F-actin homeostasis through transcriptional regulation and proteasome-mediated proteolysis

Significance

Cytoskeletal actin microfilaments have roles in cell-shape determination, motility, membrane trafficking, and cell division. Actin filaments respond dynamically to environmental changes and shifting cellular needs, and functionally different actin subtypes may play important roles in such responses. The alga Chlamydomonas has two actins: IDA5, an actin of conventional sequence that is expressed in normal growing cells, and NAP1, a divergent actin that is normally not expressed. Disruption of IDA5 filaments results in rapid transcriptional induction of NAP1 and hundreds of other genes, rapidly replacing all IDA5 filaments with NAP1 filaments, in part by proteasome-mediated degradation of IDA5. This system allows resistance of Chlamydomonas to actin-depolymerizing drugs and probably also compensates for other, diverse actin cytoskeletal perturbations, whether intrinsic or induced.

Abstract

Many organisms possess multiple and often divergent actins whose regulation and roles are not understood in detail. For example, Chlamydomonas reinhardtii has both a conventional actin (IDA5) and a highly divergent one (NAP1); only IDA5 is expressed in normal proliferating cells. We showed previously that the drug latrunculin B (LatB) causes loss of filamentous (F-) IDA5 and strong up-regulation of NAP1, which then provides essential actin function(s) by forming LatB-resistant F-NAP1. RNA-sequencing analyses now show that this up-regulation of NAP1 reflects a broad transcriptional response, much of which depends on three proteins (LAT1, LAT2, and LAT3) identified previously as essential for NAP1 transcription. Many of the LAT-regulated genes contain a putative cis-acting regulatory site, the “LRE motif.” The LatB transcriptional program appears to be activated by loss of F-IDA5 and deactivated by formation of F-NAP1, thus forming an F-actin–dependent negative-feedback loop. Multiple genes encoding proteins of the ubiquitin-proteasome system are among those induced by LatB, resulting in rapid degradation of IDA5 (but not NAP1). Our results suggest that IDA5 degradation is functionally important because nonpolymerizable LatB-bound IDA5 interferes with the formation of F-NAP1. The genes for the actin-interacting proteins cofilin and profilin are also induced. Cofilin induction may further the clearance of IDA5 by promoting the scission of F-IDA5, whereas profilin appears to function in protecting monomeric IDA5 from degradation. This multifaceted regulatory system allows rapid and quantitative turnover of F-actin in response to cytoskeletal perturbations and probably also maintains F-actin homeostasis under normal growth conditions.

Actin is one of the most highly conserved eukaryotic proteins and plays important roles in a wide range of biological processes, including the determination of cell shape and polarization, vesicle transport and endocytosis, cell motility, and cytokinesis (1–4). Actin functions primarily in its filamentous form (F-actin) rather than as globular monomers (G-actin). F-actin can be nucleated by either of two major nucleators: formins, which nucleate the formation of long, unbranched filaments and bundles, and Arp2/3 complexes, which induce the formation of branched meshworks (5). Different species contain different numbers of genes encoding actins, ranging from 1 in yeast and Giardia (6–8), through 6 in vertebrates (9), 2 to 21 in land plants (10, 11), and >30 in the slime mold Dictyostelium discoideum (12). When multiple actin genes are present, they are often transcribed in distinct patterns under different environmental conditions and/or at different stages of development and in different cell types of multicellular organisms. In most organisms, it is not known how the different actin isoforms are regulated and contribute differentially to cellular processes because of the challenges posed for genetic and molecular analyses by their high degree of structural similarity and functional redundancy. In a few cases, there is good evidence that actin isoforms have functional differences (9, 13, 14), but even in these cases it is not fully understood how the abundance of the different isoforms is regulated.

The unicellular green alga Chlamydomonas reinhardtii provides an attractive system for study of the regulation and function of different actin isoforms. Its genome encodes two actins, the conventional IDA5 (∼90% identical to vertebrate actins) and the unconventional NAP1 (∼65% identical to IDA5; refs. 15 and 16). Vegetative wild-type cells express only IDA5, which forms filaments that localize to the cell cortex, around the basal bodies, and in cage-like structures around the nucleus (17–20). NAP1 is expressed transiently during mating and flagellar regeneration and localizes to the fertilization tubule, basal bodies, and flagella (19, 21), where IDA5 is also present (17–19, 22, 23). NAP1 is also constitutively up-regulated in ida5 null mutants (17). Despite its divergent primary sequence, NAP1 provides essential actin functions, as shown by the viability of ida5 null mutants and inviability of ida5 nap1 double mutants (17, 20).

We reported previously that vegetative Chlamydomonas cells lose F-actin (F-IDA5) within ∼10 min when treated with the drug latrunculin B (LatB) (19, 20), which blocks actin polymerization in most organisms by binding strongly to G-actin subunits (24–26). However, the LatB-treated cells up-regulate transcription of NAP1 within 10–30 min, and NAP1 protein then forms filaments that are highly resistant to LatB (20). As a result, Chlamydomonas cells continue to proliferate without a detectable pause after addition of LatB. A genetic screen for LatB-sensitive mutants yielded mutations in NAP1 and in three other genes, LAT1, LAT2, and LAT3, whose products are all required for the induction of NAP1 upon LatB treatment. LAT2 is a protein of unknown function for which homologs can be found only in closely related green algae (Volvocale lineage), whereas LAT1 and LAT3 are predicted protein kinases with homology to p21-activated kinases and MAPKKKs (20). Such kinases are conserved widely in eukaryotes, and some of them have been implicated in cellular responses to perturbations of F-actin (27–29). Taken together, the results suggested that Chlamydomonas has a pathway that monitors the integrity of the F-actin cytoskeleton and maintains its homeostasis.

Our previous study (20) left several major questions open. First, how does the actin-homeostasis pathway sense the perturbation in F-actin? Second, what other factors are involved in the pathway? Third, does the pathway induce genes in addition to NAP1 that are also important for the ability of Chlamydomonas cells to continue proliferating in the presence of LatB? Here, we used transcriptomic and genetic analyses to explore these questions further.

Results

Global Transcriptional Response to LatB.

To identify genes that are regulated in response to LatB, we used RNA sequencing (RNA-seq) to analyze the transcriptome of LatB-treated wild-type cells through a time course of 120 min (Materials and Methods, RNA-seq Experiment 1). An ida5-1 null mutant (15, 20) was included as a control because this strain neither shows an obvious cellular response nor induces NAP1 in response to LatB treatment (20). After normalization of read counts (Materials and Methods), k-means clustering was used to group the genes into six clusters based on the similarity of their expression profiles (Fig. 1A). Clusters 1 (2,063 genes) and 5 (2,216 genes) showed significant induction and repression, respectively, in the presence of LatB; NAP1 was a conspicuous member of cluster 1 (Fig. 1B, top line). The genes in these clusters were largely unresponsive to LatB in the ida5-1 mutant (Fig. 1A), indicating that the transcriptional responses were specifically dependent on the LatB–IDA5 interaction. In contrast, the genes in clusters 2, 3, 4 (Fig. 1A), and 6 (see Fig. 1A legend) showed only small changes, and/or changes that were similar in wild-type and ida5-1 strains, in response to LatB. Thus, LatB causes changes in the expression of large but specific sets of genes in Chlamydomonas, in addition to NAP1.

Fig. 1.

LAT-pathway-dependent transcription of a large but specific set of Chlamydomonas genes in response to depolymerization of F-actin by LatB. (A) Heat-map display of gene expression in wild-type and ida5-1 cells treated with LatB (see Materials and Methods, RNA-seq Experiment 1). All genes were grouped into six clusters based on similarity of expression patterns (k-means clustering using the data from both the wild-type and ida5-1 strains), followed by a hierarchical sorting within each cluster. Log₂ fold changes in read counts relative to wild type at 0 min are shown for each gene in clusters 1–5, which contain 2,063, 1,265, 3,114, 1,569, and 2,216 genes, respectively. The 7,514 genes in cluster 6 all showed little or no change in expression in either strain, so this cluster is omitted from the figure for simplicity. (B) Expression of actin-cytoskeleton genes in LatB-treated cells. Data for genes encoding actins, actin-related proteins (ARPs), formins, Arp2/3 complex subunits (ARPCs), myosins, profilin, and cofilin were extracted from A. (C and D) Expression in mutants of genes that were differentially regulated in wild type during a 90-min exposure to LatB (see Materials and Methods, RNA-seq Experiment 3). (C) Venn diagram showing the numbers of genes up-regulated ≥twofold in the indicated strains (P_ADJ < 0.05). The “LAT-target” genes used for GO-term analysis (see text) are highlighted in yellow. (D) Heat-map display of gene expression for the 222 and 35 genes that were up-regulated or down-regulated ≥fourfold (P_ADJ < 0.05), respectively. (E) Smoothened density histograms of the expression changes for the up-regulated (Top) and down-regulated (Bottom) genes from D. Shown are the log₂ fold changes of the genes in the indicated strains after LatB treatment relative to wild type before LatB treatment.

The transcriptional response to LatB was almost completely distinct from the responses to the stress-inducing treatments of pH shock (which is known to induce NAP1; ref. 21) and heat stress at 33 °C (SI Appendix, Fig. S1A, Left and Center). In contrast, we observed some overlap between the response to LatB and that to translational inhibition with cycloheximide (CHX) (SI Appendix, Fig. S1A, Right), perhaps reflecting a nonspecific stress component to the LatB response. Alternatively, the overlap may reflect an effect of CHX in compromising F-actin structure or function, as has been observed in other organisms (30). When the 137 genes with the greatest differences in expression between the LatB-alone samples and the CHX-alone samples were examined, 40 genes were observed to be up-regulated specifically by LatB and not CHX (SI Appendix, Fig. S1B). In most cases, this up-regulation was partially or entirely suppressed when CHX was present in addition to LatB (SI Appendix, Fig. S1B), suggesting that new protein synthesis is required for at least some aspects of the LatB response.

Because it seemed likely that the transcriptional response to LatB would involve multiple genes encoding proteins of the actin cytoskeleton, we extracted the data for 18 such genes from the experiment of Fig. 1A. In addition to NAP1 and IDA5 (as reported previously; ref. 20), genes encoding cofilin, profilin, and a formin were significantly up-regulated in wild-type cells (by ∼20-fold, ∼7-fold, and ∼4-fold, respectively) but not in the ida5-1 mutant (Fig. 1B); the possible significance of these results is considered further below. The other 13 genes showed no significant up-regulation or even a mild down-regulation. We also examined a set of 21 genes encoding proteins of the microtubule cytoskeleton; they also showed little or no specific response to LatB (Dataset S1).

Role of the LAT Pathway and of a NAP1-Mediated Negative-Feedback Loop in LatB-Regulated Gene Expression.

The induction of NAP1 (but not of IDA5) by LatB requires the LAT1, LAT2, and LAT3 gene products (the “LAT pathway”; ref. 20). To ask if the LAT pathway is involved in the broader transcriptional response to LatB, we carried out an RNA-seq analysis using two clones of each relevant genotype and a 90-min exposure to LatB (Materials and Methods, RNA-seq Experiment 3). We then called differentially expressed genes by treating the pairs of clones as biological replicates. With a false discovery rate (FDR)-adjusted P value of <0.05 and a fold change of ≥2 as cutoffs, 801 genes were called as up-regulated (Fig. 1C) and 583 genes as down-regulated during the LatB treatment of wild-type cells. With a more stringent fold-change cutoff of ≥4, there were 222 genes up-regulated and 35 down-regulated. We then examined the expression patterns of these genes in the mutants. Of the 222 genes up-regulated ≥fourfold, none was called as differentially expressed in the ida5-1 mutant; like NAP1, most of these genes already showed modestly elevated expression in ida5-1 cells in the absence of LatB treatment (Fig. 1D). In the lat mutants, only 28% (lat3) to 46% (lat1) of the 222 genes were called as differentially up-regulated, and even in many of these cases, the absolute levels of expression were diminished (Fig. 1 D and E). (For example, the distribution of expression levels in the lat3 mutant differed significantly from that in wild type: Kolmogorov–Smirnov test D = 0.70, P = 9.2E-48.) In contrast, in the nap1 mutants, all of the 222 up-regulated genes were called as differentially expressed, and, in most cases, their expression during LatB treatment was significantly enhanced relative to wild type (Fig. 1 D and E; Kolmogorov–Smirnov test D = 0.55, P = 7.6E-30), suggesting the presence of an NAP1-dependent negative-feedback mechanism.

To explore this possibility further, we conducted a similar RNA-seq experiment but over a longer time course. In wild-type cells, all of the 222 genes up-regulated ≥fourfold showed sharp declines in transcript levels at time points beyond 90 min (SI Appendix, Fig. S2A). In striking contrast, transcript levels for nearly all of these genes remained high even after 300 min of LatB treatment in the nap1 mutants (SI Appendix, Fig. S2A), supporting the hypothesis of a NAP1-dependent negative-feedback loop.

To ask if preexisting NAP1 protein is sufficient to activate the negative-feedback function and thus reduce or eliminate the initial transcriptional response to LatB, we used a strain in which NAP1 is transcribed from the constitutively active TUB2 promoter. However, the onset and subsequent attenuation of the transcriptional response were essentially the same in this strain as in wild type (SI Appendix, Fig. S2B). Perhaps IDA5 is preferred to NAP1 for forming filaments in the strain expressing both proteins, and it is the assembly of F-NAP1 filaments, rather than merely the presence of NAP1 protein, that is the key element for the negative-feedback loop.

In contrast to the up-regulated genes, most of the 35 genes down-regulated ≥fourfold were also repressed in the lat1-lat3 and nap1 mutants (Fig. 1D), suggesting that they are not targets of the LAT pathway.

LAT-Response Element.

Using the sets of genes that are differentially expressed in response to LatB, we searched for cis-regulatory elements that might control their transcription. First, using the MEME program (SI Appendix, Materials and Methods), we searched for DNA motifs that are significantly enriched in the regions upstream of the annotated transcription start sites (“TSS-upstream regions”). Although searches using the sets of genes up-regulated in the ida5-1, lat1, lat2, or lat3 mutants did not return any significant hits, a DNA motif with a core conserved sequence (Fig. 2A and SI Appendix, Fig. S3A) was found to be enriched in the 250-bp TSS-upstream regions of the genes up-regulated in wild type; a very similar sequence was found to be enriched in the genes up-regulated in a nap1 mutant (SI Appendix, Materials and Methods). We designated this motif the “LAT-response element” (LRE). We next used the FIMO algorithm to search for LRE sequences in various gene sets and genomic regions. As expected, this analysis revealed an enrichment of LREs in the TSS-upstream regions of genes up-regulated by LatB (Fig. 2B, rows 1–4), and this enrichment was abolished by randomization of the DNA sequences (Fig. 2B, rows 7–9). A similar enrichment was observed using expression data from the nap1 mutant, but not with the data from the lat1-lat3 or ida5 mutants (SI Appendix, Table S1), and no enrichment of LREs was detected in LatB-down-regulated genes (Fig. 2B, rows 5 and 6).

Fig. 2.

Identification and functional analysis of a cis-regulatory element for LAT-pathway-responsive gene expression. (A) Logo representation of the LRE consensus sequence that was identified by MEME in the TSS-upstream regions of genes highly induced by LatB in wild-type cells (SI Appendix, Materials and Methods). (B) Enrichment of LREs in TSS-upstream regions of genes up-regulated by LatB; probabilities of the observed enrichment’s occurring by chance given the background expectations (Fisher’s exact tests) are indicated. N.S., not significant. Row 1, background LRE discovery rate in the TSS-upstream regions of all genes. Rows 2–4, enrichment of LREs in TSS-upstream regions of genes induced to varying extents (P_ADJ < 0.05) by LatB. Rows 5–6, lack of enrichment of LREs in TSS-upstream regions of genes repressed by LatB to varying extents. Rows 7–9, lack of enrichment of LREs when the sequences used in rows 2–4 were randomized. (C) Correlation between numbers of LREs and levels of transcriptional induction by LatB in wild-type and nap1 but not in lat3 strains. Smoothened density histograms of genes with the indicated numbers of LREs in the upstream regions (250-bp upstream of the annotated TSSs plus 5′-UTRs) were plotted as functions of the log₂ fold changes in expression in response to LatB addition (0 versus 90 min). P values indicate the probabilities (from Kolmogorov–Smirnov tests) that a given distribution is equivalent to that for all genes. (D–F) Evidence that the NAP1 LREs can be both sufficient and necessary for transcription in response to LatB. (D) Schematic representations of the original and modified upstream regions of NAP1 and VIPP2 in the reporter constructs used (see SI Appendix, Materials and Methods and Fig. S3C for details). P_NAP1 (pTJ001 and pTJ008): Six LREs found by FIMO (P < 0.0001) in the NAP1 upstream region are indicated by arrows; LRE1 and LRE2 were used to construct P_VIPP2^2xLRE. P_NAP1^∆LRE1 (pTJ002 and pTJ009): A 54-bp deletion removes five of the six LREs in the NAP1 upstream region. P_VIPP2 (pRAM103): The VIPP2 (Cre11.g468050) upstream region lacking LREs. P_VIPP2^2xLRE (pMO606): The VIPP2 upstream region in which two LREs from NAP1 have been inserted immediately upstream of a putative TATA box. (E) Venus-3FLAG expression from reporter constructs with various TSS-upstream-regions sequences. For each construct, 48 randomly selected transformants were grown in liquid Tris-acetate-phosphate (TAP), and Venus fluorescence was measured using a plate reader before and after a 6-h treatment with 10 µM LatB (strong lot). Values above 1,000 arbitrary units (A.U.) are highlighted in red. Where indicated, the significance of the difference in distributions was inferred using the Kolmogorov–Smirnov test. Constructs used: pMO448 (which expresses only the selection marker), pTJ001, pTJ002, pRAM103, and pMO606. (F) Relative abilities of the normal and LRE-deficient NAP1 upstream regions to drive NAP1 expression as judged by rescue of the LatB sensitivity of a nap1 mutant. The mutant strain was transformed with the indicated constructs (from plasmids pTJ008, pTJ009, and pRAM103), and 48 randomly chosen clones for each construct were spotted on TAP plates with and without 1.0 µM LatB (strong lot).

The strongest positional enrichment of LREs was found immediately upstream of the annotated TSSs (SI Appendix, Table S1). Interestingly, across all genes, LRE sequences are underrepresented in the regions immediately upstream of the TSSs compared with random expectation, increasing the significance of the enrichment observed upstream of the LatB-induced genes. Like the previously reported zygotic-response element (31), the LREs are also enriched in the 5′-UTRs and first introns of LatB-responsive genes (SI Appendix, Table S1), but no such enrichment was detected in the coding sequences, introns other than the first, 3′-UTRs, or downstream (presumed terminator) regions. Although these observations may in part reflect misannotation of the TSSs, we confirmed by examination of the available expressed-sequence-tag sequences that in many cases the LREs are indeed within the transcribed regions. Thus, because the method used to discover the LRE motif focused on the TSS-upstream regions, the observation of enrichment also in the 5′-UTRs and first introns suggests that the motif can activate transcription from an upstream TSS.

To ask if the LRE plays a functional role in LatB-induced gene expression, we first analyzed the levels of such expression as a function of the numbers of LREs in the TSS-upstream regions and 5′-UTRs. A genome-wide search identified 5,046 genes with at least one LRE, and we observed a positive correlation between the number of LREs and the level of LatB-induced expression in wild-type cells (Fig. 2C, WT). The correlation was also present in nap1 mutants but was absent in lat3 mutants (Fig. 2C), suggesting a specific connection between the LAT pathway and the LREs. When the expression patterns of the 43 genes with five or more upstream LREs were examined in different mutants, many of them showed patterns very similar to that of NAP1: modest and constitutive up-regulation in ida5, diminished or complete loss of response in lat1-lat3, and enhanced expression in nap1 (SI Appendix, Fig. S3B). Taken together, these results suggest that the LREs play a role in gene induction by LatB treatment in a LAT-pathway-dependent manner.

We next used two assays to ask more directly if the LREs in the region upstream of NAP1 are indeed important for the promotion of transcription in response to LatB. First, we cloned the 543-bp TSS-upstream region (which contains six LREs) and placed it upstream of a codon-optimized VENUS-3FLAG sequence (Fig. 2D, P_NAP1 and SI Appendix, Fig. S3C). Many transformants containing this construct showed clear induction of Venus-3FLAG in response to LatB treatment, as judged by a fluorescence plate-reader assay of 48 clones (Fig. 2E), Western blotting (SI Appendix, Fig. S3D), and fluorescence microscopy (SI Appendix, Fig. S3E). [Note that the expression of transgenes from two-promoter constructs like the ones used here is highly variable among transformants in Chlamydomonas (32).] However, when we deleted a 54-bp region immediately upstream of the TSS that contains five of the six LREs (Fig. 2D, P_NAP1^∆LRE), the transformants showed little or no induction of Venus-3FLAG in response to LatB treatment (Fig. 2E and SI Appendix, Fig. S3 D and E), indicating that the LRE cluster is important for the LatB response. Second, when we used the same two promoters to drive expression of NAP1 itself, most transformants containing the full-length P_NAP1:NAP1 showed rescue of a nap1-1 mutant, whereas most transformants containing P_NAP1^∆LRE1:NAP1 (lacking the 54-bp region) did not (Fig. 2F).

Finally, to test further whether LREs can be sufficient to drive LatB-responsive expression, we used a P_VIPP2:VENUS-3FLAG construct that contains no endogenous LREs (Fig. 2D, P_VIPP2) (33) and showed little induction of Venus-3FLAG above background when transformants were exposed to LatB (Fig. 2E and SI Appendix, Fig. S3F). When two LREs from the NAP1 upstream region were inserted immediately upstream of the TSS in this construct (Fig. 2D, P_VIPP2^2xLRE), transformants showed LatB-responsive Venus expression (Fig. 2E and SI Appendix, Fig. S3F).

Taken together, the results suggest that LREs can be both sufficient and necessary for LatB-induced gene expression, at least in appropriate sequence contexts, and the global survey suggests that such LRE-mediated control may account for a substantial fraction of the overall transcriptional response to LatB.

Induction by LatB of Genes Encoding Proteins of the Ubiquitin-Proteasome System.

To explore the functional significance of the widespread gene induction by the LAT pathway, we performed a Gene Ontology (GO)-term analysis on the 363 “LAT-target” genes that were up-regulated ≥twofold in wild type but not in lat3 mutants (Fig. 1C). Because of the paucity of annotations in the Chlamydomonas database, we relied primarily on annotations in the yeast Saccharomyces cerevisiae to calculate enrichment of GO terms in the up-regulated genes. To do so, we identified S. cerevisiae homologs of the Chlamydomonas gene products by using a method we named R/UBH (reciprocal or unidirectional best blast hit), which looks for pairs of genes in two organisms whose products form a best BLAST hit in at least one direction and a statistically meaningful BLAST hit (E < 0.1) in the other direction (SI Appendix, Materials and Methods). Despite the large evolutionary distance, we found in this way 2,651 candidate S. cerevisiae homologs of 6,473 Chlamydomonas proteins (some yeast proteins had more than one Chlamydomonas homolog), including 148 of the 363 LAT-target gene products. The higher fraction of yeast homologs among the LAT-target gene products (39%) than among all proteins (15%) indicates a significant enrichment of highly conserved genes among the LAT targets (χ² test P < 0.0001). The GO-term-enrichment analysis of the 148 proteins found a highly significant enrichment of terms related to protein degradation mediated by the ubiquitin-proteasome system and essentially nothing else (Fig. 3A and SI Appendix, Table S2). Independent searches against the Chlamydomonas, Arabidopsis thaliana, and mouse databases (SI Appendix, Materials and Methods) gave similar results, indicating that nothing of consequence had been missed because of an absence of related genes in yeast. Most of these genes were not induced in the pH-shock and CHX-treatment gene sets (SI Appendix, Fig. S1), suggesting that the induction is a specific response to LatB.

Fig. 3.

Degradation of IDA5 through a mechanism dependent on the LAT and ubiquitin/proteasome pathways. (A) LAT-pathway-dependent induction of genes whose S. cerevisiae homologs are annotated with the GO term “Proteolysis.” Data are from RNA-seq Experiment 3 (cf. Fig. 1 C–E). (B) Rapid degradation of IDA5 during exposure of wild-type cells (strain CC-124) to LatB. Levels of IDA5 and NAP1 were analyzed by Western blotting using an anti-actin antibody (SI Appendix, Materials and Methods) and then (after stripping the blots) an anti-NAP1 antibody after treatment of cells for various times with 10 µM LatB (strong lot), 10 µg/mL CHX, or both. For each sample, 30 µg of whole-cell extract were analyzed; Coomassie Brilliant Blue (CBB) staining of the membranes provides a loading control. Blots are numbered for ease of reference in the text. (C) Dependence of IDA5 degradation on the LAT pathway. Wild-type (CC-124) and mutant strains of the indicated genotypes were grown at 25 °C and treated with 10 µM LatB (strong lot) for 2 h. Samples were analyzed by Western blotting as in B. (D and E) Dependence of IDA5 degradation on the ubiquitin/proteasome pathway. Samples were analyzed by Western blotting as in B. (D) Lack of IDA5 degradation in a mutant with a temperature-sensitive proteasome subunit. Wild-type (CC-124) and rpt5-1 strains were grown at 21 °C, each culture was split into two, and the subcultures were incubated at 21 °C or 33 °C for 6 h before treatment with 10 µM LatB (weak lot) for 2 h. (E) Dose-dependent inhibition of IDA5 degradation by a proteasome inhibitor. Wild-type (CC-124) cells were grown at 25 °C and treated (or not) with 10 µM LatB (strong lot) and with the indicated concentrations of the proteasome inhibitor MG-132 for 2 h.

Ubiquitin/Proteasome-Dependent Degradation of IDA5 Actin During the Response to LatB.

The general up-regulation of genes of the ubiquitin/proteasome pathway during exposure of cells to LatB suggests that proteolysis of one or more proteins may play an important role in actin homeostasis under these conditions. Indeed, IDA5 was rapidly degraded during LatB treatment of wild-type cells (Fig. 3 B, 1) despite the induction of IDA5 transcription (Fig. 1B and ref. 20) and the accumulation of NAP1 protein (Fig. 3 B, 4) that occur at the same time. This degradation was substantially more rapid than that occurring by baseline protein turnover when new protein synthesis was blocked by CHX (Fig. 3 B, 2). Furthermore, it was slowed when CHX was added together with LatB (Fig. 3 B, 3), suggesting that new protein synthesis is important for the degradation pathway. We hypothesized that the newly synthesized proteins included the products of the specifically up-regulated ubiquitin/proteasome-pathway genes. This hypothesis was supported by three additional lines of evidence. First, IDA5 degradation was largely eliminated in the lat mutants, which do not up-regulate the ubiquitin/proteasome components, but not in a nap1 mutant, which does (Fig. 3C). Second, IDA5 degradation was greatly reduced when a strain carrying a temperature-sensitive mutation in a gene for an essential proteasome subunit was exposed to LatB at its restrictive temperature (Fig. 3D). Third, a dose-dependent reduction in IDA5 degradation was seen when wild-type cells were exposed to LatB in the presence of the proteasome inhibitor MG-132 (Fig. 3E). Importantly, MG-132 treatment did not block LatB induction of NAP1 (Fig. 3E), indicating that the inhibitor did not simply abrogate all response to LatB.

Ubiquitination almost invariably occurs on lysine residues. By sequence alignment, we found eight lysines in IDA5 that are not present in NAP1 (Fig. 4A). To ask if these residues might account for the fact that IDA5 is degraded during exposure to LatB while NAP1 is not, we replaced these lysines in IDA5 with the corresponding residues from NAP1. Tetrad analysis showed that the IDA5^8Kmut transgene rescued the inviability of ida5-1; nap1-1 double mutants (Fig. 4B), indicating that the IDA5^8Kmut protein retains essential actin functions. However, upon LatB treatment, this protein was stable, unlike either endogenous or transgenic wild-type IDA5 (Fig. 4C, lanes 2, 4, and 6), suggesting that ubiquitination of one or more of the eight lysines targets the protein for degradation, although we have not yet been able to demonstrate such ubiquitination biochemically because of the lack of a reliable method to purify IDA5. Interestingly, the lack of degradation of IDA5^8Kmut protein did not prevent the up-regulation of NAP1 during exposure to LatB (Fig. 4C, lane 6), suggesting that IDA5 degradation is not a prerequisite for function of the LAT pathway of transcriptional induction. In addition, IDA5^8Kmut, but not wild-type IDA5 expressed from the same promoter, conferred partial resistance to LatB in in ida5-1; nap1-1 cells (Fig. 4D), possibly because the undegraded IDA5 helps to drive the G-actin/F-actin equilibrium toward F-actin in the presence of low concentrations of LatB.

Fig. 4.

Importance of IDA5-specific lysines for its degradation during LatB exposure. (A) Absence in NAP1 of eight lysines (marked by #) that are conserved in IDA5 and human skeletal α-actin; a single lysine is found specifically in NAP1 (marked by *). (B) Provision of essential actin function(s) by IDA5^8Kmut, which lacks the eight lysines that are absent in NAP1. An mt+ ida5-1 strain was transformed with P_TUB2:IDA5 (from pMO589) or P_TUB2:IDA5^8Kmut (from pMO600). After confirmation of transgene expression by Western blotting, a transformant of each genotype was crossed with a mt− nap1-1 strain, tetrads (shown in columns) were dissected, and segregants were genotyped by allele-specific PCR (for ida5-1), sensitivity to 10 µM LatB (strong lot; for nap1-1), and paromomycin resistance (for the transgenes). Arrowheads, ida5-1; nap1-1; P_TUB2:IDA5 and ida5-1; nap1-1; P_TUB2:IDA5^8Kmut segregants. All inviable segregants were inferred to be ida5-1; nap1-1 based on the presumed 2:2 segregation of all markers in the tetrads. (C) Lack of IDA5^8Kmut degradation, but up-regulation of NAP1, during exposure of cells to LatB. Strains of the indicated genotypes were incubated with or without 3.0 µM LatB (strong lot) for 2 h at 25 °C, and protein levels were analyzed as in Fig. 3B. Results for IDA5, IDA5^8Kmut, and NAP1 were similar for each of two transgenic lines of each genotype. (D) Partial resistance to LatB of nap1-1 cells expressing IDA5^8Kmut. Segregants of the indicated genotypes (taken from the crosses shown in B) were spotted in fivefold dilution series on TAP plates without LatB or with LatB (strong lot) at the indicated concentrations and grown at 25 °C for 4 d.

Roles of Profilin and Cofilin in Modulating the Degradation of IDA5 Actin.

Profilin mRNA (Fig. 1B, PRF1) and protein (Fig. 5A) were modestly induced when wild-type cells were treated with LatB. To explore the possible role(s) of profilin in actin homeostasis during exposure to LatB, we used the prf1-1 temperature-sensitive-lethal mutant (ref. 34, there called “div68-1”). Interestingly, although prf1-1 cells are viable at 21 °C and inviable at 33 °C, they had a barely detectable level of profilin protein even at the lower temperature (Fig. 5B), indicating that the mutation destabilizes the protein at both temperatures even though the effect is lethal only at the higher temperature. Surprisingly, IDA5 was essentially absent in prf1-1 cells even when grown at 21 °C in the absence of LatB, while NAP1 was highly elevated (Fig. 5C), suggesting that IDA5 degradation can be triggered by a lack of binding to profilin and that the lack of IDA5, and thus also of F-IDA5, can promote NAP1 up-regulation even in the absence of drugs.

Fig. 5.

Roles of profilin and cofilin in preventing and accelerating the degradation of IDA5. (A and B) Analysis of profilin levels by Western blotting. For each sample, 15 µg of whole-cell extract were subjected to Western blotting using an antibody to Chlamydomonas profilin. (A) Induction of profilin expression by LatB. Wild-type strain CC-124 was treated with 10 µM LatB (weak lot) at 25 °C for the indicated times. (B) Nearly total absence of profilin in a temperature-sensitive profilin mutant even when grown at a temperature permissive for growth. Wild-type (CC-124) and prf1-1 strains were grown at 21 °C or shifted to 33 °C for 6 h. Images of the same blot with short and long exposure times are shown. (C) Loss of IDA5 and induction of NAP1 in the temperature-sensitive profilin mutant even when grown at 21 °C. (D and E) Synthetic lethality of the profilin mutation with nap1-1 but not with lat1, lat2, or lat3 mutations. (D) For each of the indicated crosses, tetrads were dissected and segregants scored for viability. (E) Representative tetratype tetrads from the indicated crosses were imaged after 2 d at 21 °C. Genotypes of the segregants were determined based on temperature sensitivity (prf1-1) and LatB sensitivity (nap1-1 and lat3-1), and the double-mutant segregants are indicated by the arrowheads. The prf1-1; nap1-1 cells (plate 1) divided several times after germination (presumably due to maternal mRNA and/or protein) before lysing. (F) Rescue of IDA5 protein levels in the profilin mutant by lat mutations. A tetratype tetrad from each relevant cross (D, rows 3–5) was analyzed by Western blotting as in Fig. 3B. (G) Reduced degradation of IDA5 during LatB treatment of a cofilin mutant. Strains of the indicated genotypes were grown at 25 °C with or without treatment for 2 h with 10 µM LatB (weak lot).

Because ida5-1; nap1-1 double mutants (lacking both actins) are inviable (20), we expected that prf1-1; nap1-1 double mutants would also be inviable for the same reason, and this was indeed the case (Fig. 5D, rows 1 and 2 and E, 1). Because the LAT pathway is essential for the induction of NAP1 synthesis, we also expected that prf1-1 lat double mutants would lack both actins and thus be inviable, but, surprisingly, this was not the case (Fig. 5D, rows 3–5 and E, 2). The apparent discrepancy might be explained if the lat mutations rescue IDA5 levels in prf1-1 cells because they block up-regulation not only of NAP1 but also of the ubiquitin-mediated-proteolysis genes (discussed above). Direct examination of IDA5 and NAP1 levels in the single- and double-mutant strains supported this model: Whereas prf1-1 single-mutant cells had high levels of NAP1 and undetectable IDA5, prf1-1 lat double-mutant cells had the opposite pattern (Fig. 5F). Taken together, the results suggest that the LAT pathway targets IDA5 for destruction even in the absence of LatB (probably through induction of the ubiquitin/proteasome system), and that profilin has a previously unknown role in protecting G-IDA5 from such degradation.

Another gene that was highly induced by LatB treatment was COF1, encoding the F-actin-severing protein cofilin (Fig. 1B and SI Appendix, Fig. S4). We described previously a single recessive suppressor of the LatB sensitivity of a nap1 null mutation (20), and we have now shown that this mutation causes a Leu-to-Pro substitution at position 18 in COF1 (Materials and Methods and SI Appendix, Fig. S4). Suppression of nap1 LatB sensitivity by cof1-1 might arise if the loss of F-IDA5 and consequent degradation of IDA5 were slowed by reduced severing of F-IDA5 filaments. Consistent with this hypothesis, IDA5 was substantially stabilized in a LatB-treated cof1-1 strain (Fig. 5G). In wild-type cells, the induction of COF1 by LatB might speed F-IDA5 disassembly, and thus IDA5 degradation, thereby promoting the replacement of F-IDA5 by F-NAP1.

Role of IDA5 Degradation in Actin Homeostasis During Exposure to LatB.

To investigate the possible importance of IDA5 degradation in the response to actin depolymerization by LatB, we used strains carrying a nap1 null mutation and expressing wild-type NAP1 from the constitutive TUB2 promoter, an arrangement that simplifies the analysis by removing transcriptional induction of NAP1 as a factor. In the absence of other mutations, such a strain showed a partial resistance to LatB that was distinctly less than that of wild type (Fig. 6A, columns 1–3 and ref. 20), probably because the levels of NAP1 were less than the induced level in wild-type cells exposed to LatB (Fig. 6B; compare lanes 5, 7, and 8 to lanes 2 and 6). When the strain also harbored a lat3 mutation, its LatB resistance was further decreased (Fig. 6A, column 4 and ref. 20). This observation cannot be explained by a failure of NAP1 induction as a result of the lat3 mutation, because NAP1 in this strain is only expressed ectopically from the TUB2 promoter; therefore, we hypothesized that it might be due to failure of expression of other LatB-induced genes. Specifically, the decreased resistance might be due to reduced expression in the lat3 mutant of the ubiquitin-proteasome system, leading to increased persistence of LatB-bound IDA5 monomers. Indeed, elimination of IDA5 from these strains by an ida5-1 mutation resulted in greatly enhanced LatB resistance, whether or not LAT3 was present (Fig. 6A, columns 5 and 6). Taken together, these results suggest that proteolytic elimination of LatB-bound IDA5 monomers can facilitate actin homeostasis, probably by reducing interference with the assembly and/or function of F-NAP1.

Fig. 6.

Importance of IDA5 degradation for full resistance to LatB. (A) Single cells of the indicated genotypes were placed by micromanipulation on TAP plates containing the indicated concentrations of LatB (strong lot), and the plates were incubated for 3 d at 25 °C. Images are representative of at least 10 cells for each strain and LatB concentration. The strains used in columns 3–6 are segregants from a single cross and therefore have P_TUB2:NAP1 inserted at the same chromosomal location (and thus presumably expressed at approximately the same level); columns 1 and 2 are internal controls. (B) Expression of NAP1 protein from the endogenous locus and the P_TUB2:NAP1 transgene. Strains of the indicated genotypes were treated with 3.0 µM LatB (strong lot) for 120 min at 21 °C, and extracts were analyzed by Western blotting as in Fig. 3B.

Discussion

A Broad Transcriptional Response to Disturbance of F-Actin.

Depolymerization of F-IDA5 by LatB leads to strong transcriptional induction of the gene encoding the divergent actin NAP1 (20) and of several hundred other genes (this study). In most cases, this induction depends on the LAT pathway involving the LAT1, LAT2, and LAT3 gene products and appears to be under the control of a negative-feedback loop that requires NAP1 (and thus also the LAT pathway). Several possible models might explain these observations. First, the pathway might be activated by the transient overabundance of G-IDA5 (or G-IDA5–LatB complexes). However, this model is inconsistent with the up-regulation of NAP1 in the ida5-1 and prf1-1 mutants, which lack G-IDA5. Second, the pathway might be activated by the subsequent loss of IDA5 protein. However, this model is inconsistent with the evidence that NAP1 is induced when IDA5 degradation does not occur (the IDA5^8Kmut mutant and wild-type cells treated with the proteasome inhibitor MG132). Thus, the most attractive model (Fig. 7) appears to be that the structure and/or function of F-actin (either F-IDA5 or F-NAP1) suppresses the transcriptional-response pathway. Upon LatB treatment of wild-type cells, F-IDA5 is lost, which activates the pathway; as NAP1 is produced, F-NAP1 assembles, shutting off the pathway. This model also appears consistent with the observations on the ida5-1 and prf1-1 mutants. In the former, it appears that a low level of constitutive pathway activation is sufficient to maintain levels of NAP1 adequate for cell function, and because F-NAP1 does not depolymerize upon LatB addition, there is no additional transcriptional response upon exposure to the drug. In the prf1-1 mutant, the instability of IDA5 leads to a loss of F-IDA5, and thus activation of the pathway, until sufficient F-NAP1 is present to achieve a steady state.

Fig. 7.

Proposed model for the F-actin–homeostasis pathway in Chlamydomonas. In normal vegetative cells, IDA5 is the only actin expressed; it cycles between the monomeric (G) and filamentous (F) forms with the aid of both profilin (which protects it from degradation and probably delivers it to a formin for polymerization) and cofilin (which probably promotes filament turnover). In LatB-treated cells, binding of G-IDA5 by LatB leads to loss of F-IDA5, which in turn activates a transcriptional pathway that up-regulates multiple genes including the alternative actin gene NAP1 and genes encoding proteins of the ubiquitin-proteasome system. This system then degrades the G-IDA5-LatB complexes (which could otherwise interfere with the polymerization of NAP1) but not NAP1. As NAP1 also does not bind LatB, it polymerizes (probably also with the aid of profilin, as prf1-1 is a lethal mutation at high temperature) into filaments, which in turn shut down the transcriptional pathway, forming a negative-feedback loop.

Examination of the sequences upstream of the LatB-induced genes revealed a potential cis-acting regulatory site, the LRE motif, variants of which were found in the promoters, 5′-UTRs, and introns of many such genes. We found that these motifs are highly enriched near the TSSs and that there was a strong correlation between the number of LREs and the induction of expression by LatB. Thus, although the majority of the 5,046 genes with small numbers of LREs do not respond strongly to LatB, genes with multiple, closely spaced LREs are likely to do so, perhaps because the actual mechanism involves some form of cooperativity of transcription-factor binding. Such homotypic site clustering is a commonly observed feature of transcriptional regulation in diverse systems. We have not yet identified a candidate transcription factor for binding to this motif, and the amino acid sequences of LAT1, LAT2, and LAT3 do not suggest that any of them has the ability to bind specific DNA motifs.

There have been few, if any, genome-wide analyses of transcriptional responses to perturbations of F-actin. However, it has been reported that in mammalian cells, actin and actin-binding proteins participate directly in the transcriptional regulation of many genes in response to shocks to the actin cytoskeleton (35–38). In addition, latrunculin treatment has been shown to activate MAP kinases in yeasts (27, 29). Although most characterization of the downstream responses to activation of the yeast kinases has focused on nontranscriptional regulation, an additional transcriptional component may well exist. In land plants, pathogen infection can cause disturbances in F-actin structures, which in turn induce genes related to the innate-immune response (39, 40). Thus, it appears that the ability to recognize and respond to perturbations of the actin cytoskeleton arose early in eukaryotic evolution.

Regulation of Actin Levels by Ubiquitin-Proteasome-Mediated Proteolysis and Protection from Proteolysis by Profilin.

In GO analyses of the genes induced by LatB through the LAT pathway, the only highly enriched terms were related to the ubiquitin-proteasome system. Many genes encoding components of E3-ubiquitin ligases (including multiple proteins of the SCF complex) and the proteasome itself were up-regulated 2- to 10-fold. Such levels of up-regulation seem very likely to be significant given that the expression of proteasome components is frequently tightly regulated by negative-feedback mechanisms that keep their overall levels constant, so that changes of this magnitude are rarely observed (41).

IDA5 is a likely target of the up-regulated ubiquitin-proteasome system. In LatB-treated cells, IDA5 is rapidly degraded in a process that depends on proteasome function. IDA5 degradation under these conditions also depends on the LAT pathway and new protein synthesis, presumably because components of the ubiquitin-proteasome system must be up-regulated to achieve efficient degradation. Blocking IDA5 degradation by replacement of lysine residues significantly attenuated the LatB sensitivity of a nap1 strain (Fig. 4D), indicating that the ubiquitination of IDA5 is functionally significant.

Even in cells not treated with LatB, IDA5 is degraded (with concomitant up-regulation of NAP1) when profilin function is compromised by the prf1-1 mutation. Thus, binding to profilin may help to maintain the proper folding of G-IDA5, shield it from the ubiquitination machinery, and/or protect it from degradation by promoting its assembly into filaments. Consistent with the second possibility, at least three of the eight lysine residues in IDA5 that are possible targets for ubiquitination (K120, K286, and K361) are located in or near the actin–profilin interfaces where this has been studied (42–44). The adaptive benefit of PRF1 up-regulation in response to LatB may be sequestration of the transiently overabundant G-IDA5, assistance in NAP1 polymerization, or both. Interestingly, IDA5 degradation in the prf1-1 mutant also depends on the LAT pathway, an observation that suggests several possibilities that are not mutually exclusive. First, the LAT pathway might be required for production of some factor(s) involved in targeting of IDA5 by the ubiquitin-proteasome system even in the absence of a major perturbation of the actin cytoskeleton. Second, the loss of profilin function in the prf1-1 mutant, with resulting loss of G- and F-IDA5, might itself constitute enough of a stress to the F-actin cytoskeleton to trigger activation of the LAT pathway, resulting in up-regulation of the ubiquitin-proteasome system and enhancement of IDA5 degradation. In this regard, it is of interest that our preliminary RNA-seq analysis of a prf1-1 mutant showed gene-expression changes with strong similarity to those seen in wild-type cells treated with LatB. To our knowledge, such a role of profilin in protecting actin from degradation has not been reported in other systems.

We hypothesize that IDA5 degradation in response to LatB eliminates nonpolymerizable G-IDA5–LatB complexes that could competitively interfere with NAP1 polymerization. This hypothesis is supported by our observations on a nap1 P_TUB2:NAP1 strain, which has lower levels of NAP1 than a wild-type strain whose endogenous NAP1 gene has been induced by LatB treatment. In the absence of other mutations, the nap1 P_TUB2:NAP1 strain showed only weak resistance to LatB. This resistance was further weakened when the degradation of IDA5 was reduced by a lat mutation, but it was restored to wild-type levels by an ida5 mutation that simply eliminates IDA5 protein.

Interactions of actin with the ubiquitin-proteasome system have also been observed in other organisms. For example, in mammals, mass-spectrometry studies found evidence that some of the many actin isoforms are ubiquitinated in vivo (45, 46), and some of the putative ubiquitination sites correspond to the lysines that are present in IDA5 but not in NAP1. In addition, a recombinant actin expressed in mouse muscle cells was found to be ubiquitinated and degraded (47), and in synapses, ubiquitination was found to regulate γ-actin levels (48) and thus add an important layer of regulation of the balance of actin isoforms during the establishment of neuronal connections (49, 50).

Summary: The Actin-Homeostasis System of Chlamydomonas.

In response to the insult to their F-IDA5–based actin cytoskeletons produced by LatB, Chlamydomonas cells up-regulate production of (i) an alternative actin, NAP1, that is insensitive to the drug; (ii) the ubiquitin-proteasome system, which degrades LatB-IDA5 complexes, probably to prevent them from interfering with the formation of F-NAP1; (iii) profilin, which appears to protect G-IDA5 from uncontrolled degradation and may promote F-NAP1 polymerization; and (iv) cofilin, which may promote the prompt removal of IDA5 by severing the LatB-sensitive filaments (Fig. 7). Remarkably, these responses together produce an F-actin–homeostasis system so efficient that a sudden exposure of wild-type cells to a high dose of LatB produces little or no interruption in their growth rate (20), while within the cells the entire complement of IDA5 has been depolymerized, degraded, and replaced with NAP1. This system seems likely to have evolved as a defense against compounds in the environment that damage the actin cytoskeleton (20), but the same system appears to operate also in the absence of drugs, for example when IDA5 protein (and thus F-IDA5) is lost in a profilin mutant. It will be interesting to see the extent to which this system operates also during other specific insults to the actin cytoskeleton in Chlamydomonas and to learn the degree to which the elements of this system (or analogous functions) are conserved in other organisms.

Materials and Methods

Strains.

C. reinhardtii wild-type strains CC-124 (mt−) and iso10 (mt+, congenic to CC-124; provided by S. Dutcher, Washington University in St. Louis, St. Louis) were the parental strains. Hygromycin-resistant (M11H; mt−) and paromomycin-resistant (P11P; mt+) derivatives were obtained by transformation of iso10 with aph7″ and APHVIII genes followed by several backcrosses (51). The lat1, lat2, lat3, nap1, prf1-1 (originally isolated as div68-1), and rpt5-1 (originally isolated as gex36-1) mutants used in this study were isolated previously in the CC-124 background (20, 34) and had been back-crossed at least twice with CC-124 or iso10 (or their drug-resistant derivatives). The cof1-1 mutation was originally isolated in a phenotypic revertant of a nap1-1 strain (20), which was back-crossed several times with M11H and M12P to obtain the cof1-1 single mutant; the mutation was then identified by bulked-segregant sequencing (34). The ida5-1 strains were derivatives of CC-3421 (obtained from the Chlamydomonas Resource Center) that had been back-crossed three times with CC-124 or iso10. ida5-1 had originally been isolated (as ida5) and shown to be a null mutant by Kato-Minoura et al. (15). For RNA-seq Experiment 5 (discussed below), a nap1-2 mutant that had been transformed with a P_TUB2:NAP1 construct (20) was crossed to iso10, and two NAP1⁺ and two NAP1⁺ P_TUB2:NAP1 segregants were used for the analyses.

Transcriptome Analysis.

We performed five transcriptome analyses by RNA-seq (Dataset S1), as follows. (i) CC-124 and an ida5-1 mt− strain were grown to OD₇₅₀ ∼0.3 at 21 °C, and LatB (weak lot) was added to a final concentration of 10 µM. Samples (10 mL) were collected before LatB addition and at intervals thereafter by brief centrifugation, and the pellets were frozen immediately in liquid nitrogen and stored at −80 °C. (ii) Like Experiment 1 except that strain CC-124 was treated with 10 µM LatB (weak lot), 10 µg/mL CHX, or both. In addition, a separate culture of CC-124 was shifted to 33 °C and incubated at that temperature for 6 h. (iii) M11H, P11P, ida5-1 mt−, ida5-1 mt+, lat1-1, lat1-5, lat2-1, lat2-2, lat3-1, lat3-3, nap1-1, and nap1-2 strains were treated as in Experiment 1, except that samples were added to tubes with crushed ice and centrifuged briefly; the pellets were then transferred to 2-mL cryogenic storage tubes and frozen at −80 °C. (iv) Like Experiment 3 but with samples taken over a more extended time course. (v) Two wild-type and two P_TUB2-NAP1 strains (all segregants from the same cross; discussed above) were grown to OD₇₅₀ ∼0.1 at ∼23 °C, and LatB (strong lot) was added to a final concentration of 3.0 µM. Samples (9 mL) were collected before LatB addition and at intervals thereafter as in Experiment 3.

To prepare samples for RNA-seq, prechilled glass beads, 500 µL phenol-chloroform, and 500 µL RNA extraction buffer (10 mM Tris⋅Cl, pH 8.0, 1 mM EDTA, 0.2 M NaCl, and 0.2% SDS) were added in this order to each frozen cell pellet, and the cells were disrupted by vigorous vortexing. After centrifugation for 15 min at 12,000 × g, the aqueous phase was recovered; RNA was then precipitated with ethanol, dissolved in RNase-free water, and further purified using the RNeasy mini kit (Qiagen) following the manufacturer’s instructions. RNA concentrations were measured using a NanoDrop (Thermo Scientific), and RNA integrity was assessed using an Agilent 2100 Bioanalyzer using the Plant RNA assay; all RNA-integrity numbers were between 7.1 and 9.9. cDNA libraries were generated from 1 µg total RNA per sample using a TruSeq RNA kit (Illumina) and sequenced (50-bp single-end reads) by Genewiz using an Illumina HiSeq 2500. Alignments of reads to the reference genome and read-count calculations were performed as described previously (52). For the heat-map displays, read counts for each gene were first corrected for the library size and then normalized against the value for wild type at time = 0.

For Experiment 2, differential-expression analysis was performed using EdgeR (53) and treating the 0-min and 10-min samples as one pair of replicates and the 90-min and 120-min samples as a second pair of replicates (Dataset S2). For Experiment 3, differential-expression analysis was performed using DESeq2 (54) and the two replicate samples for each genotype and time point examined (Dataset S3).

Smoothened density histograms were generated in R using the density function with default parameters.

Other Methods.

Growth conditions, plasmids, genetic and transformation methods, and the methods used for identification of the LRE, cross-species-homolog, and GO-term analysis, microscopy and plate-reader analyses, and Western blotting are described in SI Appendix, Materials and Methods.