlying the exo-endocytic recycling of synaptic vesicles. In what turned out to be a seminal finding, in the mid-1990s Peter McPherson, a postdoc in my laboratory, identified a nerve terminal–enriched protein that shared some interactions with the GTPase dynamin, another abundant presynaptic protein studied in our lab (McPherson et al., 1994). As dynamin mediates the fission reaction of endocytosis by cutting the neck of endocytic buds to generate free vesicles (Takei et al., 1995; Roux et al., 2006; Ferguson and De Camilli, 2012), we speculated that this new protein would also have a role in endocytic traffic. Peter set out to clone it, and I vividly recall my surprise when, upon my returning from a summer vacation in Italy he reported to me that a domain of this protein, which we called synaptojanin, had homology to an inositol 5-phosphatase that dephosphorylates PI(4,5)P₂ (McPherson et al., 1996). A second domain of this protein was subsequently found to have inositol 4-phosphatase activity (Guo et al., 1999), so that synaptojanin can dephosphorylate PI(4,5)P₂ all the way to PI. Thus, I was back to lipids, specifically to inositol phospholipids, and to a potential role of these lipids not in the transduction of extracellular signals, but in membrane traffic. With a mind primed by my project in medical school, I delved with enthusiasm into this topic. PI(4,5)P₂ had been shown to interact with clathrin adaptors (Beck and Keen, 1991), but the physiological significance of these findings had remained unclear. Our studies led to a model in which the selective concentration of PI(4,5)P₂ in the plasma membrane is required for the recruitment of clathrin adaptors and other endocytic factors to this membrane, while the tight coupling of PI(4,5)P₂ dephosphorylation by synaptojanin to the dynamin-dependent fission reaction of endocytosis is required to allow the shedding of such factors once the vesicle has undergone separation from the plasma membrane (Cremona et al., 1999; Cremona and De Camilli, 2001; Wenk et al., 2001). We also discovered curvature-generating/sensing proteins (BAR domain–containing proteins, primarily endophilin) that are responsible for achieving this coupling by coordinating the formation of the narrow neck of endocytic buds with the recruitment of synaptojanin (Takei et al., 1999; Farsad et al., 2001; Frost et al., 2009; Wu et al., 2010; Milosevic et al., 2011). This model, which is encapsulated in Figure 1 and which is supported by genetic studies in mice and other model organisms, posits that a cycle of PI phosphorylation and dephosphorylation is nested within the synaptic vesicle cycle. The implication of a phosphoinositide phosphatase in membrane transport converged with the identification of a PI kinase (the PI 3-kinase Vps34) involved in “vacuolar protein sorting” in yeast (Schu et al., 1993). This raised the possibility that the phosphorylation and dephosphorylation of inositol phospholipids could have a general significance in membrane traffic (De Camilli et al., 1996).Open in a separate windowFIGURE 1:Schematic cartoon depicting the occurrence of a cycle of phosphatidylinositol phosphorylation–dephosphorylation nested within the exo-endocytic recycling of synaptic vesicles at synapses. PI(4,5)P₂ in the plasma membrane helps define this membrane as the acceptor for the fusion of synaptic vesicles and functions as a coreceptor for endocytic factors responsible for their reinternalization. PI(4,5)P₂ dephosphorylation by synaptojanin allows the shedding of endocytic factors and the reutilization of vesicles for a new cycle of secretion. Recruitment of synaptojanin is coupled to dynamin-dependent fission via its binding to the BAR domain–containing protein endophilin, a curvature-generating/sensing protein that also binds dynamin.The field grew very rapidly. It was found that reversible phosphorylation of phosphatidylinositol at the 3,4 and 5 positions of the inositol ring by a multiplicity of enzymes generates seven phosphoinositide species whose differential protein-binding specificities help control, often via a dual key mechanism (Wenk and De Camilli, 2004), membrane cytosol interfaces (Di Paolo and De Camilli, 2006; Vicinanza et al., 2008; Balla, 2013). Along with small GTPases (Behnia and Munro, 2005), phosphoinositides are key determinants of the identity of membranes or membrane subdomains and thus control the multiplicity of reactions that occur at membrane–cytosol interfaces, such as the recruitment and shedding of trafficking proteins, the function of integral membrane proteins, and the assembly and disassembly of signaling and cytoskeleton complexes. Importantly, it became clear that, as we had found for synaptojanin, interconversion of phosphoinositide species along the secretory and endocytic pathways functions as a switch to release membrane-associated factors that define one compartment and to recruit factors that define the next compartment (Odorizzi et al., 2000; Di Paolo and De Camilli, 2006; Zoncu et al., 2009). I think I would not have immediately grasped the significance of the identification of synaptojanin and then invested much of our work in this field had it not been for my early research experience in medical school.The roles of phosphoinositides in membrane identity imply the existence of mechanisms to tightly control their localizations and also to ensure availability of PI in membranes—primarily the plasma membrane—where the backbone of this lipid is consumed by PLC. Surprisingly, genetic evidence suggested that one mechanism to control PI4P levels in membranes is via their direct contacts with the endoplasmic reticulum (ER), where a main PI4P phosphatase is localized (Foti et al., 2001; Stefan et al., 2011; Mesmin et al., 2013). Concerning availability of PI, DAG generated by PLC and its phosphorylated downstream product phosphatidic acid must be returned to the ER for metabolic recycling by the ER-localized phosphatidylinositol synthase, and newly synthesized PI then needs to be rapidly transported to the plasma membrane to replace its depleted pool. Strong evidence indicates that these reactions are mediated at least in part by lipid transfer proteins, and we have now learned that much of this protein-mediated lipid transport occurs at membrane contact sites (Mesmin et al., 2013; Wong et al., 2017; Cockcroft and Raghu, 2018; Pemberton et al., 2020). This is what motivated my lab to enter the young field of membrane contact sites. Here, once again, my early appreciation of the importance of lipid transfer proteins during medical school had an impact on my decision to invest in this rapidly developing area of cell biology.Not only has it been rewarding to help expand the inventory of proteins known to function both as membrane tethers and as lipid transporters, and to elucidate their function and regulation (Giordano et al., 2013; Schauder et al., 2014; Chung et al., 2015; Dong et al., 2016; Saheki et al., 2016), but this has been an opportunity, in collaboration with my Yale colleague Karin Reinisch, to participate in the discovery of something unexpected and new in the biology of eukaryotic cells. Until recently, it was thought that nonvesicular lipid transport, including the transport occurring at membrane contact sites, occurred only via protein modules that function as shuttles between two bilayers, often acting as countertransporters, delivering different cargoes as they move back and forth between two closely apposed membranes. However, our investigations of VPS13, a protein originally identified in yeast for its function in membrane transport and subsequently linked to lipid dynamics (Lang et al., 2015; Park et al., 2016), raised the possibility that this protein, and thus also the closely related autophagy factor ATG2, may be the founding members of a protein superfamily with bulk lipid transport properties, that is, proteins that could facilitate the net fluxes of lipids between adjacent membranes, and thus in membrane expansion (Kumar et al., 2018). As was subsequently shown by structural studies (Valverde et al., 2019; Li et al., 2020), these rod-like proteins harbor a hydrophobic groove that spans their length, thus supporting a bridge-like model of lipid transport to support bulk lipid flow (Li et al., 2020; Guillen-Samander et al., 2021; Leonzino et al., 2021). Inspection of protein sequence and structure databases suggests the existence of several other proteins with these properties, implying that this mode of lipid transport (a bridge-like mechanism), previously described for the transport of lipids from the inner to the outer membrane of Gram-negative bacteria (Bishop, 2019), may be of broad relevance in cell biology. As mutations in VPS13 family proteins result in neurodegenerative diseases, including a Huntington-like disease and Parkinson’s disease (Ugur et al., 2020), this is a field rich in medical implications. In fact, my curiosity about disease mechanisms stemming from my medical school training contributed to making me specially interested in studying these proteins.It has been a wonderful journey, rich with twists and turns, exploring new territories and continuously discovering that there are untapped frontiers whose existence we still do not know about. During the course of my career, I found especially rewarding bringing together different fields and finding new connections between fundamental cell biology and medicine, more so in recent years, where genetic studies of diseases often provide the first clues to molecular and cellular mechanisms. As I have emphasized in this essay, each of our experiences, no matter how successful, becomes part of the body of knowledge that defines us and has an impact on our research trajectory. I typically encourage students to search for training opportunities in different settings rather than follow a linear path, as this will enable them to bring together in their independent careers the expertise and knowledge of different worlds and thus to carry out original science. Our specific and unique paths are what make our research innovative and special.     

Induction of Robust Cellular and Humoral Virus-Specific Adaptive Immune Responses in Human Immunodeficiency Virus-Infected Humanized BLT Mice

The generation of humanized BLT mice by the cotransplantation of human fetal thymus and liver tissues and CD34⁺ fetal liver cells into nonobese diabetic/severe combined immunodeficiency mice allows for the long-term reconstitution of a functional human immune system, with human T cells, B cells, dendritic cells, and monocytes/macrophages repopulating mouse tissues. Here, we show that humanized BLT mice sustained high-level disseminated human immunodeficiency virus (HIV) infection, resulting in CD4⁺ T-cell depletion and generalized immune activation. Following infection, HIV-specific humoral responses were present in all mice by 3 months, and HIV-specific CD4⁺ and CD8⁺ T-cell responses were detected in the majority of mice tested after 9 weeks of infection. Despite robust HIV-specific responses, however, viral loads remained elevated in infected BLT mice, raising the possibility that these responses are dysfunctional. The increased T-cell expression of the negative costimulator PD-1 recently has been postulated to contribute to T-cell dysfunction in chronic HIV infection. As seen in human infection, both CD4⁺ and CD8⁺ T cells demonstrated increased PD-1 expression in HIV-infected BLT mice, and PD-1 levels in these cells correlated positively with viral load and inversely with CD4⁺ cell levels. The ability of humanized BLT mice to generate both cellular and humoral immune responses to HIV will allow the further investigation of human HIV-specific immune responses in vivo and suggests that these mice are able to provide a platform to assess candidate HIV vaccines and other immunotherapeutic strategies.An ideal animal model of human immunodeficiency virus (HIV) infection remains elusive. Nonhuman primates that are susceptible to HIV infection typically do not develop immunodeficiency (63), and although the simian immunodeficiency virus (SIV) infection of rhesus macaques has provided many critically important insights into retroviral pathogenesis (30), biological and financial considerations have created some limitations to the wide dissemination of this model. The great need for an improved animal model of HIV itself recently has been underscored by the disappointing results of human trials of MRKAd5, an adenovirus-based HIV type 1 (HIV-1) vaccine. This vaccine was not effective and actually may have increased some subjects'' risk of acquiring HIV (53). In the wake of these disappointing results, there has been increased interest in humanized mouse models of HIV infection (54). The ability of humanized mouse models to test candidate vaccines or other immunomodulatory strategies will depend critically on the ability of these mice to generate robust anti-HIV human immune responses.Mice have provided important model systems for the study of many human diseases, but they are unable to support productive HIV infection, even when made to express human coreceptors for the virus (7, 37, 52). A more successful strategy to humanize mice has been to engraft human immune cells and/or tissues into immunodeficient severe combined immunodeficiency (SCID) or nonobese diabetic (NOD)/SCID mice that are unable to reject xenogeneic grafts (39, 42, 57). Early versions of humanized mice supported productive HIV infection and allowed investigators to begin to address important questions in HIV biology in vivo (23, 40, 43-45). More recently, human cord blood or fetal liver CD34⁺ cells have been used to reconstitute Rag2^−/− interleukin-2 receptor γ chain-deficient (γ_c^−/−) and NOD/SCID/γ_c^−/− mice, resulting in higher levels of sustained human immune cell engraftment (27, 29, 61). These mice have allowed for stable, disseminated HIV infection (2, 4, 24, 65, 67), including mucosal transmission via vaginal and rectal routes (3). These mice recently have been used to demonstrate an important role for Treg cells in acute HIV infection (29) and to demonstrate that the T-cell-specific delivery of antiviral small interfering RNA is able to suppress HIV replication in vivo (31). These mice also have demonstrated some evidence of adaptive human immune responses, including the generation of HIV-specific antibody responses in some infected mice (2, 65), and some evidence of humoral and cell-mediated responses to non-HIV antigens or pathogens (24, 61). Most impressively, Rag2^−/− γ_c^−/− mice reconstituted with human fetal liver-derived CD34⁺ cells have generated humoral responses to dengue virus infection that demonstrated both class switching and neutralizing capacity (32). In spite of these advances, however, these models have not yet been reported to generate de novo HIV-specific cell-mediated immune responses, which are considered to be a crucial arm of host defense against HIV infection in humans.In contrast to humanized mouse models in which only human hematopoietic cells are transferred into immunodeficient mice, the surgical implantation of human fetal thymic and liver tissue has been performed in addition to the transfer of human hematopoietic stem cells (HSC) to generate mice in which human T cells are educated by autologous human thymic tissue rather than by the xenogeneic mouse thymus. Melkus and colleagues refer to mice they have reconstituted in this way as NOD/SCID-hu BLT (for bone marrow, liver, and thymus), or simply BLT, mice (41). We previously referred to mice that we have humanized in a similar way as NOD/SCID mice cotransplanted with human fetal thymic and liver tissues (Thy/Liv) and CD34⁺ fetal liver cells (FLC) (33, 60) but now adopt the designation BLT mice as well. BLT mice demonstrate the robust repopulation of mouse lymphoid tissues with functional human T lymphocytes (33, 41, 60) and can support the rectal and vaginal transmission of HIV (13, 59). Further, BLT mice demonstrate antigen-specific human immune responses against non-HIV antigens and/or pathogens (41, 60). The ability of these mice to generate human immune responses against HIV, however, has not yet been reported. In this study, we investigated whether the provision of autologous human thymic tissue in BLT mice generated by the cotransplantion of human fetal Thy/Liv tissues and CD34⁺ FLC would allow for the maturation of human T cells in humanized mice capable of providing improved cellular responses to HIV as well as providing adequate help for improved humoral responses. To describe the cells contributing to human immune responses in BLT mice, we also characterized the phenotypes of multiple subsets of T cells, B cells, dendritic cells (DCs), and monocytes/macrophages present in uninfected humanized mice. The generation of robust HIV-directed human cellular and humoral immune responses in these mice would further demonstrate the ability of humanized mice to provide a much needed platform for the evaluation of HIV vaccines and other novel immunomodulatory strategies.     

Discovery of keratin function and role in genetic diseases: the year that 1991 was

In 1991, a set of transgenic mouse studies took the fields of cell biology and dermatology by storm in providing the first credible evidence that keratin intermediate filaments play a unique and essential role in the structural and mechanical support in keratinocytes of the epidermis. Moreover, these studies intimated that mutations altering the primary structure and function of keratin filaments underlie genetic diseases typified by cellular fragility. This Retrospective on how these studies came to be is offered as a means to highlight the 25th anniversary of these discoveries.Although intermediate filaments (IFs) have been characterized at some level for a longer period of time (Oshima, 2007 ), they were officially discovered as such as recently as 1968 by Howard Holtzer and colleagues while studying the developing skeletal muscle (Ishikawa et al., 1968 ). The advent of gene cloning methods and monospecific antibody production in the late 1970s and throughout the 1980s led to an explosion of data and knowledge about IFs that established them as a large family of genes and proteins that are individually regulated in a tight and evolutionarily conserved tissue- and differentiation-specific manner. Researchers also uncovered some of the remarkable properties of IFs as purified elements in vitro and in living systems and recognized that they occur in the nucleus as well as in the cytoplasm. In spite of the fast pace of progress during that period, however, it was not possible to produce evidence that spoke unequivocally about the functional importance of IFs in cells and tissues, let alone their role in disease.Beginning in the mid- to late 1980s, pioneering experimentation along two distinct lines was underway in the laboratory of Elaine Fuchs, then at the University of Chicago. The eventual merger of these approaches yielded the first formal insight into IF function in vivo, as well as into their direct involvement in human disease. In an effort to define structure–function relationships with regard to the assembly and network formation properties of IFs, one such approach was the application of systematic deletion mutagenesis to keratin 14 (K14), a type I IF that is expressed with its type II partner keratin 5 (K5) in the progenitor basal layer of the epidermis and related complex epithelia. These studies demonstrated that deleting sequences from either end of the central α-helical rod domain of the K14 protein was deleterious for filament formation in a dominant manner both in transfected cells (Albers and Fuchs, 1987 , 1989 ; Figure 1) and the setting of IF polymerization assays involving purified proteins in vitro (e.g., Coulombe et al., 1990 ). The second key effort in the Fuchs lab in the late 1980s resulted in the demonstration that the proximal 2.5 kb and distal 700 base pairs corresponding respectively to the 5′ upstream and 3′ downstream regions of the cloned human K14 gene were sufficient to confer tissue-specific, that is, K14-like, regulation in transgenic mice in vivo (Vassar et al., 1989 ; Figure 1). This tour de force paved the way for the production of a human K14 gene promoter–based cassette (e.g., Saitou et al., 1995 ) that could reliably direct the expression of any open reading frame in a K14-like manner in transgenic mice. As an aside, this tool has had a profound effect on epithelial and skin biology research.Open in a separate windowFIGURE 1:Schematic representation of the strategy and outcome of the experiments that led to the discovery of keratin function and role in genetic disease. Original figures are reproduced to give a realistic account of the data. (A) Examples of a disrupted keratin filament network in cultured epithelial cells transfected with and expressing a dominantly acting K14 deletion mutant (arrows). (Reproduced from Albers and Fuchs, 1987 , with permission.) (B) Preferential expression of a substance P-epitope–tagged transgenic human K14 protein in the basal layer of tail skin epidermis in mouse, conveying the tissue- and differentiation-specific behavior of the transgene. (Reproduced from Vassar et al., 1989 , with permission.) (C) The two experimental approaches described in A and B were combined to assess the consequences of tissue-specific expression of dominantly acting K14 mutants in skin tissue in vivo. (D) Newborn mouse littermates. The mouse at the top is transgenic (Tg) and expresses a mutated form of K14 in the epidermis. It is showing severe skin blistering (arrows), particularly in its front paws, which are heavily used by mouse newborns to feed from their mother. The bottom mouse is a nontransgenic control showing no such blistering. (E, F) Hematoxylin-eosin–stained skin tissue sections showing the location of subepidermal cleavage within the epidermis of a K14 mutant–expressing transgenic mouse (opposing arrows in E). Cleavage occurs at the level of the basal layer, where the mutant keratin is expressed. Again, this is never seen in control wild-type (Wt) skin (F). Bar, 100 μm (E, F). (D–F are from Coulombe et al., 1991b , with permission.) (G) Leg skin in a patient suffering from the Dowling–Meara form of epidermolysis bullosa simplex. Characteristic of this severe variant of this disease, several skin blisters are often grouped in a herpetiform manner (Fine et al., 1991 ).Subsequent use of the human K14 promoter–based cassette to direct the expression of epitope-tagged and selected deletion mutants of K14 gave rise to transgenic mouse pups that exhibited extensive blistering of the skin preferentially at sites of frictional trauma (Coulombe et al., 1991b ; Vassar et al., 1991 ; Figure 1). Electron microscopy showed that skin blistering occurred secondary to a loss of the integrity of keratinocytes located in the basal layer of the epidermis, that is, the precise site of mutant K14 protein accumulation. Such blistering did not occur in transgenic mice expressing a full-length version of human K14 modified to carry only an epitope tag at the C-terminus at similar or higher levels (Coulombe et al., 1991b ; Vassar et al., 1991 ). In addition, the severity of skin blistering in mutant K14–expressing transgenic pups could be directly related to the extent to which the mutant protein had been shown to disrupt filament assembly in transfected cell assays and in IF reconstitution assays in vitro. For instance, tissue-specific expression of a K14 mutant that could severely disrupt 10-nm filament assembly was associated with whole-body skin blistering and the untimely death of mouse pups and, from a pathology perspective, with “tonofilament clumping” and a paucity of visible keratin IFs in transgenic basal keratinocytes. By comparison, expression of another K14 mutant with a less deleterious effect on 10-nm IF assembly was compatible with the survival of transgenic mouse pups and resulted in skin blistering largely limited to the front paws in newborn mice together with altered organization of keratin IFs in basal keratinocytes of transgenic epidermis in situ, albeit without tonofilament clumping. This initial set of mouse strains thus revealed the existence of a direct link between the so-called “genotype” (i.e., mutant K14 characteristics) and the skin phenotype (Coulombe et al., 1991b ; Vassar et al., 1991 ; Fuchs and Coulombe, 1992 ). Electrophoretic analyses of protein samples confirmed that the K14 mutant proteins acted dominantly to produce such spectacular phenotypes in transgenic mouse skin. Finally, blistering also occurred in the mutant K14–expressing transgenic mice in other stratified epithelia known both to express K14 and experience trauma, notably in the oral mucosa (Coulombe et al., 1991b ; Vassar et al., 1991 ).It is worth celebrating the 25th anniversary of these pioneering experiments for the following two reasons. First, the study of these mice provided the first formal demonstration that keratin IFs play a fundamentally important role in structural support in surface epithelia such as the epidermis and oral mucosa. Without proper IF support, epidermal keratinocytes are rendered fragile and cannot sustain trivial frictional stress (Coulombe et al., 1991b ; Fuchs and Coulombe, 1992 ). The second reason is the observation that the phenotype of these K14 mutant–expressing mice proved eerily similar to those of individuals afflicted with the disease epidermolysis bullosa simplex (EBS), a rare, dominantly inherited and debilitating skin condition in which the epidermis and oral mucosa undergo blistering after exposure to trivial mechanical trauma. As observed in the mouse model, tissue cleavage had been shown to result from the loss of integrity of keratinocytes located in the basal layer (Fine et al., 1991 ). Further, other researchers had previously reported on anomalies in the organization of keratin IFs in the basal epidermal keratinocytes of EBS patients (Anton-Lamprecht, 1983 ; Ito et al., 1991 ) or in cultures of epidermal keratinocytes established from EBS patients (Kitajima et al., 1989 ). The Fuchs laboratory thus teamed up with Amy Paller, a physician-scientist and pediatric dermatologist with deep expertise in genodermatoses, and mutations were soon discovered in the K14 gene of two independent and sporadic cases of a severe variant of the disease known as Dowling–Meara EBS (Coulombe et al., 1991a ; Figure 1). The two mutations were heterozygous missense alleles that affected the very same codon in K14 (Arg-125) and were correctly predicted at the time to correspond to a mutational hot spot in type I keratin genes. The mutations were shown to dominantly disrupt 10-nm IF assembly in vitro and/or in transfected keratinocytes in culture (Coulombe et al., 1991a ). Soon thereafter, a team led by Ervin Epstein at University of California, San Francisco (San Francisco, CA), reported on the use of classical linkage analysis to uncover a missense mutation in the K14 gene of a small pedigree with Koebner-type EBS, a less severe variant of the disease (Bonifas et al., 1991 ). The next year, Birgit Lane and colleagues (Lane et al., 1992 ) reported on the occurrence of mutations in keratin 5 (K5), the formal type II keratin assembly partner for K14 in vivo, in another instance of Dowling–Meara EBS.In the years since 1991, a role in structural support has been formally demonstrated for all classes of IFs (Coulombe et al., 2009 ), including the nuclear-localized lamins (e.g., Lammerding et al., 2004 ). Moreover, we now know of several hundred independent instances of mutations in either K5 or K14 in the setting of the EBS disease, with the vast majority of those consisting of dominantly acting missense alleles (Szeverenyi et al., 2008 ; Human Intermediate Filament Database, www.interfil.org, maintained at the Centre for Molecular Medicine and Bioinformatics Institute, Singapore). We also learned that, as anticipated, EBS largely represents a loss-of-function phenotype, since K14-null mice (Lloyd et al., 1995 ), K14-null individuals (Chan et al., 1994 ; Rugg et al., 1994 ), and K5-null mice (Peters et al., 2001 ) all exhibit an EBS-like skin-blistering phenotype (Coulombe et al., 2009 ). Mutations such as Arg125Cys in K14 markedly compromise the remarkable mechanical properties of keratin filaments (Ma et al., 2001 ), as well as the steady-state dynamics of keratin filaments in transfected keratinocytes in culture (Werner et al., 2004 ). Finally, mutations affecting the coding sequence of IF genes have been shown to underlie >100 diseases affecting the human population (Omary et al., 2004 ; Szeverenyi et al., 2008 ; www.interfil.org). Consistent with the exquisite tissue- and cell type–specific regulation of IF genes, these diseases collectively affect a myriad of tissues and organs and are relevant to nearly all branches of medicine. These observations attest to the importance and profound effect that the generation and characterization of mutant K14–expressing transgenic mice has had for cell biology, epithelial physiology, dermatology, and medicine.Many thoughts spring to mind when reminiscing about my involvement with this body of work. First, this effort was prescient of the power of team science and, in particular, of the potential effect of close collaborations involving biologists and physician-scientists. I learned a great deal and benefited immensely from working closely with many colleagues on this project, including Bob Vassar, Kathryn Albers, Linda Degenstein, Liz Hutton, Anthony Letai, Amy Paller, and, last but not least, my postdoctoral mentor and the laboratory head, Elaine Fuchs. Second, there is no substitute for elements such as innovation, hard work, perseverance, boldness, accountability, and great leadership. Elaine had the vision and created the exceptional circumstances necessary to make this set of discoveries possible, and, of equal importance, she was an integral part of the day-to-day progress and maturation of the entire project. Finally, as we all know, there is an intangible element of luck involved in discovery research. In this instance, a strong argument can be made that the studies highlighted here may not have had such a deep and defining effect had the effort been devoted to any IF other than the K5–K14 keratin pairing.What are some of the lingering issues regarding this specific topic that preoccupy us still, 25 years later? Two challenges loom particularly large. First, we have yet to achieve a satisfactory understanding of how mutations in keratin proteins can cause disease. This is due in part to the lack of an atomic-level understanding of the core structure of IFs (which has been a tough nut to crack; Lee et al., 2012 ), along with the reality that, for any relevant IF gene, there is a broad variety of disease-associated (mostly missense) mutations that pepper their primary structure (www.interfil.org). Second, we have yet to achieve success toward the treatment of EBS or any IF-based disorder. Disease characteristics such as low incidence, a dominantly inherited character, genetic heterogeneity (e.g., broad mutational landscape), and, in the case of EBS and related conditions, an intrinsically high rate of cell turnover within the main target tissue significantly add to the challenge of devising safe and effective therapeutic strategies (Coulombe et al., 2009 ). Although efforts are still underway to foster progress on these two challenging issues, the field as a whole has made significant progress in uncovering a plethora of noncanonical functions of keratin IFs (Hobbs et al., 2016 ) in addition to understanding their regulation, dynamics, and many remarkable properties.     

Infectious etiology and amyloidosis in Alzheimer's disease: The puzzle continues

Recent studies have renewed the debate on infectious etiology in late-onset Alzheimer''s disease. Bocharova et al. reported that abundant expression of human beta amyloid (Aβ) in the mouse brain (5XFAD animals) failed to protect against acute herpes simplex virus type 1 infection relative to control mice. While this study does not confirm the antiviral actions of Aβ, it neither supports nor disproves the hypothesis that infection with microbial pathogens is the major cause of Alzheimer''s disease.

Alzheimer''s disease (AD) is a devastating and progressive neurodegenerative disorder and the most common form of dementia in the elderly. AD is associated with the pathological deposition of neurofibrillary tau tangles and extracellular amyloid β (Aβ) plaques as well as with chronic neuroinflammation. Numerous studies showed that specific microbial infections, including with herpes simplex virus type 1 (HSV1), Chlamydia pneumoniae, and several types of spirochaete are linked to AD etiology, as these pathogens have been detected in some AD brains, particularly in senile plaques (1). HSV1 is the most common of these and strongly links human pathogens to AD etiology (supported by over 100 publications reporting direct or indirect association), whereas only a handful of reports are contradictory. The following lines of evidence support this connection: (i) the presence of viruses and other microbes in the brain of most elderly individuals; (ii) in AD brains, pathogen signatures like HSV1 DNA specifically colocalize with AD amyloid plaques; (iii) active HSV1 infection can trigger chronic neuroinflammatory responses that lead to herpes simplex encephalitis, known to cause severe damage in brain regions associated with memory and cognitive functions; (iv) circulating levels of anti-HSV antibodies, an indicator of HSV1 reactivation, is positively correlated with AD pathology; (v) HSV1 infection activates neurotoxic pathways and causes an AD-like phenotype in mice, while in clinical studies, treatment with antiherpetic drugs like valacyclovir show improved cognitive functions in patients with AD compared with controls; and (vi) key features of AD pathology are transmissible upon intracerebral injection of AD brain homogenates (1, 2), indicating that microbial infection could represent an important contributor to late-onset AD and supporting “the Aβ antimicrobial protection hypothesis.” This hypothesis proposes that Aβ deposition is an early response against microbial infection, which consequently drives chronic neuroinflammation and neurodegeneration, which is not necessarily contrary to with the established “amyloid hypothesis” that proposes that amyloidosis of Aβ is the root cause of AD (1). Recently, contrasting evidence was provided to continue the debate on the antiviral role of Aβ and the causative role of HSV1 infection in AD and to highlight that chronic neuroinflammation could represent a central mechanism in infectious etiology of late-onset AD (Fig. 1).Open in a separate windowFigure 1Schematic presentation of AD pathogenesis according to the “Aβ antimicrobial protection hypothesis” and the “amyloid β hypothesis.” Chronic microbial infections activate the immune system and lead to sustained inflammatory responses, which allow penetration of microbial pathogens and/or their products to cross the blood–brain barrier. In the brain, they colocalize with Aβ and induce Aβ fibrillation to form senile plaques. Disruption of the blood–brain barrier causes penetration of peripheral inflammatory molecules as well as inflammatory cells into the brain and promotes gliosis. Furthermore, various risk factors, including genetics, stress, sleep, diet, head injury, and aging, influence disease progression. Together, these form a vicious inflammatory response, perpetuated by chronic infections or reactivation, which further stimulate chronic neuroinflammation. This process eventually leads to neuron death and AD pathology. AD, Alzheimer''s disease.Previously, an elegant and high profile study by Eimer et al. (3) reported that human Aβ interacts with viral surface glycoproteins, which mediates Aβ oligomerization and viral entrapment, leading to protection against HSV1 infection in AD mice and 3D human neuronal cell cultures. Others also showed that Aβ amyloidosis protects against bacterial and fungal infections in animal and worm models of AD (4), suggesting that in addition to various genetic, biochemical, and environmental factors, microbial infection could represent one of the possible triggers of amyloidosis in AD neuropathology (Fig. 1).In a recent article in the Journal of Biochemistry, Bocharova et al. (5) questioned the antiviral properties of Aβ and the role of viral infection in AD pathology in contrast to the findings of Eimer et al. (3), demonstrating protective antiherpetic actions mediated by human Aβ-expressing mice (5XFAD mice). Bocharova et al. used three different age groups of 5XFAD mice expressing transgenes for mutant human amyloid precursor protein and human presenilin 1, one of the four core proteins of the gamma secretase complex that generates Aβ from amyloid precursor protein, with five AD-linked mutations. WT control and 5XFAD mice were infected with three different doses of two strains of HSV1 ranging from 5- to 10-fold below or above the dose lethal to 50% of the mice (LD₅₀). Despite the use of different viral strains, doses, and ages of the animals, survival analysis revealed no statistically significant differences between 5XFAD and WT mice, which confirmed the lack of protective effect of the 5XFAD genotype against HSV1-induced encephalitis (5). Bocharova et al. also noted the region-specific or cell-specific tropisms of HSV1 strains that were not affected in 5XFAD mice compared with controls, which suggested that host–pathogen interactions remained unaltered by Aβ overexpression. Also contrary to the study by Eimer et al. (3), Bocharova et al. found no evidence of HSV1-induced Aβ aggregation and Aβ-mediated viral entrapment. This observation was attributed to the microglia transitioning to a chronic reactive stage and subsequently phagocytosing the virus in Aβ aggregate-dense brain regions. Reactive microglia, often seen in human AD brains, might cause chronic neuroinflammation in response to recurrent reactivation or repetitive HSV1 infections (6). This repeated and sustained microglia activation could eventually potentiate deregulated chronic neuroinflammation and development of AD pathology (7).These studies raise obvious fundamental questions, such as what are the targets for antimicrobial actions of Aβ in human? It is noteworthy to mention that mouse models are not natural hosts for HSV1 and hence do not adequately mimic spontaneous viral shedding or recurrent symptomatic diseases in humans (8). Furthermore, antimicrobial proteins show varying activity against different microbial pathogens, and therefore, the lack of protection against HSV1 infection in 5XFAD mice contrasts the strong protective effects against bacterial and fungal infections (5). This could correlate with diverse microbial infections in human brain. The discrepancies between the two studies could also be due to the use of different doses and variations of HSV1 strains. Nevertheless, the current work neither supports nor refutes the hypothesis of the viral etiology of late-onset AD. Indeed, since no protection was found against acute HSV1 infection in 5XFAD mice, it indicates that viral pathogens could even increase the risk of late-onset AD through multiple Aβ-dependent and independent mechanisms.In the future, it would be of interest to examine the antimicrobial actions of Aβ against diverse microbial pathogens, including several types of spirochete, Treponema pallidum, C. pneumoniae, and the protozoan Toxoplasma gondii, implicated in human brain diseases. Certainly, HSV1 is not the sole contributor to late-onset AD, as it is a multifactorial disease with many contributing factors (including other potential pathogens). HSV1 may contribute to a minority of AD cases, and even this etiology may depend on the presence or the absence of other risk factors such as genetic factors (apolipoprotein E4 variant carriers), age, stress, sleep, diet, head injury, cardiovascular disease, and many others (7). Collectively, these contradictory studies question the contributory role of HSV1 to AD development and the antiviral activity of Aβ. Fundamentally, the findings of Bocharova et al. (5) support the notion that chronic infections with viral, bacterial, and fungal pathogens might cause deregulated neuroinflammatory responses, which subsequently increase amyloidosis in the brain and contribute to AD pathogenesis (2, 9, 10). Thus, inflammatory responses against infections might provide the missing link between infectious etiology and late-onset AD, if such a link exists.It is worth mentioning that the infectious etiology in late-onset AD is a puzzle not yet solved, as—at least so far—no specific microbial infection has been conclusively linked to causation of AD in humans. However, interesting data from many laboratories renewed the concern of infectious etiology in late-onset AD, which will provide new opportunities for anti-inflammatory therapy development.     

Engineering therapeutic protein disaggregases

Therapeutic agents are urgently required to cure several common and fatal neurodegenerative disorders caused by protein misfolding and aggregation, including amyotrophic lateral sclerosis (ALS), Parkinson’s disease (PD), and Alzheimer’s disease (AD). Protein disaggregases that reverse protein misfolding and restore proteins to native structure, function, and localization could mitigate neurodegeneration by simultaneously reversing 1) any toxic gain of function of the misfolded form and 2) any loss of function due to misfolding. Potentiated variants of Hsp104, a hexameric AAA+ ATPase and protein disaggregase from yeast, have been engineered to robustly disaggregate misfolded proteins connected with ALS (e.g., TDP-43 and FUS) and PD (e.g., α-synuclein). However, Hsp104 has no metazoan homologue. Metazoa possess protein disaggregase systems distinct from Hsp104, including Hsp110, Hsp70, and Hsp40, as well as HtrA1, which might be harnessed to reverse deleterious protein misfolding. Nevertheless, vicissitudes of aging, environment, or genetics conspire to negate these disaggregase systems in neurodegenerative disease. Thus, engineering potentiated human protein disaggregases or isolating small-molecule enhancers of their activity could yield transformative therapeutics for ALS, PD, and AD.We urgently need to pioneer game-changing solutions to remedy a number of increasingly prevalent and fatal neurodegenerative disorders, including amyotrophic lateral sclerosis (ALS), Parkinson’s disease (PD), and Alzheimer’s disease (AD; Cushman et al., 2010 ; Jackrel and Shorter, 2015 ). These disorders relentlessly erode our morale and economic resources. Aging is the major risk factor for all of these diseases, which threaten public health on a global scale and represent a severe impediment to living longer lives. A number of promising drugs have emerged to treat cancer and heart disease, but, distressingly, this is not the case for these and other neurodegenerative diseases, for which drug pipelines lie dormant and empty. This situation is unacceptable, and an impending healthcare crisis looms worldwide as population demographics inexorably shift toward older age groups.ALS, PD, AD, and related neurodegenerative disorders are unified by a common underlying theme: the misfolding and aggregation of specific proteins (characteristic for each disease) in the CNS (Cushman et al., 2010 ; Eisele et al., 2015 ). Thus, in ALS, typically an RNA-binding protein with a prion-like domain, TDP-43, mislocalizes from the nucleus to cytoplasmic inclusions in degenerating motor neurons (Neumann et al., 2006 ; Gitler and Shorter, 2011 ; King et al., 2012 ; Robberecht and Philips, 2013 ; March et al., 2016 ). In PD, α-synuclein forms toxic soluble oligomers and cytoplasmic aggregates, termed Lewy bodies, in degenerating dopaminergic neurons (Dehay et al., 2015 ). By contrast, in AD, amyloid-β (Aβ) peptides form extracellular plaques and the microtubule-binding protein, tau, forms cytoplasmic neurofibrillary tangles in afflicted brain regions (Jucker and Walker, 2011 ). Typically, these disorders are categorized into ∼80–90% sporadic cases and ∼10–20% familial cases. Familial forms of disease often have clear genetic causes, which might one day be amenable to gene editing via clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 therapeutics if critical safety and ethical concerns can be successfully addressed and respected (Doudna and Charpentier, 2014 ; Baltimore et al., 2015 ; Rahdar et al., 2015 ; Callaway, 2016 ). However, the more common sporadic forms of disease often have no clear underlying genetics, and wild-type proteins misfold (Cushman et al., 2010 ; Jucker and Walker, 2011 ; Robberecht and Philips, 2013 ; Dehay et al., 2015 ). Consequently, therapeutic agents that directly target and safely reverse deleterious protein misfolding are likely to have broad utility (Eisele et al., 2015 ).There are no treatments that directly target the reversal of the protein-misfolding phenomena that underlie these disorders (Jackrel and Shorter, 2015 ). Strategies that directly reverse protein misfolding and restore proteins to native form and function could, in principle, eradicate any severely damaging loss-of-function or toxic gain-of-function phenotypes caused by misfolded conformers (Figure 1; Jackrel and Shorter, 2015 ). Moreover, therapeutic disaggregases would dismantle self-templating amyloid or prion structures, which spread pathology and nucleate formation of neurotoxic, soluble oligomers (Figure 1; Cushman et al., 2010 ; Cohen et al., 2013 ; Guo and Lee, 2014 ; Jackrel and Shorter, 2015 ). My group has endeavored to engineer and evolve Hsp104, a hexameric AAA+ ATPase and protein disaggregase from yeast (DeSantis and Shorter, 2012 ; Sweeny and Shorter, 2015 ), to more effectively disaggregate misfolded proteins connected with various neurodegenerative disorders, including ALS (e.g., TDP-43 and FUS) and PD (e.g., α-synuclein). Although wild-type Hsp104 can disaggregate diverse amyloid and prion conformers, as well as toxic soluble oligomers (Lo Bianco et al., 2008 ; DeSantis et al., 2012 ), its activity against human neurodegenerative disease proteins is suboptimal. Is it even possible to improve on existing Hsp104 disaggregase activity, which has been wrought over the course of millions of years of evolution?Open in a separate windowFIGURE 1:Therapeutic protein disaggregases. Two malicious problems are commonly associated with protein misfolding into disordered aggregates, toxic oligomers, and cross–β amyloid or prion fibrils: 1) a toxic gain of function of the protein in various misfolded states; and 2) a loss of function of the protein in the various misfolded states. These problems can contribute to the etiology of diverse neurodegenerative diseases in a combinatorial or mutually exclusive manner. A therapeutic protein disaggregase would reverse protein misfolding and recover natively folded functional proteins from disordered aggregates, toxic oligomers, and cross–β amyloid or prion fibrils. In this way, any toxic gain of function or toxic loss of function caused by protein misfolding would be simultaneously reversed. Ideally, all toxic misfolded conformers would be purged. Therapeutic protein disaggregases could thus have broad utility for various fatal neurodegenerative diseases.Remarkably, the answer to this question is yes! We used nimble yeast models of neurodegenerative proteinopathies (Outeiro and Lindquist, 2003 ; Gitler, 2008 ; Johnson et al., 2008 ; Sun et al., 2011 ; Khurana et al., 2015 ) as a platform to isolate enhanced disaggregases from large libraries of Hsp104 variants generated by error-prone PCR (Jackrel et al., 2014b ). In this way, we reprogrammed Hsp104 to yield the first disaggregases that reverse TDP-43, FUS (another RNA-binding protein with a prion-like domain connected to ALS), and α-synuclein (connected to PD) aggregation and proteotoxicity (Jackrel et al., 2014a , 2015 ; Jackrel and Shorter, 2015 ; Torrente et al., 2016 ). Remarkably, a therapeutic gain of Hsp104 function could be elicited by a single missense mutation (Jackrel et al., 2014a , 2015 ; Jackrel and Shorter, 2015 ). Under conditions in which Hsp104 was ineffective, potentiated Hsp104 variants dissolved protein inclusions, restored protein localization (e.g., TDP-43 returned to the nucleus from cytoplasmic inclusions), suppressed proteotoxicity, and attenuated dopaminergic neurodegeneration in a Caenorhabditis elegans PD model (Jackrel et al., 2014a ; Jackrel and Shorter, 2015 ). Remarkably, these therapeutic modalities originated from degenerate loss of amino acid identity at select positions of Hsp104 rather than specific mutation (Jackrel et al., 2014a ; Jackrel and Shorter, 2015 ). Some of these changes were remarkably small (Jackrel et al., 2014a , 2015 ; Jackrel and Shorter, 2015 ). Thus, potentiated Hsp104 variants could be generated by removal of a methyl group from a single side chain or addition or removal of a single methylene bridge from a single side chain (Jackrel et al., 2014a , 2015 ; Jackrel and Shorter, 2015 ). Thus, small molecules that bind in accessible regions of Hsp104 rich in potentiating mutations might also be able to enhance activity. However, a small-scale screen for small-molecule modulators of Hsp104 revealed only inhibitors (Torrente et al., 2014 ). Nonetheless, our work has established that disease-associated aggregates and amyloid are tractable targets and that enhanced artificial disaggregases can restore proteostasis and mitigate neurodegeneration (Jackrel and Shorter, 2015 ).One surprising aspect of this work is just how many Hsp104 variants we could isolate with potentiated activity. We now have hundreds (Jackrel et al., 2014a ; Jackrel et al., 2015 ). Typically, potentiated Hsp104 variants displayed enhanced activity against several neurodegenerative disease proteins. For example, Hsp104^A503S rescued the aggregation and toxicity of TDP-43, FUS, TAF15, and α-synuclein (Jackrel et al., 2014a ; Jackrel and Shorter, 2014 ). By contrast, some potentiated Hsp104 variants rescued only the aggregation and toxicity of a subset of disease proteins. For example, Hsp104^D498V rescued only the aggregation and toxicity of FUS and α-synuclein (Jackrel et al., 2014a ). A challenge that lies ahead is to engineer potentiated Hsp104 variants that are highly substrate specific to mitigate any potential off-target effects, should they arise (Jackrel and Shorter, 2015 ).Very small changes in primary sequence yield potentiated Hsp104 variants. However, Hsp104 has no metazoan homologue (Erives and Fassler, 2015 ). Now comes the important point. Neuroprotection could be broadly achieved by making very subtle modifications to existing human chaperones with newly appreciated disaggregase activity—for example, Hsp110, Hsp70, and Hsp40 (Torrente and Shorter, 2013 ) and HtrA1 (Poepsel et al., 2015 ).Whether Metazoa even possess a powerful protein disaggregation and reactivation machinery had been a long-standing enigma (Torrente and Shorter, 2013 ). However, it has recently emerged that two metazoan chaperone systems—1) Hsp110, Hsp70, and Hsp40 (Torrente and Shorter, 2013 ) and 2) HtrA1 (Poepsel et al., 2015 )—possess disaggregase activity that could be therapeutically harnessed or stimulated to reverse deleterious protein misfolding in neurodegenerative disease. I suspect that Metazoa harbor additional disaggregase systems that remain to be identified (Guo et al., 2014 ). Whether due to vicissitudes of aging, environment, or genetic background, these disaggregase systems fail in the context of ALS, PD, and AD. Based on the surprising precedent of our potentiated versions of Hsp104 (Jackrel et al., 2014a ; Jackrel and Shorter, 2015 ), I hypothesize that it is possible to engineer and evolve potentiated variants of these human protein disaggregases to more effectively counter deleterious misfolding events in ALS, PD, and AD (Torrente and Shorter, 2013 ; Mack and Shorter, 2016 ).Using classical biochemical reconstitution, it was discovered that one mammalian protein-disaggregase system comprises three molecular chaperones—Hsp110, Hsp70, and Hsp40—which synergize to dissolve and reactivate model proteins trapped in disordered aggregates and can even depolymerize amyloid fibrils formed by α-synuclein from their ends (Shorter, 2011 ; Duennwald et al., 2012 ; Torrente and Shorter, 2013 ). Hsp110, Hsp70, and Hsp40 isoforms are found in the cytoplasm, nucleus, and endoplasmic reticulum, which suggest that protein disaggregation and reactivation can occur in several compartments (Finka et al., 2015 ). Subsequent studies suggest that this system may be more powerful than initially anticipated (Rampelt et al., 2012 ; Mattoo et al., 2013 ; Gao et al., 2015 ; Nillegoda et al., 2015 ). Nonetheless, this system must become overwhelmed in neurodegenerative disorders. Perhaps selectively vulnerable neurons display particular deficits in this machinery. Hence, potentiating the activity of this system via engineering could be extremely valuable. It is promising that directed evolution studies yielded DnaK (Hsp70 in Escherichia coli) variants with improved ability to refold specific substrates (Aponte et al., 2010 ; Schweizer et al., 2011 ; Mack and Shorter, 2016 ), but whether this can be extended to human Hsp70 and the disaggregation of neurodegenerative disease proteins is uncertain.It is exciting that recent studies have revealed that HtrA1, a homo-oligomeric PDZ serine protease, can dissolve and degrade AD-linked tau and Aβ42 fibrils in an ATP-independent manner (Tennstaedt et al., 2012 ; Poepsel et al., 2015 ). HtrA1 first dissolves tau and Aβ42 fibrils and then degrades them, as protease-defective HtrA1 variants dissolve fibrils without degrading them (Poepsel et al., 2015 ). HtrA1 is found in the cytoplasm (∼30%) but is also secreted (∼70%; Poepsel et al., 2015 ). Indeed, HtrA1 is known to degrade substrates in both the extracellular space and the cytoplasm (Chien et al., 2009 ; Campioni et al., 2010 ; Tiaden and Richards, 2013 ). Thus HtrA1 could dissolve Aβ42 fibrils in the extracellular space and tau fibrils in the cytoplasm and simultaneously destroy the two cardinal features of AD (Poepsel et al., 2015 ). I suspect that this system becomes overwhelmed or is insufficient in AD, and thus potentiating and tailoring HtrA1 disaggregase activity could be a valuable therapeutic strategy. For example, it might be advantageous to simply degrade Aβ42 after disaggregation if the peptide has no beneficial function. Thus HtrA1 variants with enhanced disaggregation and degradation activity against Aβ42 could be extremely useful. However, Aβ42 (and related Aβ peptides) may have physiological functions that are presently underappreciated (Soscia et al., 2010 ; Fedele et al., 2015 ), in which case HtrA1 variants with enhanced disaggregase activity but reduced proteolytic activity could be vital. HtrA1 variants with enhanced disaggregase activity but reduced proteolytic activity may also be particularly important to recover functional tau from neurofibrillary tangles to reverse loss of tau function in AD and various tauopathies (Santacruz et al., 2005 ; Trojanowski and Lee, 2005 ; Dixit et al., 2008 ).I suggest that relatively small changes in primary sequence will yield large increases in disaggregase activity for these systems as they do for Hsp104 (Jackrel et al., 2014a ; Jackrel and Shorter, 2015 ). If true, this would further suggest that small molecules that bind in the appropriate regions of Hsp110, Hsp70, Hsp40, or HtrA1 might also enhance disaggregase activity. Thus, isolating small-molecule enhancers of the Hsp110, Hsp70, and Hsp40 or HtrA1 disaggregase systems could yield important therapeutics. Indeed, I hypothesize that enhancing the activity of the Hsp110, Hsp70, and Hsp40 or HtrA1 disaggregase system with specific small molecules will enable dissolution of toxic oligomeric and amyloid forms of various disease proteins and confer therapeutic benefits in ALS, PD, AD, and potentially other neurodegenerative disorders.It is intriguing that several small molecules are already known to enhance various aspects of Hsp70 chaperone activity (Pratt et al., 2015 ; Shrestha et al., 2016 ). These include MKT-077, JG-98, YM-1, YM-8, and 115-7c (Wisen et al., 2010 ; Pratt et al., 2015 ). It is not known whether any of these stimulates the disaggregase activity of the Hsp110, Hsp70, and Hsp40 system. MKT-077, JG-98, YM-1, and YM-8 are rhodocyanines that bind with low micromolar affinity to the nucleotide-binding domain of ADP- but not ATP-bound Hsp70, stabilizing the ADP-bound state (Pratt et al., 2015 ). The ADP-bound state of Hsp70 engages clients with higher affinity, and consequently MKT-077, JG-98, and YM-1 activate binding of Hsp70 to misfolded proteins (Wang et al., 2013 ; Pratt et al., 2015 ). Thus, under some conditions, these small molecules can promote folding of certain Hsp70 clients (Morishima et al., 2011 ; Pratt et al., 2015 ). However, prolonged interaction of clients with Hsp70 promotes their CHIP-dependent ubiquitylation and degradation in vivo (Morishima et al., 2011 ; Wang et al., 2013 ; Pratt et al., 2015 ). Intriguingly, YM-1 promotes clearance of polyglutamine oligomers and aggregates in cells (Wang et al., 2013 ; Pratt et al., 2015 ). MKT-0777, YM-1, JG-98, and YM-8 also promote clearance of tau and confer therapeutic benefit in tauopathy models (Abisambra et al., 2013 ; Miyata et al., 2013 ; Fontaine et al., 2015 ). Of importance, YM-8 is long lived in vivo and crosses the blood–brain barrier (Miyata et al., 2013 ). The dihydropyrimidine 115-7c activates Hsp70 ATPase turnover rate, promotes Hsp70 substrate refolding, and reduces α-synuclein aggregation in cell culture (Wisen et al., 2010 ; Kilpatrick et al., 2013 ). It binds to the IIA subdomain of Hsp70 and promotes the active Hsp70–Hsp40 complex (Wisen et al., 2010 ). Small-molecule enhancers of HtrA1 protease activity have also emerged (Jo et al., 2014 ). Thus it will be important to assess whether these small molecules enhance the activity of their respective disaggregases against various neurodegenerative substrates.Although small molecules that enhance disaggregase activity of endogenous human proteins might be the most immediately translatable, gene-, mRNA-, or protein-based therapies can also be envisioned. For example, adeno-associated viruses expressing enhanced disaggregases might be used to target degenerating neurons (Dong et al., 2005 ; Lo Bianco et al., 2008 ; Deverman et al., 2016 ). Alternatively, if viral vectors are undesirable, modified mRNAs might serve as an alternative to DNA-based gene therapy (Kormann et al., 2011 ). Protein-based therapeutics could also be explored. For example, intraperitoneal injection of human Hsp70 increased lifespan, delayed symptom onset, preserved motor function, and prolonged motor neuron viability in a mouse model of ALS (Gifondorwa et al., 2007 ; Gifondorwa et al., 2012 ). Several other studies suggest that exogenous delivery of Hsp70 can have beneficial, neuroprotective effects in mice (Nagel et al., 2008 ; Bobkova et al., 2014 ; Bobkova et al., 2015 ).Ultimately, if safety and ethical concerns can be overcome in a circumspect, risk-averse manner, CRISPR-Cas9–based therapeutics might even be used to genetically alter the underlying disaggregase to a potentiated form in selectively vulnerable neuronal populations. This approach might be particularly valuable if enhanced disaggregase activity is not detrimental in the long term. Moreover, stem cell–based therapies for replacing lost neurons could also be fortified to express enhanced disaggregase systems. Thus they would be endowed with resistance to potential infection by prion-like conformers that might have accumulated during disease progression (Cushman et al., 2010 ).Enhanced disaggregase activity is likely to be highly advantageous to neurons under circumstances in which protein misfolding has overwhelmed the system (Jackrel et al., 2014a ; Jackrel and Shorter, 2015 ). However, inappropriate hyperactivity of protein disaggregases might also have detrimental, off-target effects under regular conditions in which protein misfolding is not an overwhelming issue (Jackrel et al., 2014a ; Jackrel and Shorter, 2015 ). Thus it may be advantageous to engineer enhanced protein disaggregases to be highly substrate specific. In this way, off-target effects would be readily avoided. There is strong precedent for directed evolution or engineering of specialized chaperone or protein activity from a generalist antecedent (Wang et al., 2002 ; Farrell et al., 2007 ; Smith et al., 2015 ). Thus, engineering specialist disaggregases for each disease substrate could be achieved. Alternatively, transient or intermittent doses of enhanced disaggregases at specific times or places where they are most needed would also minimize potentially toxic side effects. For example, enhanced disaggregase activity might be applied ephemerally to clear existing misfolded conformers and then be withdrawn once the endogenous proteostasis network regains control. Similarly, it is straightforward to envision administration of small-molecule enhancers of disaggregase activity in intermittent protocols that enable facile recovery from potential side effects (Fontaine et al., 2015 ). In this way, any adverse effects of enhanced protein-disaggregase activity under normal physiological conditions would be avoided. Many barriers will need to be safely overcome to implement a successful therapeutic disaggregase, including how to deliver enhanced disaggregase activity to exactly where it is needed. However, these obstacles are not a reason to be pessimistic. On the contrary, the isolation of engineered disaggregases that efficaciously reverse deleterious misfolding of neurodegenerative disease proteins directs our attention to considerably expand the environs in which they should be sought. My closing sentences, therefore, are intended to be provocative.I suspect that neuroprotection could be broadly actualized via precise but subtle alterations to existing protein-disaggregase modalities. The engineering and evolution of protein disaggregases could yield important solutions to avert an imminent plague of neurodegenerative disorders that promises to devastate our society. I strongly suspect that cures for various neurodegenerative disorders will be realized by pioneering as-yet-uncharted regions of disaggregase sequence space or chemical space to elucidate small-molecule enhancers of disaggregase activity.     

Occludin OCEL-domain interactions are required for maintenance and regulation of the tight junction barrier to macromolecular flux

In vitro and in vivo studies implicate occludin in the regulation of paracellular macromolecular flux at steady state and in response to tumor necrosis factor (TNF). To define the roles of occludin in these processes, we established intestinal epithelia with stable occludin knockdown. Knockdown monolayers had markedly enhanced tight junction permeability to large molecules that could be modeled by size-selective channels with radii of ∼62.5 Å. TNF increased paracellular flux of large molecules in occludin-sufficient, but not occludin-deficient, monolayers. Complementation using full-length or C-terminal coiled-coil occludin/ELL domain (OCEL)–deficient enhanced green fluorescent protein (EGFP)–occludin showed that TNF-induced occludin endocytosis and barrier regulation both required the OCEL domain. Either TNF treatment or OCEL deletion accelerated EGFP-occludin fluorescence recovery after photobleaching, but TNF treatment did not affect behavior of EGFP-occludin^ΔOCEL. Further, the free OCEL domain prevented TNF-induced acceleration of occludin fluorescence recovery, occludin endocytosis, and barrier loss. OCEL mutated within a recently proposed ZO-1–binding domain (K433) could not inhibit TNF effects, but OCEL mutated within the ZO-1 SH3-GuK–binding region (K485/K488) remained functional. We conclude that OCEL-mediated occludin interactions are essential for limiting paracellular macromolecular flux. Moreover, our data implicate interactions mediated by the OCEL K433 region as an effector of TNF-induced barrier regulation.Tight junctions seal the paracellular space in simple epithelia, such as those lining the lungs, intestines, and kidneys (Anderson et al., 2004 ; Fanning and Anderson, 2009 ; Shen et al., 2011 ). In the intestine, reduced paracellular barrier function is associated with disorders in which increased paracellular flux of ions and molecules contributes to symptoms such as diarrhea, malabsorption, and intestinal protein loss. Recombinant tumor necrosis factor (TNF) can be used to model this barrier loss in vitro or in vivo (Taylor et al., 1998 ; Clayburgh et al., 2006 ), and TNF neutralization is associated with restoration of intestinal barrier function in Crohn''s disease (Suenaert et al., 2002 ). Further, in vivo and in vitro studies of intestinal epithelia show that TNF-induced barrier loss requires myosin light chain kinase (MLCK) activation (Zolotarevsky et al., 2002 ; Clayburgh et al., 2005 , 2006 ; Ma et al., 2005 ; Wang et al., 2005 ). The resulting myosin II regulatory light chain (MLC) phosphorylation drives occludin internalization, which is required for cytokine-induced intestinal epithelial barrier loss (Clayburgh et al., 2005 , 2006 ; Schwarz et al., 2007 ; Marchiando et al., 2010 ). In addition, transgenic EGFP-occludin expression in vivo limits TNF-induced depletion of tight junction–associated occludin, barrier loss, and diarrhea (Marchiando et al., 2010 ). Conversely, in vitro studies show that occludin knockdown limits TNF-induced barrier regulation (Van Itallie et al., 2010 ). The basis for this discrepancy is not understood.One challenge is that, despite being identified 20 yr ago (Furuse et al., 1993 ), the contribution of occludin to tight junction regulation remains incompletely defined. The observation that occludin-knockout mice are able to form paracellular barriers and do not have obvious defects in epidermal, respiratory, or bladder tight junction function (Saitou et al., 2000 ; Schulzke et al., 2005 ) led many to conclude that occludin is not essential for tight junction barrier function. It is important to note, however, that barrier regulation in response to stress has not been studied in occludin-deficient animals.We recently showed that dephosphorylation of occludin serine-408 promotes assembly of a complex composed of occludin, ZO-1, and claudin-2 that inhibits flux across size- and charge-selective channels termed the pore pathway (Anderson and Van Itallie, 2009 ; Turner, 2009 ; Raleigh et al., 2011 ; Shen et al., 2011 ). Although this demonstrates that occludin can serve a regulatory role, it does not explain the role of occludin in TNF-induced barrier loss, which increases flux across the size- and charge-nonselective leak pathway (Wang et al., 2005 ; Weber et al., 2010 ). In vitro studies, however, do suggest that occludin contributes to leak pathway regulation, as occludin knockdown in either Madin–Darby canine kidney (MDCK) or human intestinal (Caco-2) epithelial monolayers enhances leak pathway permeability (Yu et al., 2005 ; Al-Sadi et al., 2011 ; Ye et al., 2011 ). Taken as a whole, these data suggest that occludin organizes the tight junction to limit leak pathway flux, whereas occludin removal, either by knockdown or endocytosis, enhances leak pathway flux.To define the mechanisms by which occludin regulates the leak pathway, we analyzed the contributions of occludin, as well as specific occludin domains, to basal and TNF-induced barrier regulation. The data indicate that TNF destabilizes tight junction–associated occludin via interactions mediated by the C-terminal coiled-coil occludin/ELL domain (OCEL). Further, these OCEL-mediated events are required for TNF-induced barrier regulation. Thus these data provide new insight into the structural elements and mechanisms by which occludin regulates leak pathway paracellular flux.     

Map-Based Cloning of the Gene Associated With the Soybean Maturity Locus E3

Photosensitivity plays an essential role in the response of plants to their changing environments throughout their life cycle. In soybean [Glycine max (L.) Merrill], several associations between photosensitivity and maturity loci are known, but only limited information at the molecular level is available. The FT3 locus is one of the quantitative trait loci (QTL) for flowering time that corresponds to the maturity locus E3. To identify the gene responsible for this QTL, a map-based cloning strategy was undertaken. One phytochrome A gene (GmPhyA3) was considered a strong candidate for the FT3 locus. Allelism tests and gene sequence comparisons showed that alleles of Misuzudaizu (FT3/FT3; JP28856) and Harosoy (E3/E3; PI548573) were identical. The GmPhyA3 alleles of Moshidou Gong 503 (ft3/ft3; JP27603) and L62-667 (e3/e3; PI547716) showed weak or complete loss of function, respectively. High red/far-red (R/FR) long-day conditions enhanced the effects of the E3/FT3 alleles in various genetic backgrounds. Moreover, a mutant line harboring the nonfunctional GmPhyA3 flowered earlier than the original Bay (E3/E3; PI553043) under similar conditions. These results suggest that the variation in phytochrome A may contribute to the complex systems of soybean flowering response and geographic adaptation.FLOWERING represents the transition from the vegetative to the reproductive phase in plants. Various external cues, such as photoperiod and temperature, are known to initiate plant flowering under the appropriate seasonal conditions. Of these cues, light is the most important, being received by several photoreceptors, including the red light (R) and the far-red light (FR)-absorbing phytochromes and the blue/UV-A absorbing cryptochromes and phototorpins (Chen et al. 2004).Phytochrome is the best characterized of these photoreceptors. All higher plant phytochromes are thought to exist as specific dimer combinations (Sharrock and Clack 2004), with each monomer being attached to a light-absorbing linear tetrapyrrole, phytochromobilin. The phytochrome apoproteins are synthesized within the cytosol and assemble autocatalytically with a chromophore to form the phytochrome holoproteins. The R-absorbing form (Pr) is thought to be inactive but is then converted to the active FR-absorbing form (Pfr) by R absorption. The absorption of light triggers the transfer of the phytochrome to the nucleus, where it regulates gene expression. In most plant species, the phytochrome apoproteins are encoded by a small gene family. Type I phytochrome is degraded in the light and is abundant in dark-grown seedlings, whereas type II phytochrome is relatively stable in the light (reviewed by Bae and Choi 2008). In Arabidopsis, five phytochromes (PhyA–E) have been characterized (Clack et al. 1994; Quail et al. 1995). PhyA is type I and is responsible for the very low fluence response and high irradiance response, whereas the other phytochromes are type II and are responsible for red-far/red reversible low fluence response (reviewed by Whitelam et al. 1998).It is well known that mutations in the phytochrome A gene affect the photoperiodic control of flowering. In Arabidopsis, a phyA mutant flowered later in either long-day or short-day conditions with a night break (Johnson et al. 1994; Reed et al. 1994). In rice, combinations of mutant alleles of phytochrome genes conferred various effects on the flowering phenotype. For example, the phyA phyB and phyA phyC double mutants grown under natural-day-length conditions showed earlier flowering phenotypes than wild-type plants (Takano et al. 2005). In pea, a long-day plant, loss- or gain-of-function phyA mutants displayed late or early flowering phenotypes, respectively (Weller et al. 1997, 2001). It is likely that photoperiodic response via phyA signaling is important for crop adaptation to a wide range of growing conditions.In soybean [Glycine max (L.) Merrill], several maturity loci, designated as E loci (Cober et al. 1996a), have been characterized by classical methods. These are E1 and E2 (Bernard 1971), E3 (Buzzell 1971), E4 (Buzzell and Voldeng 1980), E5 (McBlain and Bernard 1987), E6 (Bonato and Vello 1999), and E7 (Cober and Voldeng 2001). Of these, the E1, E3, and E4 loci have been suggested to be related to photoperiod sensitivity under various light conditions (Saidon et al. 1989; Cober et al. 1996b; Abe et al. 2003). In previous studies, using the same populations as in this study, three flowering-time quantitative trait loci (QTL)—FT1, FT2, and FT3 loci—were identified and considered to be identical with the maturity loci E1, E2, and E3, respectively (Yamanaka et al. 2001; Watanabe et al. 2004). Although many loci related to soybean flowering and maturity have been identified, and some candidate genes were recognized using near isogenic lines (NILs) (Tasma and Shoemaker 2003), most of the genes responsible for these loci have not yet been isolated except for the E4 gene. Liu et al. (2008) reported an association between phytochrome A and photoperiod sensitivity. A retrotransposon sequence inserted into the exon of the e4 allele conferred an early flowering phenotype under long-day conditions extended by incandescent lighting.A relationship between the E3 gene and some photoreceptor genes was suggested from different photosensitivity responses of various soybean NILs (Cober et al. 1996a). Cober and Voldeng (1996) also reported a linkage relationship between the E3 and Dt1 loci, which is related to a determinate or indeterminate growth habit phenotype. Additionally, Molnar et al. (2003) reported that Satt229, on linkage group (LG) L, was a proximal simple sequence repeat (SSR) marker to the E3 loci. According to the Soybean Genome Database (Shultz et al. 2006a,b, 2007; http://soybeangenome.siu.edu/) and the Legume Information System (LIS; http://www.comparative-legumes.org/), there are numerous QTL and >60 loci associated with various agronomic traits in the region between Dt1 and Satt373 (∼30–40 cM). This extremely large number of QTL may be the result of linkage between the Dt1 and E3 loci because both loci can affect many aspects of plant morphology. Among these QTL, several associations with the E3 gene have been reported (Mansur et al. 1996; Orf et al. 1999; Funatsuki et al. 2005; Kahn et al. 2008).To identify the genes responsible for the target QTL, fine mapping and map-based cloning strategies are necessary (Salvi and Tuberosa 2005). QTL analysis using intercross-derived populations, such as F₂ and recombinant inbred lines (RILs), have some limitations in genome resolution (10–30 cM) because of the simultaneous segregation of several loci affecting the same trait (Kearsey and Farquhar 1998). Additional strategies are therefore required to locate QTL more precisely. The use of NILs that differ at a single QTL is an effective approach for fine mapping and characterization of an individual locus (Salvi and Tuberosa 2005). However, the development of NILs through repeated backcrossing is time-consuming and laborious (Tuinstra et al. 1997). The use of a residual heterozygous line (RHL), as proposed by Yamanaka et al. (2004), and which is derived from RIL, is a powerful tool for precisely evaluating QTL (Haley et al. 1994). An RHL harbors a heterozygous region where the target QTL is located and a homozygous background in most other regions of the genome. Tuinstra et al. (1997) used a similar term, heterogeneous inbred family, for a selfed RHL population to identify the QTL associated with seed weight in sorghum.This RHL strategy has already been used to identify loci underlying resistance to pathogens in soybean (Njiti et al. 1998; Meksem et al. 1999; Triwitayakorn et al. 2005). After identification of the target loci, novel DNA markers tightly linked to the loci were developed using the amplified fragment length polymorphism (AFLP) method (Meksem et al. 2001a,b). Physical contigs, screened by sequence-characterized amplified region (SCAR) markers converted from these AFLP fragments, are ideal sources for identifying candidate genes for the target traits (Ruben et al. 2006).The aim of this study is to characterize the FT3 locus using a map-based cloning strategy and to confirm the gene responsible for the E3/FT3 locus by allelism tests through comparisons of gene sequences and photosensitivity of several alleles.     

Synapse reorganization—a new partnership revealed

Changes in synaptic function require both qualitative and quantitative reorganization of the synaptic components. Ca²⁺ plays a central role in this process, but the mechanism has not been fully elucidated. Zhang et al report a novel mechanism whereby Ca²⁺/calmodulin (CaM) regulates the stability of the postsynaptic scaffold. Ca²⁺/CaM interacts with PSD-95, a core protein in the postsynaptic density (PSD) that supports synaptic signaling and structural components. Ca²⁺/CaM interferes with the palmitoylation of PSD-95, resulting in the dissociation of PSD-95 from the postsynaptic membrane. This process may explain the reduction of surface glutamate receptor observed during synaptic depression and homeostatic regulation of the synaptic response after prolonged neuronal activity.The adaptation of synaptic strength is a fundamental process for learning and memory. This form of synaptic plasticity is triggered by influx of Ca²⁺ from NMDA-type glutamate receptor (NMDAR), leading to the synaptic insertion or removal of AMPA-type glutamate receptor (AMPAR) and determines the strength of synaptic transmission. This process is mediated by the gross reorganization of the postsynaptic composition in a qualitative and quantitative fashion (Bosch et al, 2014). Importantly, the number of synaptic AMPAR is regulated by the number and affinity of postsynaptic ‘slots’, a hypothetical receptor binding site within a synapse, which is regulated during synaptic plasticity processes.PSD-95, a scaffolding protein at excitatory synapses, has been considered as a major candidate for the slot. It interacts with AMPAR through the TARP/stargazin protein family and modulates the synaptic localization of the receptor as well as the strength of synaptic transmission (El-Husseini et al, 2000). In turn, the localization of PSD-95 at the synapse is regulated by a constant cycle of palmitoylation by protein palmitoyl acyltransferases (PAT) at cysteines 3 and 5, which is required for efficient synaptic targeting of the protein, and depalmitoylation by palmitoyl protein thioesterases (PPT) (El-Husseini et al, 2002; Noritake et al, 2009). Activation of glutamate receptors increases depalmitoylated PSD-95 and releases it from the postsynaptic site (El-Husseini et al, 2002; Sturgill et al, 2009), whereas blockage of neuronal activity by TTX increases palmitoylated PSD-95 and targets it to the synapse (Noritake et al, 2009). However, it remains unclear how neuronal activity controls the palmitoylation/depalmitoylation cycle and the subsequent trafficking of PSD-95 to and from the synapse.In this issue of The EMBO Journal, using a combination of structural biological, biochemical, and cell biological approaches, Zhang et al (2014) revealed a novel mechanism by which Ca²⁺ regulates the synaptic localization of PSD-95. The authors found that Ca²⁺ complexed to CaM can bind PSD-95 within the first 13 residues—the exact location where palmitoylation takes place. Ca²⁺/CaM binds preferentially to unmodified PSD-95, thereby blocking the accessibility of the PAT to the palmitoylation sites. In contrast, Ca²⁺/CaM does not bind to palmitoylated PSD-95, and therefore, palmitoylated PSD-95 is subject to depalymitoylation by PPT. Overall, the net effect of Ca²⁺/CaM is a reduction of PSD-95 bound to the synaptic membrane (Fig​(Fig1).1). Consistent with this, a PSD-95 mutant that was unable to bind Ca²⁺/CaM does not leave the synapse following glutamate/glycine treatment. Furthermore, the mutant PSD-95 showed an increased in synaptic accumulation, indicating that the treatment also increases palmitoylation activity but it is normally dominated by the depalmitoylating action of Ca²⁺/CaM binding (Noritake et al, 2009).Open in a separate windowFigure 1Calcium influx-induced release of PSD-95 and glutamate receptor from the synapseAt a synapse, cysteine residues of PSD-95 (C3 and C5) are under a continuous cycle of palmitoylation by protein palmitoyl acyltransferase (PAT), and depalmitoylation by palmitoyl protein thioesterase (PPT). A. Palmitoylated PSD-95 associates with the synaptic membrane and CDKL5 and serves as a ‘slot’ for AMPAR at the synapse through the interaction with TARP/stargazin. B. Upon glutamate stimulation, Ca²⁺ influx through NMDARs induces binding of Ca²⁺/CaM to PSD-95. Ca²⁺/CaM blocks the accessibility of PAT, thereby facilitating the depalmitoylation of PSD-95, which subsequently allows PSD-95 to dissociate from the synaptic membrane and CDKL5. C. The dissociation of PSD-95 from the synaptic membrane reduces the number of available ‘slots’ for AMPAR on the postsynaptic membrane, leading to a reduction of AMPAR-mediated synaptic transmission.The beauty of the work by Zhang et al (2014) is that the structure fully explains the biology. However, several important questions remain. Above all, it is still unclear when this mechanism would operate. Generally, the palmitoylation/depalmitoylation cycle is considered to be in the order of minutes. Therefore, for Ca²⁺/CaM to effectively reduce the palmitoylation of PSD-95, a prolonged influx of Ca²⁺ is required. In contrast, the rise in intracellular Ca²⁺ induced by a single excitatory postsynaptic current (EPSC) or dendritic action potential is in the order of milliseconds to seconds. A slow repetitive stimulation protocol, such as that used to induce long-term depression (LTD) (1 Hz, 15 min), may be effective in increasing Ca²⁺/CaM for a sufficiently long period of time. Homeostatic scaling induced by a prolonged increase in network activity (for example, via the pharmacological blockade of inhibitory synaptic transmission) may also be a possible mechanism for the Ca²⁺/CaM-mediated removal of synaptic PSD-95. In contrast, stimulation used to induce long-term potentiation (LTP) (a brief high-frequency stimulation such as 100 Hz, 1 s) may not be effective. Indeed, Bosch et al (2014) found that the induction of LTP at single dendritic spines in hippocampal CA1 pyramidal neurons does not decrease or increase PSD-95 during the first hour after induction even if the dendritic spine enlarges during this period.In this context, it is important to understand what impact Ca²⁺/CaM-mediated removal of synaptic PSD-95 has on synaptic transmission. If PSD-95 is indeed the slot for AMPAR, the Ca²⁺/CaM-mediated removal of synaptic PSD-95 is expected to reduce the synaptic transmission. The stimulation protocol used here by Zhang et al (bath application of glutamate/glycine) is similar to previously reported approaches to induce ‘chemical’ LTD and hence can provide an explanation for the decrease in PSD-95 from the synapse for 10–15 min after stimulation. The PSD-95 mutant that is unable to bind Ca²⁺/CaM will be useful to further analyze the link between the observed PSD-95-Ca²⁺/CaM interaction and synaptic plasticity.PSD-95 is also known to interact with the cyclin-dependent protein kinase-like kinase 5 (CDKL5) at the first 19 residues in a palmitoylation-dependent manner (Zhu et al, 2013). As expected, Ca²⁺/CaM binding also regulates CDKL5 association with PSD-95. The treatment of neurons with NMDA reduces the palmitoylation of PSD-95 and, concomitantly, the association with CDKL5. Mutations of CDKL5 and netrin-G1 gene have been reported in patients with an atypical form of Rett syndrome. Netrin-G1 ligand (NGL1) has been identified as an interaction partner and substrate of CDKL5 (Ricciardi et al, 2012). CDKL5 phosphorylates NGL1, and this phosphorylation stabilizes the interaction of NGL1 with PSD-95. Given that the Ca²⁺ signal only lasts a few milliseconds to seconds, it is important to investigate the spatiotemporal interaction between CaM, PSD-95, CDKL5, and NGL1 in dendritic spines during physiological and pathological conditions. In addition, CDKL5 can function as an upstream modulator of Rac signaling during development (Chen et al, 2010). Together with the fact that Rac also plays an important role in long-term potentiation, the PSD-95/CDKL5 complex might regulate Rac activity in the vicinity of the PSD.Other neuronal proteins including AMPAR subunit GluR1/2, glutamate receptor interacting protein (GRIP), G-protein-coupled receptors, δ-catenin, and small and trimeric G-proteins can also undergo palmitoylation, suggesting that this process plays an essential role in subcellular targeting and that these proteins can be regulated by activity (Kang et al, 2008). The question of whether Ca²⁺/CaM can regulate the palmitoylation of these proteins remains. Interestingly, the Ca²⁺/CaM interaction site of PSD-95 does not conform to a canonical IQ-motif, a CaM binding motif found in many other proteins (Zhang et al, 2014). Therefore, a bioinformatic approach is not possible at this point in time. This opens further directions of research.