Operons Are a Conserved Feature of Nematode Genomes

The organization of genes into operons, clusters of genes that are co-transcribed to produce polycistronic pre-mRNAs, is a trait found in a wide range of eukaryotic groups, including multiple animal phyla. Operons are present in the class Chromadorea, one of the two main nematode classes, but their distribution in the other class, the Enoplea, is not known. We have surveyed the genomes of Trichinella spiralis, Trichuris muris, and Romanomermis culicivorax and identified the first putative operons in members of the Enoplea. Consistent with the mechanism of polycistronic RNA resolution in other nematodes, the mRNAs produced by genes downstream of the first gene in the T. spiralis and T. muris operons are trans-spliced to spliced leader RNAs, and we are able to detect polycistronic RNAs derived from these operons. Importantly, a putative intercistronic region from one of these potential enoplean operons confers polycistronic processing activity when expressed as part of a chimeric operon in Caenorhabditis elegans. We find that T. spiralis genes located in operons have an increased likelihood of having operonic C. elegans homologs. However, operon structure in terms of synteny and gene content is not tightly conserved between the two taxa, consistent with models of operon evolution. We have nevertheless identified putative operons conserved between Enoplea and Chromadorea. Our data suggest that operons and “spliced leader” (SL) trans-splicing predate the radiation of the nematode phylum, an inference which is supported by the phylogenetic profile of proteins known to be involved in nematode SL trans-splicing.     

Time to reconsider the role of ribavirin in Lassa fever

Ribavirin is the only available Lassa fever treatment. The rationale for using ribavirin is based on one clinical study conducted in the early 1980s. However, reanalysis of previous unpublished data reveals that ribavirin may actually be harmful in some Lassa fever patients. An urgent reevaluation of ribavirin is therefore needed.

Fifty years after its discovery, Lassa fever remains uncontrolled, and mortality remains unacceptably high. Since 2015, Nigeria has been experiencing increasingly large outbreaks of Lassa fever, with new peaks reached in 2016, 2017, and 2018. In 1987, McCormick and colleagues reported a case fatality rate (CFR) of 16.5% among 441 patients hospitalized in Sierra Leone [1]. In Nigeria in 2019, 124 deaths were recorded among 554 laboratory-confirmed cases for a CFR of 22% [2].Ribavirin is the only available Lassa fever–specific treatment and has been used routinely for over 25 years. However, intravenous ribavirin is not licensed for Lassa fever. Its mechanism of action is unclear, it is expensive and hard to source, and it has well-known toxicities [3]. Therefore, the evidence for using ribavirin in Lassa fever deserves careful scrutiny. The emergence of potential new therapeutics for Lassa fever, such as favipiravir and monoclonal antibodies, adds further weight to the case for reconsidering the role of ribavirin since the evaluation of new drugs in clinical trials requires a comparison against existing treatments with a known efficacy and safety profile [4,5].The rationale for using ribavirin in Lassa fever is primarily based on one clinical study conducted in Sierra Leone in the late 1970s and early 1980s. McCormick and colleagues [6] reported that in Lassa fever patients with a serum aspartate aminotransferase (AST) level of ≥150 IU/L, the use of intravenous ribavirin within the first 6 days of illness reduced the fatality rate from 61% (11/18) with no ribavirin to 5% (1/20) (p = 0.002). These authors concluded that ribavirin is effective in the treatment of Lassa fever. However, there are long-standing concerns about the methods used in this study. Although randomization was used to assign patients to treatment groups, the comparisons presented were not according to original randomized groups, and we have reconstructed their derivation (Fig 1). Serious limitations to the comparisons presented include the use of historic controls, inclusion of pregnant women in the control group but their exclusion from the ribavirin group (case fatality is around 2-fold higher in pregnant women than nonpregnant patients), and post hoc merging of treatment groups. Despite this and the fact that the results only supported the use of ribavirin in nonpregnant adult patients with AST ≥150 IU/L, this study is the basis upon which ribavirin is now used in all patients with Lassa fever, including children, pregnant women, and people with normal liver function.Open in a separate windowFig 1Reconstruction of the McCormick et al. data.AST, aspartate aminotransferase; PW, pregnant women. † Discrepancy within McCormick et al, with 39 patients reported treated with oral ribavirin but only 38 (14+24) outcomes reported. ‡ Discrepancy within McCormick et al, with table 1 reporting 12/63 but text reporting 13/62.It has been well known among Lassa specialists that the McCormick study reports a subset of a much larger dataset assembled by the Lassa treatment unit in Sierra Leone and that a report on the full dataset was commissioned by the United States Army Medical Research and Development Command. One of us (PH) therefore submitted a freedom of information (FOI) request to access this report. The full report and an accompanying memo are available, and we encourage readers to access and read the materials [7,8]. The memo states that some of the original trial records were unavailable, and the data should be “interpreted with extreme caution.” Nonetheless, the report presents data from 1977 through to 1991 on 807 Lassa fever patients with a known outcome that were assigned to different ribavirin treatment regimens. These newly available data raise important questions about the safety and efficacy of ribavirin for the treatment of Lassa fever.The original data were lost during the civil war in Sierra Leone, but the report contains tables showing the distribution of characteristics of the whole population according to treatment group, an appendix showing individual data for the 405 patients who died, and results of a logistic regression analysis comparing the effect of ribavirin with no treatment for some of the ribavirin regimens, after adjusting for patient characteristics. Based on these data, we derived aggregated datasets containing the number of deaths according to treatment groups and individual characteristics. We combined groups I (“No treatment given”) and X (“Drugs were not available”) as no treatment and all groups in which ribavirin was administered (II, III, and V to IX) as ribavirin. Exhibit III-8 in the FOI report presented case fatality by treatment group and AST, from which we derived crude odds ratios (ORs) comparing ribavirin with no treatment. The logistic regression reported in Exhibit III-9 was restricted to “those treatment groups that yielded the lowest case fatality rates with respect to untreated patients in the high severity patient illness category” (groups II, III, V, and VII). It was adjusted for age, gender, time to admission, time to treatment, length of stay, and log(AST). We also reconstructed analyses by digitizing the data on individuals who died in Appendix D, calculating the number of deaths according to treatment group and AST, and subtracting these numbers from the totals presented in Exhibit III-2. These allowed us to estimate overall mortality ORs before and after adjusting for ribavirin, although the numbers did not entirely match, and so the number of deaths was reduced in some small groups.Estimates of the effect of oral and intravenous ribavirin from the McCormick study and of all ribavirin from the full report are shown in Fig 2. Based on the crude ORs derived from Exhibit III-8, ribavirin reduced mortality only in patients with serum AST ≥150 IU/L, with less benefit (OR 0.48 [95% CI 0.30 to 0.78]) than reported by McCormick and colleagues. However, ribavirin appeared to increase mortality in patients with serum AST <150 IU/L (2.90 [1.42 to 5.95]). In fact, in our analysis, the only stratum in which ribavirin appeared protective (0.38 [0.21 to 0.70]) was serum AST >300 IU/L (Table H in S1 Text). The logistic regression reported in the FOI report suggested a modest reduction in mortality, but the reasons for the choice of treatment groups compared were unclear. In the reconstructed analyses, ribavirin was associated with overall increased mortality (2.12 [1.67, 2.68]), although this was attenuated after adjustment for AST (1.48 [1.05, 2.08]).Open in a separate windowFig 2Forest plot of the OR of death in treatment and risk subgroups.AST, aspartate aminotransferase; FOI, freedom of information; OR, odds ratio.In our view, there is a compelling case to reevaluate the role of ribavirin in the care of patients with Lassa fever. The data suggest that ribavirin treatment may harm Lassa fever patients with AST <150 IU/L. The limitations revealed by the US Army report, such as large amounts of missing data, unclear treatment allocation practices, imbalances in treatment groups, and errors in coding serology results, cast further doubt on the conclusions of the McCormick study. This aligns with 2 recent systematic reviews by Eberhardt and colleagues and Cheng and colleagues, which concluded that the efficacy of ribavirin in Lassa fever was uncertain because of critical risk of bias in existing studies [9,10].Challenging a quarter of century of clinical practice is difficult. The first step is to acknowledge inadequacies in our knowledge and to ensure that treatment recommendations for Lassa fever better reflect the (weak) strength of evidence for ribavirin in different patient populations. Vigorous efforts should be made to engage clinicians and patients in designing a placebo-controlled trial to assess the safety and efficacy of ribavirin treatment in Lassa fever patients, particularly in those with milder disease (as may be indicated by an admission AST <150 IU/L) in whom the available evidence is compatible with ribavirin causing more harm than good.In conclusion, Lassa fever patients are receiving a drug that may lack efficacy or cause harm. It is incumbent on us to ensure that the next 25 years of Lassa fever treatment are built on more solid foundations.     

Mitochondrial impairment observed in fibroblasts from South African Parkinson’s disease patients with parkin mutations

Parkinson’s disease (PD), defined as a neurodegenerative disorder, is characterized by the loss of dopaminergic neurons in the substantia nigra in the midbrain. Loss-of-function mutations in the parkin gene are a major cause of autosomal recessive, early-onset PD. Parkin has been implicated in the maintenance of healthy mitochondria, although previous studies show conflicting findings regarding mitochondrial abnormalities in fibroblasts from patients harboring parkin-null mutations. The aim of the present study was to determine whether South African PD patients with parkin mutations exhibit evidence for mitochondrial dysfunction. Fibroblasts were cultured from skin biopsies obtained from three patients with homozygous parkin-null mutations, two heterozygous mutation carriers and two wild-type controls. Muscle biopsies were obtained from two of the patients. The muscle fibers showed subtle abnormalities such as slightly swollen mitochondria in focal areas of the fibers and some folding of the sarcolemma. Although no differences in the degree of mitochondrial network branching were found in the fibroblasts, ultrastructural abnormalities were observed including the presence of electron-dense vacuoles. Moreover, decreased ATP levels which are consistent with mitochondrial dysfunction were observed in the patients’ fibroblasts compared to controls. Remarkably, these defects did not manifest in one patient, which may be due to possible compensatory mechanisms. These results suggest that parkin-null patients exhibit features of mitochondrial dysfunction. Involvement of mitochondria as a key role player in PD pathogenesis will have important implications for the design of new and more effective therapies.     

ATP sensitive bi-quinoline activator of the AMP-activated protein kinase

The AMP-activated protein kinase (AMPK) regulates cellular and whole-body energy balance in response to changes in adenylate charge and hormonal signals. Activation of AMPK in tissues such as skeletal muscle and liver reverses many of the metabolic defects associated with obesity and Type 2 diabetes. Here we report a bi-quinoline (JJO-1) that allosterically activates all AMPK αβγ isoforms in vitro except complexes containing the γ3 subunit. JJO-1 does not directly activate the autoinhibited α subunit kinase domain and differs among other known direct activators of AMPK in that allosteric activation occurs only at low ATP concentrations, and is not influenced by either mutation of the γ subunit adenylate-nucleotide binding sites or deletion of the β subunit carbohydrate-binding module. Our findings indicate that AMPK has multiple modes of allosteric activation that may be exploited to design isoform-specific activators as potential therapeutics for metabolic diseases.     

Nucleotide Regulation of the Structure and Dynamics of G-Actin

Actin, a highly conserved cytoskeletal protein found in all eukaryotic cells, facilitates cell motility and membrane remodeling via a directional polymerization cycle referred to as treadmilling. The nucleotide bound at the core of each actin subunit regulates this process. Although the biochemical kinetics of treadmilling has been well characterized, the atomistic details of how the nucleotide affects polymerization remain to be definitively determined. There is increasing evidence that the nucleotide regulation (and other characteristics) of actin cannot be fully described from the minimum energy structure, but rather depends on a dynamic equilibrium between conformations. In this work we explore the conformational mobility of the actin monomer (G-actin) in a coarse-grained subspace using umbrella sampling to bias all-atom molecular-dynamics simulations along the variables of interest. The results reveal that ADP-bound actin subunits are more conformationally mobile than ATP-bound subunits. We used a multiscale analysis method involving coarse-grained and atomistic representations of these simulations to characterize how the nucleotide affects the low-energy states of these systems. The interface between subdomains SD2–SD4, which is important for polymerization, is stabilized in an actin filament-like (F-actin) conformation in ATP-bound G-actin. Additionally, the nucleotide modulates the conformation of the SD1-SD3 interface, a region involved in the binding of several actin-binding proteins.     

The Wall-stress Footprint of Blood Cells Flowing in Microvessels

It is well known that mechanotransduction of hemodynamic forces mediates cellular processes, particularly those that lead to vascular development and maintenance. Both the strength and space-time character of these forces have been shown to affect remodeling and morphogenesis. However, the role of blood cells in the process remains unclear. We investigate the possibility that in the smallest vessels blood’s cellular character of itself will lead to forces fundamentally different than the time-averaged forces usually considered, with fluctuations that may significantly exceed their mean values. This is quantitated through the use of a detailed simulation model of microvessel flow in two principal configurations: a diameter D = 6.5μm tube—a model for small capillaries through which red blood cells flow in single-file—and a D = 12μm tube—a model for a nascent vein or artery through which the cells flow in a confined yet chaotic fashion. Results in both cases show strong sensitivity to the mean flow speed U. Peak stresses exceed their means by greater than a factor of 10 when U/D?10 s⁻¹, which corresponds to the inverse relaxation time of a healthy red blood cell. This effect is more significant for smaller D cases. At faster flow rates, including those more commonly observed under normal, nominally static physiological conditions, the peak fluctuations are more comparable with the mean shear stress. Implications for mechanotransduction of hemodynamic forces are discussed.     

Open and Closed Conformations of the Isolated Transmembrane Domain of Death Receptor 5 Support a New Model of Activation

It has long been presumed that activation of the apoptosis-initiating Death Receptor 5, as well as other structurally homologous members of the TNF-receptor superfamily, relies on ligand-stabilized trimerization of noninteracting receptor monomers. We and others have proposed an alternate model in which the TNF-receptor dimer—sitting at the vertices of a large supramolecular receptor network of ligand-bound receptor trimers—undergoes a closed-to-open transition, propagated through a scissorslike conformational change in a tightly bundled transmembrane (TM) domain dimer. Here we have combined electron paramagnetic resonance spectroscopy and potential-of-mean force calculations on the isolated TM domain of the long isoform of DR5. The experiments and calculations both independently validate that the opening transition is intrinsic to the physical character of the TM domain dimer, with a significant energy barrier separating the open and closed states.Death receptor 5 (DR5) is a member of the tumor necrosis factor receptor (TNFR) superfamily that mediates apoptosis when bound by its cognate ligand, TNF-related apoptosis-inducing ligand (1). Upregulated in cancer cells, DR5 is among the most actively pursued anticancer targets (2). TNF-related apoptosis-inducing ligand binds to preassembled DR5 trimers at their extracellular domains, causing the formation of oligomeric ligand-receptor networks that are held together by receptor dimers (3). In the long-isoform of DR5, this dimer is crosslinked via ligand-induced disulfide bond formation between two transmembrane (TM) domain α-helices at Cys-209, and is further stabilized by a GxxxG motif one helix-turn downstream (3).Our recent study of the structurally homologous TNFR1 showed that receptor activation involves a conformational change that propagates from the extracellular domain to the cytosolic domain through a separation (or opening) of the TM domains of the dimer (4). We have therefore hypothesized that the activation of DR5, and indeed all structurally homologous TNF-receptors, involves a scissorslike opening of the TM domain dimer (Fig. 1).Open in a separate windowFigure 1Activation model of the DR5-L TM dimer. The sequence and positions of the disulfide bond and TOAC spin label (top), along with our previously published model (bottom, left) are shown. We propose an activation model (bottom, right) in which the transmembrane dimer pivots at its disulfide bond to reach an active open conformation.Using electron paramagnetic resonance (EPR) spectroscopy, a technique that has been used previously to study TM helix architecture and dynamics (5,6), and potential-of-mean force (PMF) calculations (7,8), this study addresses the question of whether the isolated disulfide-linked DR5-L TM domain dimer occupies distinct open and closed states (Fig. 1), and how its dynamic behavior contributes to the free-energy landscape of the opening transition of the full-length receptor.The DR5-L TM domain was synthesized with TOAC, an amino acid with a nitroxide spin label rigidly fixed to the α-carbon (9), incorporated at position 32 (Fig. 1), with some minor modification to facilitate EPR measurements. Previous work confirmed that this peptide forms disulfide-linked dimers (e.g., via comparison to 2-ME treated sample) and a negligible population of higher-order oligomers (further supported by model fitting of the EPR data below). For peptide work, residues were renumbered such that Thr-204 corresponds to Thr-1, and so on. The cytosolic Cys-29 (which we previously showed does not participate in a disulfide bond in cells) was replaced with serine to prevent the formation of antiparallel disulfide-linked dimers, and Trp-34 was replaced with tyrosine to prevent intrinsic fluorescence in fluorescence studies (not published). Continuous-wave (CW) dipolar EPR (sensitive only to spin-spin distances <25 Å) was used to measure TOAC-TOAC distances within the TM dimers and revealed an ordered Gaussian distribution centered at 16 Å (full width half-maximum (FWHM) = 4 Å), corresponding to a closed state (Fig. 2 A). Double electron-electron resonance (DEER) (sensitive to spin-spin distances from 15 to 60 Å) also detected a short distance consistent with the dipolar EPR data, along with a longer, disordered component (32.9 Å, FWHM = 28 Å) (Fig. 2 B). Together, these measurements indicate the presence of a compact, ordered closed state and a broader, disordered open state. EPR on oriented membranes also indicated two structural states. Global fitting revealed two populations of spin-label tilt angles (orientation of the nitroxide principal axis relative to the membrane normal): a narrow conformation (24°, FWHM = 20°), and a disordered conformation (50°, FWHM = 48°) (Fig. 2 C). This bimodal orientational distribution (Fig. 2 C) is remarkably consistent with the bimodal distance distribution (Fig. 2 B).Open in a separate windowFigure 2EPR spectra (left) of 32-TOAC-DR5 in lipid, and resulting structural distributions (right). (A) CW dipolar EPR spectra (left) of dimer (1 mM diamide) and monomer (1 mM 2-mercaptoethanol). Best-fit spin-spin distance distribution was a single Gaussian centered at 16 ± 2 Å (right). (B) The DEER waveform (left) of 32-TOAC-DR5 dimer was best fit (right) to a two-Gaussian distribution. The short distance was constrained to agree with the CW data, because DEER has poor sensitivity for distances <20 Å. The long-distance distribution is centered at 32.9 Å and is much broader. (C) CW EPR spectra (left) of 32-TOAC-DR5, with the membrane-normal oriented parallel (red) and perpendicular (blue) to the field. Simultaneous (global) fitting of these spectra reveals narrow and broad components (right). (In panels B and C, the overall distribution is plotted as black, while the closed and open components are plotted as green and magenta, respectively.)We subsequently conducted a PMF calculation (10) using the DR5-L TM dimer starting configuration developed by our group previously (3), embedded in a DMPC bilayer, with the Leu-32/Leu-32 C_α distance as the reaction coordinate. Three calculations were run from independent starting configurations, each using 50 windows spaced in 0.5° increments, and run for 20 ns at each window (totaling 3 μs). Each of the calculations yielded a similar result, and the averaged free energy curve (Fig. 3 A) agrees remarkably well with our EPR measurements: a narrow distribution at the closed conformation (∼16 Å, Fig. 3 B) separated by an ∼3 kcal/mol energy barrier from a broad distribution of accessible open conformations at ∼27 Å, (Fig. 3 C). Each of the three individual PMF plots can be found in Fig. S1 in the Supporting Material.Open in a separate windowFigure 3(A) PMF calculation of the DR5 TM domain dimer along the Leu-32/Leu-32 distance reaction coordinate. The PMF calculation reveals a narrow closed state and a broader open state separated by a free energy barrier. Representative snapshots of the (B) closed state and (C) open state.In the closed state, the helices are tightly packed at the GxxxG interfacial motif and all the way down the juxtaposed helix faces at residues Ala-18, Leu-22, Ala-25, and Val-26. The tight packing is aided by kinking and twisting of the two helices around their common axis, increasing the interacting surface area. In the open conformations, the Ala-18, Leu-22, Ala-25, and Val-26 pairs are dissociated and, interestingly, the GxxxG motif at Gly-10 and Gly-14 remains tightly packed. The open state energy well is only slightly less favorable than the closed state (by ∼2 kcal/mol), and its free energy profile is relatively broad and flat. The increased crossing angle in the open state is facilitated by straightening of the helix kink and is not accommodated by a change in bilayer thickness (see Fig. S3, A and B).The observed change in helix-helix distance (11 Å between the two minima in the PMF) is extremely close to that observed previously in live-cell FRET studies of a constitutively active form of TNFR1 (∼8 Å change between states using large fluorescence probes at the cytosolic domains) (4). The change observed in the EPR data (17 Å) may be an overestimate because the measurement is made between TOAC spin labels that likely protrude from the two helices, depending on rotational orientation. These results collectively show that activation of these receptors requires a small, but clearly significant conformational opening of the TM domains. One important note is that our EPR experiments recapitulate the equilibrium distribution of the two states despite there being no driving force to traverse the barrier between them (∼3 kcal/mol in the closed-to-open transition and ∼1 kcal/mol in the open-to-closed transition, Fig. 3). We do not interpret the results to mean that the dimer necessarily traverses these barriers at 4°C. Rather, there likely exist multiple reaction paths for dimerization of the abstracted TM domains. Finally, in the context of the full-length receptor, how the ligand induces a conformational change capable of overcoming the closed-to-open barrier remains an important question.Whether the observed structural transition in the TM domain dimer of the long-isoform of DR5 is a ubiquitous conformational switch that acts over the entire TNFR superfamily remains unknown. Vilar et al. (11) first proposed a similar scissors-model for activation of p75 neurotrophin receptor, which has a cysteine at the center of its TM helix. The short isoform of DR5 lacks a TM domain cysteine, but does form noncovalent dimers in cells, with likely TM domain dimer contacts (3). Among the other closely related and structurally homologous members of the TNFR superfamily, TNFR1 contains a cysteine at the center of the TM domain, but lacks any discernible small residue motifs (e.g., GxxxG). TNFR2 lacks a TM cysteine on the extracellular side, but does have a GxxxG motif positioned similarly to that of DR5. On the other hand, Death Receptor 4, whose functional distinction from DR5 has remained somewhat elusive, lacks both a cysteine and any recognizable small-residue hydrophobic motif.In summary, we have extended recent findings that point to the TM domain of DR5 as an essential structural component in the conformational change associated with activation. Our findings that the DR5-L TM domain occupies distinct open and closed states, separated by a substantial energy barrier, points the way to further studies across the TNF-receptor superfamily.     

Independent Synchronized Control and Visualization of Interactions between Living Cells and Organisms

To investigate the early stages of cell-cell interactions occurring between living biological samples, imaging methods with appropriate spatiotemporal resolution are required. Among the techniques currently available, those based on optical trapping are promising. Methods to image trapped objects, however, in general suffer from a lack of three-dimensional resolution, due to technical constraints. Here, we have developed an original setup comprising two independent modules: holographic optical tweezers, which offer a versatile and precise way to move multiple objects simultaneously but independently, and a confocal microscope that provides fast three-dimensional image acquisition. The optical decoupling of these two modules through the same objective gives users the possibility to easily investigate very early steps in biological interactions. We illustrate the potential of this setup with an analysis of infection by the fungus Drechmeria coniospora of different developmental stages of Caenorhabditis elegans. This has allowed us to identify specific areas on the nematode’s surface where fungal spores adhere preferentially. We also quantified this adhesion process for different mutant nematode strains, and thereby derive insights into the host factors that mediate fungal spore adhesion.     

A phylogenetic comparative study of flowering phenology along an elevational gradient in the Canadian subarctic

Climate change is affecting high-altitude and high-latitude communities in significant ways. In the short growing season of subarctic habitats, it is essential that the timing and duration of phenological phases match favorable environmental conditions. We explored the time of the first appearance of flowers (first flowering day, FFD) and flowering duration across subarctic species composing different communities, from boreal forest to tundra, along an elevational gradient (600–800 m). The study was conducted on Mount Irony (856 m), North-East Canada (54°90′N, 67°16′W) during summer 2012. First, we quantified phylogenetic signal in FFD at different spatial scales. Second, we used phylogenetic comparative methods to explore the relationship between FFD, flowering duration, and elevation. We found that the phylogenetic signal for FFD was stronger at finer spatial scales and at lower elevations, indicating that closely related species tend to flower at similar times when the local environment is less harsh. The comparatively weaker phylogenetic signal at higher elevation may be indicative of convergent evolution for FFD. Flowering duration was correlated significantly with mean FFD, with later-flowering species having a longer flowering duration, but only at the lowest elevation. Our results indicate significant evolutionary conservatism in responses to phenological cues, but high phenotypic plasticity in flowering times. We suggest that phylogenetic relationships should be considered in the search for predictions and drivers of flowering time in comparative analyses, because species cannot be considered as statistically independent. Further, phenological drivers should be measured at spatial scales such that variation in flowering matches variation in environment.