T lymphocytes activated by persistent viral infection differentially modify the expression of metalloproteinases and their endogenous inhibitors, TIMPs, in human astrocytes: relevance to HTLV-I-induced neurological disease

Activation of T lymphocytes by human pathogens is a key step in the development of immune-mediated neurologic diseases. Because of their ability to invade the CNS and their increased secretion of proinflammatory cytokines, activated CD4+ T cells are thought to play a crucial role in pathogenesis. In the present study, we examined the expression of inflammatory mediators the cytokine-induced metalloproteinases (MMP-2, -3, and -9) and their endogenous inhibitors, tissue inhibitors of metalloproteinases (TIMP-1, -2, and -3), in human astrocytes in response to activated T cells. We used a model system of CD4+ T lymphocytes activated by persistent viral infection (human T lymphotropic virus, HTLV-I) in transient contact with human astrocytes. Interaction with T cells resulted in increased production of MMP-3 and active MMP-9 in astrocytes despite increased expression of endogenous inhibitors, TIMP-1 and TIMP-3. These data suggest perturbation of the MMP/TIMP balance. These changes in MMP and TIMP expression were mediated, in part, by soluble factors (presumably cytokines) secreted by activated T cells. Integrin-mediated cell adhesion is also involved in the change in MMP level, since blockade of integrin subunits (alpha1, alpha3, alpha5, and beta1) on T cells resulted in less astrocytic MMP-9-induced expression. Interestingly, in CNS tissues from neurological HTLV-I-infected patients, MMP-9 was detected in neural cells within the perivascular space, which is infiltrated by mononuclear cells. Altogether, these data emphasize the importance of the MMP-TIMP axis in the complex interaction between the CNS and invading immune cells in the context of virally mediated T cell activation.     

Semaphorin CD100 from activated T lymphocytes induces process extension collapse in oligodendrocytes and death of immature neural cells

An inappropriate cross talk between activated T lymphocytes infiltrating the CNS and neural cells can sustain the onset and progression of demyelination and axonal degeneration in neuroinflammatory diseases. To mimic this deleterious cross talk, we designed an experimental paradigm consisting of transient cocultures of T lymphocytes chronically activated by retrovirus infection (not virus productive) with human multipotent neural precursors or primary oligodendrocytes from rat brain. We showed that activated T lymphocytes induced apoptotic death of multipotent neural progenitors and immature oligodendrocytes after a progressive collapse of their process extensions. These effects were reminiscent of those induced by brain semaphorin on neural cells. Blockade by specific Abs of soluble CD100 (sCD100)/semaphorin 4D released by activated T cells, or treatment with rsCD100, demonstrated that this immune semaphorin has the ability to collapse oligodendrocyte process extensions and to trigger neural cell apoptosis, most likely through receptors of the plexin family. The specific presence of sCD100 in the cerebrospinal fluid and of CD100-expressing T lymphocytes in the spinal cord of patients suffering with neuroinflammatory demyelination pointed to the potential pathological effect of sCD100 in the CNS. Thus, our results show that CD100 is a new important element in the deleterious T cell-neural cell cross talk during neuroinflammation and suggest its role in demyelination or absence of remyelination in neuroinflammatory diseases including multiple sclerosis and human T lymphotropic virus type 1-associated myelopathy.     

Matrix metalloproteinases and their inhibitors,modulators of neuro-immune interactions and of pathophysiological processes in the nervous system

The matrix metalloproteinases (MMPs) belong to a growing family of Zn2+-dependent endopeptidases, secreted or membrane-bound (MT-MMP), that regulate or degrade by proteolytic cleavage protein components of the extracellular matrix, cytokines, chemokines, cell adhesion molecules and a variety of membrane receptors. MMP activity is counterbalanced by their physiological inhibitors, the tissue inhibitors of MMPs (TIMPs), a family of 4 secreted multifunctional proteins that have growth promoting activities. In physiological conditions MMP activity is tightly regulated and altered MMP regulation is associated with pathological processes including inflammation, cell proliferation, cell death and tissue remodeling. The MMP/TIMP system is involved in the development and function of cells of the immune system by promoting their differentiation, activation, migration across basement membranes and tissues. In the last years, data has accumulated indicating that the MMP/TIMP system is expressed in the nervous system where it regulates neuro-immune interactions and plays a major role in pathophysiological processes. In this review, we present recent in vivo and in vitro studies that highlight the contribution of the MMP/TIMP system to various diseases of the nervous system, involving blood brain barrier breakdown, neuroinflammation, glial reactivity, neuronal death, reactive plasticity, and to developmental and physiological processes including cell migration, axonal sprouting and neuronal plasticity. This review also alludes to the beneficial effects of synthetic MMP inhibitors in different animal models of neuropathology. In all, a further understanding of the role of MMPs and TIMPs in the nervous system should contribute to unravel mechanisms of neuronal plasticity and pathology and set the basis of new therapeutic strategies in nervous system disorders based on the development of synthetic MMP inhibitors.     

Activated T cell exosomes promote tumor invasion via Fas signaling pathway

Activated T cells release bioactive Fas ligand (FasL) in exosomes, which subsequently induce self-apoptosis of T cells. However, their potential effects on cell apoptosis in tumors are still unknown. In this study, we purified exosomes expressing FasL from activated CD8(+) T cell from OT-I mice and found that activated T cell exosomes had little effect on apoptosis and proliferation of tumor cells but promoted the invasion of B16 and 3LL cancer cells in vitro via the Fas/FasL pathway. Activated T cell exosomes increased the amount of cellular FLICE inhibitory proteins and subsequently activated the ERK and NF-κB pathways, which subsequently increased MMP9 expression in the B16 murine melanoma cells. In a tumor-invasive model in vivo, we observed that the activated T cell exosomes promoted the migration of B16 tumor cells to lung. Interestingly, pretreatment with FasL mAb significantly reduced the migration of B16 tumor cells to lung. Furthermore, CD8 and FasL double-positive exosomes from tumor mice, but not normal mice, also increased the expression of MMP9 and promoted the invasive ability of B16 murine melanoma and 3LL lung cancer cells. In conclusion, our results indicate that activated T cell exosomes promote melanoma and lung cancer cell metastasis by increasing the expression of MMP9 via Fas signaling, revealing a new mechanism of tumor immune escape.     

Cell and agonist-specific regulation of genes for matrix metalloproteinases and their tissue inhibitors by primary glial cells

An imbalance in the matrix metalloproteinase (MMP) : tissue inhibitor of MMP (TIMP) ratio may be associated with tissue injury. Here, we studied the regulation of TIMP and MMP gene expression in primary glial cultures to ascertain the factors involved in the regulation of these genes in conditions of inflammatory neuropathology. Astrocytes were found to basally express TIMP-1 and TIMP-3 mRNA while microglia expressed only TIMP-2 mRNA. TIMP-4 mRNA was not detectable in either cell type. Treatment with interferon-alpha (IFN-alpha), IFN-gamma, interleukin-3 (IL-3), IL-6 or tumor necrosis factor-alpha (TNF-alpha) did not alter expression of the TIMP genes. However, in astrocytes, but not in microglia, serum, IL-1beta or lipopolysaccharide (LPS) evoked a dose- and time-dependent increase in TIMP-1 mRNA and a coincident down-regulation of the TIMP-3 gene. Astrocytes were found to express mRNA constitutively for MMPs -3, -11 and -14. In contrast, microglia expressed only MMP-12 mRNA under basal conditions. IL-1beta enhanced MMP-3 mRNA levels while LPS increased the MMP-3, -9, -12, -13 and -14 mRNAs. Our findings reveal that regulatory control of TIMP and MMP gene expression by glial cells is agonist- and cell-type specific, and suggest that innate immune signals govern the temporal and spatial expression patterns of TIMP and MMP genes in neuroinflammatory conditions of the CNS.     

Expression of Fas/Fas ligand by decidual leukocytes in hydatidiform mole

Complete hydatidiform moles are entirely paternally derived and, therefore, represent a complete intrauterine allograft that might be expected to provoke an altered maternal immune response compared with that of normal pregnancy. Uterine decidua contains a large leukocyte population, of which 10%-20% are T lymphocytes. Fas ligand (FasL) expression by placental trophoblast may induce apoptosis of Fas+ lymphocytes, thereby facilitating immune tolerance and survival of the molar trophoblast. Our previous studies have shown an increase in activated CD4+ decidual T cells in molar pregnancy compared with normal pregnancy. This study was designed to characterize and quantitate Fas/FasL expression by decidual leukocytes in complete and partial hydatidiform mole compared with that in normal early pregnancy using single and double immunohistochemical labeling (i.e., avidin-biotin-peroxidase and avidin-biotin-alkaline phosphatase). A significant increase was found in Fas and FasL expression by decidual CD4+ T cells in complete (Fas+, P = 0.0106; FasL+, P = 0.0081) and partial (Fas+, P = 0.0131; FasL+, P = 0.0051) hydatidiform moles, as was a significant decrease in Fas expression by decidual CD8+ T cells in complete (P = 0.0137) and partial (P = 0.0202) hydatidiform mole compared with normal early pregnancy. The implications of altered Fas/FasL status of decidual T-cell subsets in hydatidiform mole are also discussed.     

Design and Screening of a Glial Cell-Specific,Cell Penetrating Peptide for Therapeutic Applications in Multiple Sclerosis

Multiple Sclerosis (MS) is an autoimmune, neurodegenerative disease of the central nervous system (CNS) characterized by demyelination through glial cell loss. Current and proposed therapeutic strategies to arrest demyelination and/or promote further remyelination include: (i) modulation of the host immune system; and/or (ii) transplantation of myelinating/stem or progenitor cells to the circulation or sites of injury. However, significant drawbacks are inherent with both approaches. Cell penetrating peptides (CPP) are short amino acid sequences with an intrinsic ability to translocate across plasma membranes, and theoretically represent an attractive vector for delivery of therapeutic peptides or nanoparticles to glia to promote cell survival or remyelination. The CPPs described to date are commonly non-selective in the cell types they transduce, limiting their therapeutic application in vivo. Here, we describe a theoretical framework for design of a novel CPP sequence that selectively transduces human glial cells (excluding non-glial cell types), and conduct preliminary screens of purified, recombinant CPPs with immature and matured human oligodendrocytes and astrocytes, and two non-glial cell types. A candidate peptide, termed TD2.2, consistently transduced glial cells, was significantly more effective at transducing immature oligodendrocytes than matured progeny, and was virtually incapable of transducing two non-glial cell types: (i) human neural cells and (ii) human dermal fibroblasts. Time-lapse confocal microscopy confirms trafficking of TD2.2 (fused to EGFP) to mature oligodendrocytes 3–6 hours after protein application in vitro. We propose selectivity of TD2.2 for glial cells represents a new therapeutic strategy for the treatment of glial-related disease, such as MS.     

Glial cell type-specific responses to menadione-induced oxidative stress

Glial cell types in the central nervous system are continuously exposed to reactive oxygen species (ROS) due to their high oxygen metabolism and demonstrate differential susceptibility to certain pathological conditions believed to involve oxidative stress. The purpose of the current studies was to test the hypothesis that mtDNA damage could contribute to the differential susceptibility of glial cell types to apoptosis induced by oxidative stress. Primary cultures of rat astrocytes, oligodendrocytes, and microglia were utilized, and menadione was used to produce the oxidative stress. Apoptosis was detected and quantitated in menadione-treated oligodendrocytes and microglia (but not astrocytes) using either positive annexin-V staining or positive staining for 3'-OH groups in DNA. The apoptotic pathway that was activated involved the release of cytochrome c from the intermitochondrial space and activation of caspase 9. Caspase 8 was not activated after exposure to menadione in any of the cells. Using equimolar concentrations of menadione, more initial damage was observed in mtDNA from oligodendrocytes and microglia. Additionally, using concentrations of menadione that resulted in comparable initial mtDNA damage, more efficient repair was observed in astrocytes compared to either oligodendrocytes or microglia. The differential susceptibility of glial cell types to oxidative damage and apoptosis did not appear related to cellular antioxidant capacity, because under the current culture conditions astrocytes had lower total glutathione content and superoxide dismutase activity than oligodendrocytes and microglia. These results show that the differential susceptibility of glial cell types to menadione-induced oxidative stress and apoptosis appears to correlate with increased oxidative mtDNA damage and support the hypothesis that mtDNA damage could participate in the initiation of apoptosis through the enhanced release of cytochrome c and the activation of caspase 9.     

Delta-Notch signaling controls the generation of neurons/glia from neural stem cells in a stepwise process

We examined the role of Notch signaling on the generation of neurons and glia from neural stem cells by using neurospheres that are clonally derived from neural stem cells. Neurospheres prepared from Dll1(lacZ/lacZ) mutant embryos segregate more neurons at the expense of both oligodendrocytes and astrocytes. This mutant phenotype could be rescued when Dll1(lacZ/lacZ) spheres were grown and/or differentiated in the presence of conditioned medium from wild-type neurospheres. Temporal modulation of Notch by soluble forms of ligands indicates that Notch signaling acts in two steps. Initially, it inhibits the neuronal fate while promoting the glial cell fate. In a second step, Notch promotes the differentiation of astrocytes, while inhibiting the differentiation of both neurons and oligodendrocytes.     

Improvement in Double Staining With Fluoro-Jade C and Fluorescent Immunostaining: FJC Staining Is Not Specific to Degenerating Mature Neurons

Fluoro-Jade C (FJC) staining has been used to detect degenerating neurons in tissue sections. It is a simple and easy staining procedure and does not depend on the manner of cell death. In some experiments, double staining with FJC and fluorescent immunostaining (FI) is required to identify cell types. However, pretreatment for FJC staining contains some processes that are harsh to fluorophores, and the FI signal is greatly reduced. To overcome this issue, we improved the double staining protocol to acquire clear double-stained images by introducing the labeled streptavidin–biotin system. In addition, several studies indicate that FJC can label non-degenerating glial cells, including resting/reactive astrocytes and activated microglia. Moreover, our previous study indicated that degenerating mesenchymal cells were also labeled by FJC, but it is still unclear whether FJC can label degenerating glial cells. Acute encephalopathy model mice contained damaged astrocytes with clasmatodendrosis, and 6-aminonicotinamide-injected mice contained necrotic astrocytes and oligodendrocytes. Using our improved double staining protocol with FJC and FI, we detected FJC-labeled degenerating astrocytes and oligodendrocytes with pyknotic nuclei. These results indicate that FJC is not specific to degenerating neurons in some experimental conditions:     

Bcl-2 and Fas/APO-1 regulate distinct pathways to lymphocyte apoptosis.

Activation of the cell surface receptor Fas/APO-1 (CD95) induces apoptosis in lymphocytes and regulates immune responses. The cytoplasmic membrane protein Bcl-2 inhibits lymphocyte killing by diverse cytotoxic agents, but we found it provided little protection against Fas/APO-1-transduced apoptosis in B lymphoid cell lines, thymocytes and activated T cells. In contrast, the cowpox virus protease inhibitor CrmA blocked Fas/APO-1-transduced apoptosis, but did not affect cell death induced by gamma-radiation or serum deprivation. Signalling through Fas/APO-1 did not down-regulate Bcl-2 or induce its antagonists Bax and Bcl-xS. In Fas/APO-1-deficient lpr mice, Bcl-2 transgenes markedly augmented the survival of antigen-activated T cells and the abnormal accumulation of lymphocytes (although they did not interfere with deletion of auto-reactive cells in the thymus). These data raise the possibility that Bcl-2 and Fas/APO-1 regulate distinct pathways to lymphocyte apoptosis.     

Multilineage potential of stable human mesenchymal stem cell line derived from fetal marrow

Human bone marrow contains two major cell types, hematopoietic stem cells (HSCs) and mesenchymal stem cells (MSCs). MSCs possess self-renewal capacity and pluripotency defined by their ability to differentiate into osteoblasts, chondrocytes, adipocytes and muscle cells. MSCs are also known to differentiate into neurons and glial cells in vitro, and in vivo following transplantation into the brain of animal models of neurological disorders including ischemia and intracerebral hemorrhage (ICH) stroke. In order to obtain sufficient number and homogeneous population of human MSCs, we have clonally isolated permanent and stable human MSC lines by transfecting primary cell cultures of fetal human bone marrow MSCs with a retroviral vector encoding v-myc gene. One of the cell lines, HM3.B10 (B10), was found to differentiate into neural cell types including neural stem cells, neurons, astrocytes and oligodendrocytes in vitro as shown by expression of genetic markers for neural stem cells (nestin and Musashi1), neurons (neurofilament protein, synapsin and MAP2), astrocytes (glial fibrillary acidic protein, GFAP) and oligodendrocytes (myelin basic protein, MBP) as determined by RT-PCR assay. In addition, B10 cells were found to differentiate into neural cell types as shown by immunocytochical demonstration of nestin (for neural stem cells), neurofilament protein and beta-tubulin III (neurons) GFAP (astrocytes), and galactocerebroside (oligodendrocytes). Following brain transplantation in mouse ICH stroke model, B10 human MSCs integrate into host brain, survive, differentiate into neurons and astrocytes and induce behavioral improvement in the ICH animals. B10 human MSC cell line is not only a useful tool for the studies of organogenesis and specifically for the neurogenesis, but also provides a valuable source of cells for cell therapy studies in animal models of stroke and other neurological disorders.     

Symposium 10: Differentiation Plasticity of Stem Cells. Direct differentiation of radial glial cells from embryonic stem cells

The major role of radial glial cells in neuronal development is to provide support and guidance for neuronal migration. In vitro, neurons, astrocytes and oligodendrocytes have also been generated from neural stem cells and embryonic stem cells, but the generation of radial glial cells in vitro has not yet been reported. Since radial glial cells can lead to neurons and astrocytes during brain development, neurogenesis and gliogenesis of stem cells in vitro may at least in part also utilize the same mechanisms. To test this hypothesis, we utilized five different clones of embryonic (ES) and embryonal carcinoma (EC) stem cell lines to investigate the differentiation of radial glial cells during in vitro neural differentiation. Here, we demonstrate that radial glial cells can be generated from ES/EC cell lines. These ES/EC cell‐derived radial glial cells are similar in morphology to radial glial cells in vivo. They also express several cytoskeletal markers that are characteristics of radial glial cells in vivo. The processes of these in vitro‐generated radial glial cells are organized into scaffolds that appear to support the migration of newly generated neurons in culture. Like radial glial cells in vivo, they appear to differentiate subsequently into astrocytes. Differentiation of radial glial cells may be a common pathway during in vitro neural differentiation of ES cells. This novel in vitro model system may facilitate the investigation of regulation of radial glial cell differentiation and its biological function. Acknowledgements: Supported by USPHS Grant NS11853 and a grant from the Children's Medical Research Foundation.     

Antigenic and functional characterization of a rat central nervous system-derived cell line immortalized by a retroviral vector

We have immortalized rat central nervous system (CNS) cells of primary cultures of rat optic nerve with murine leukemia virus psi-2,SV-40-6, which is defective in assembly and contains the SV-40 large T antigen and neomycin resistance genes, to produce a cell line that we named A7. After drug selection, greater than 90% of the growing cells expressed nuclear SV-40 large T cells and a fraction of these contained the astrocyte-specific marker, glial fibrillary acidic protein. The majority of these cells also expressed surface marker A4 (specific for neural tube derivatives), Ran 2, p185 (the 185-kD phosphoprotein product of the neu oncogene), and fibronectin, but did not express the astrocyte enzymes glutamine synthetase and monoamine oxidase B. Surface markers characteristic of glial progenitors (A2B5) and oligodendrocytes (galactocerebroside) were not detected. After two rounds of cell cloning, subclone A7.6-3 expressed Ran 2, fibronectin, and the neural cell adhesion molecule (N-CAM) but not glial fibrillary acidic protein and A4. The A7 cell line and subclones also displayed certain functions of type 1 astrocytes: the conditioned medium of these cells had a potent mitogenic activity for glial progenitor cells which could be neutralized by anti-platelet-derived growth factor antibodies and monolayers of these cells supported the growth of embryonic hypothalamic neurons. We conclude that a retrovirus containing SV-40 large T antigen can immortalize rat CNS cells and that such immortalized glial cells retain at least two important functions of type 1 astrocytes: the ability to secrete platelet-derived growth factor and to support the growth of embryonic CNS neurons. Moreover, such stable immortalized clonal cell lines can be used to study gene regulation in glial cells.     

Glial cell lineage in vivo in the mouse cerebellum

Glial cells of the cerebellum originate from cells of the ventricular germinative layer, but their lineage has not been fully elucidated. For studying the glial cell lineage in vivo by retrovirus-mediated gene transfer, we introduced a marker retrovirus into the ventricular germinative layer of embryonic day 13 mice. In the resulting adult cerebella, virus-labeled glial cells were grouped in discrete clusters, and statistical analysis showed that these clusters represented clones in high probability. Of 71 of the virus-labeled glial clusters, 33 clusters were composed of astrocytes/Bergmann glia, 10 were composed of only white matter astrocytes, and 24 were composed of only oligodendrocytes. No glial clusters contained virus-labeled neurons. These results suggest that astrocytes/Bergmann glia, white matter astrocytes and oligodendrocytes immediately arise from separate glial precursors: these three glial lineages may diverge in the course of cerebellar development.     

Susceptibility of astrocytes to class I MHC antigen-specific cytotoxicity

Cell-mediated immune mechanisms contribute to tissue injury within the central nervous system (CNS) in a number of experimental diseases, including experimental allergic encephalomyelitis and some viral infections, and may mediate lesion formation in multiple sclerosis. We investigated the conditions under which murine astrocytes can become susceptible targets of cytotoxic T cells. We demonstrate that mouse astrocytes in vitro can be susceptible targets of class I major histocompatibility complex (MHC)-specific cytotoxicity mediated by L3 cytotoxic T lymphocytes (CTL). Expression of appropriate class I MHC antigen on the astrocytes is a requirement, because only cells bearing the H-2d phenotype are susceptible to lysis by L3 cells. BALB/c-H-2dm2 astrocytes lacking the specific determinant recognized by L3 cells are not susceptible to lysis. Astrocyte lysis can, however, occur under culture conditions in which MHC antigen expression is immunocytochemically low or undetectable. Cytolysis can be inhibited by pretreatment of the effector L3 cells with either anti-Lyt-2 monoclonal antibody (mAb) or anti-clonotypic mAb and by preincubation of the glial target cells with an appropriate anti-H-2 antibody (anti-H-2Ld). mAb to lymphocyte function-associated antigen does not inhibit cytotoxicity of the L3 clone against glial cells. Knowledge regarding the role of CTL within the CNS, including the surface molecules involved in glial cell lysis, could further the development of immunotherapies designed to effect immune reactivity within the CNS.     

Cells behaving badly: a theoretical model for the Fas/FasL system in tumour immunology

One proposed mechanism of tumour escape from immune surveillance is tumour up-regulation of the cell surface ligand FasL, which can lead to apoptosis of Fas receptor (Fas) positive lymphocytes. Based upon this 'counterattack', we have developed a mathematical model involving tumour cell-lymphocyte interaction, cell surface expression of Fas/FasL, and their secreted soluble forms. The model predicts that (a) the production of soluble forms of Fas and FasL will lead to the down-regulation of the immune response; (b) matrix metalloproteinase (MMP) inactivation should lead to increased membrane FasL and result in a higher rate of Fas-mediated apoptosis for lymphocytes than for tumour cells. Recent studies on cancer patients lend support for these predictions. The clinical implications are two-fold. Firstly, the use of broad spectrum MMP inhibitors as anti-angiogenic agents may be compromised by their adverse effect on tumour FasL up-regulation. Also, Fas/FasL interactions may have an impact on the outcome of numerous ongoing immunotherapeutic trials since the final common pathway of all these approaches is the transduction of death signals within the tumour cell.     

Inhibition of cell growth and induction of apoptotic cell death by the human tumor-associated antigen RCAS1.

Tumor-associated antigens that can be recognized by the immune system include the MAGE-family, p53, MUC-1, HER2/neu and p21ras. Despite their expression of these distinct antigens, tumor elimination by the immune system is often inefficient. Postulated mechanisms include insufficient expression of co-stimulatory or adhesion molecules by tumor cells, or defective processing and presentation of antigens on their cell surfaces. Tumor cells may also evade immune attack by expressing CD95 (APO-1/Fas) ligand or other molecules that induce apoptosis in activated T cells. Here we describe RCAS1 (receptor-binding cancer antigen expressed on SiSo cells), a membrane molecule expressed on human cancer cells. RCAS1 acts as a ligand for a putative receptor present on various human cell lines and normal peripheral lymphocytes such as T, B and NK cells. The receptor expression was enhanced by activation of the lymphocytes. RCAS1 inhibited the in vitro growth of receptor-expressing cells and induced apoptotic cell death. Given these results, tumor cells may evade immune surveillance by expression of RCAS1, which would suppress clonal expansion and induce apoptosis in RCAS1 receptor-positive immune cells.