Actin is one of the most conserved eukaryotic proteins. It is thought to have multiple essential cellular roles and to function primarily or exclusively as filaments (“F-actin”). Chlamydomonas has been an enigma, because a null mutation (ida5-1) in its single gene for conventional actin does not affect growth. A highly divergent actin gene, NAP1, is upregulated in ida5-1 cells, but it has been unclear whether NAP1 can form filaments or provide actin function. Here, we used the actin-depolymerizing drug latrunculin B (LatB), the F-actin-specific probe Lifeact-Venus, and genetic and molecular methods to resolve these issues. LatB-treated wild-type cells continue to proliferate; they initially lose Lifeact-stained structures but recover them concomitant with upregulation of NAP1. Thirty-nine LatB-sensitive mutants fell into four genes (NAP1 and LAT1–LAT3) in which we identified the causative mutations using a novel combinatorial pool-sequencing strategy. LAT1–LAT3 are required for NAP1 upregulation upon LatB treatment, and ectopic expression of NAP1 largely rescues the LatB sensitivity of the lat1–lat3 mutants, suggesting that the LAT gene products comprise a regulatory hierarchy with NAP1 expression as the major functional output. Selection of LatB-resistant revertants of a nap1 mutant yielded dominant IDA5 mutations that presumably render F-IDA5 resistant to LatB, and nap1 and lat mutations are synthetically lethal with ida5-1 in the absence of LatB. We conclude that both IDA5 and the divergent NAP1 can form filaments and redundantly provide essential F-actin functions and that a novel surveillance system, probably responding to a loss of F-actin, triggers NAP1 expression and perhaps other compensatory responses.
ACTIN is one of the most highly conserved proteins across the full range of the eukaryotic phylogeny (Figure 1, A and B; Hightower and Meagher 1986; Sheterline et al. 1999); for example, human, yeast, and higher-plant actins are all ∼90% identical in sequence. This sequence conservation suggests that actin also has highly conserved and critical functions. Indeed, since the demonstration that actin is essential in nonmotile budding yeast cells (Shortle et al. 1982), it has been widely presumed that it is essential for the survival and function of most, if not all, eukaryotic cells (although this appears to have been demonstrated rigorously in few, if any, other cases). Consistent with this view, actin has been shown to be involved in a diverse range of important processes, including vesicle transport and endocytosis, the determination of cell shape, cell motility, and cytokinesis (Korn 1982; Sheterline et al. 1999; Doherty and McMahon 2008; Balasubramanian et al. 2012). In all of these roles, actin is thought to function not as globular monomers (“G-actin”) but as filaments (“F-actin”), and filament formation is known to be tightly regulated, prominently by actin-binding proteins such as profilin, the formins, and the Arp2/3 complex (Doherty and McMahon 2008).
Many organisms contain multiple genes that encode actins with unknown degrees of actual or potential functional overlap; for example, mammals contain six (Perrin and Ervasti 2010), the slime mold Dictyostelium discoideum contains 33 (Joseph et al. 2008), and the plant Arabidopsis thaliana contains at least 10 (McDowell et al. 1996). Because of the difficulties that the potential redundancy of actin poses for genetic analyses, many studies of actin function have relied instead on polymerization-blocking drugs such as the cytochalasins and the latrunculins LatA and LatB. In particular, LatA and LatB are highly effective in depolymerizing most if not all actins that have been studied to date (Spector et al. 1983, 1989; Ayscough 1998). However, such drug studies have unavoidable ambiguity because of the possibility that not all actin family members are equally sensitive to the drug and the danger of unknown side effects.
Given this background, the unicellular green alga Chlamydomonas reinhardtii has presented an opportunity but also a paradox and a challenge. It has a single gene (IDA5) encoding a conventional actin (Figure 1, A and B), and localization studies have suggested that IDA5 may have multiple roles both during vegetative growth and during mating (Harper et al. 1992; Ehler and Dutcher 1998; Kato-Minoura et al. 1998; Kovar et al. 2001; Avasthi et al. 2014). However, electron microscopy and labeling with fluorescent phallotoxin detected F-actin structures only in fertilization tubules (Detmers et al. 1983; Wilson et al. 1997; Hirono et al. 2003), raising the surprising possibility that IDA5 might function during vegetative growth as G-actin rather than in filaments, as originally suggested by Harper et al. (1992). In another surprise, a screen for mutants defective in swimming behavior and flagellar structure yielded a null mutation in IDA5 (originally ida5, henceforth ida5-1; a frameshift resulting in a complete loss of IDA5 protein) (Kato et al. 1993; Kato-Minoura et al. 1997). This mutation has essentially no effect on cell proliferation, although it does affect intraflagellar transport, flagellar regeneration, and fertilization-tubule formation (Kato et al. 1993; Kato-Minoura et al. 1997; Avasthi et al. 2014). Moreover, although very high doses of cytochalasin D caused temporary shortening of the flagella (Dentler and Adams 1992), they had no effect on proliferation (Harper et al. 1992).
These observations have suggested that Chlamydomonas cell-shape determination, vesicle transport and endocytosis, and cytokinesis might be completely independent of F-actin function. However, interpretation of this surprising hypothesis has been complicated by the presence in Chlamydomonas of a gene (NAP1) that encodes an unconventional actin (Kato-Minoura et al. 1998). Although NAP1 is only ∼65% identical in amino acid sequence to IDA5 and other conventional actins (Figure 1B), phylogenetic analyses consistently cluster it together with actins rather than with Arp1-type actin-related proteins (Figure 1A; Kato-Minoura et al. 2014). NAP1 is transcriptionally upregulated in the ida5-1 null mutant (Kato-Minoura et al. 1998; Hirono et al. 2003), so that it has been speculated that it could provide actin function under these conditions. A priori, the extensive sequence divergence of NAP1 from conventional actins provides a challenge for this model. Moreover, it is unclear whether NAP1 can polymerize into filaments: no assay has been available to examine this question in Chlamydomonas cells, and unlike conventional actins, the protein has been reported to polymerize poorly or not at all when expressed in mammalian cells (Kato-Minoura 2011). Assessment of possible NAP1 function has been blocked by the absence of a nap1 loss-of-function mutant and a lack of information about the possible sensitivity of NAP1 to latrunculins or other drugs.
In a previous study, we used the F-actin-specific probe Lifeact (Riedl et al. 2008) to obtain evidence that wild-type (WT) Chlamydomonas cells contain LatB-sensitive F-actin filaments (Avasthi et al. 2014). We have now further examined the effects of LatB and performed genetic analyses beginning with the isolation of LatB-sensitive mutants. Our results suggest that IDA5 and NAP1 can redundantly provide essential F-actin functions and that Chlamydomonas cells possess a surveillance system that actively responds to insults to their actin cytoskeletons through transcriptional upregulation of the actin genes. These studies provide the tools for future in-depth analysis of actin function in Chlamydomonas, a key microbial representative of the plant superkingdom.
Materials and Methods
Strains and growth conditions
C. reinhardtii strains CC-124 (mt−) and iso10 (mt+) (congenic to CC-124; S. Dutcher, personal communication) were the parental strains. Hygromycin- and paromomycin-resistant derivatives were obtained by transformation with PCR-amplified aph7” and APHVIII genes. Mutants were either isolated in the CC-124 background and crossed to drug-resistant derivatives of iso10 or isolated directly in a drug-resistant background. ida5-1 strains CC-3420 (mt+) and CC-3421 (mt−) were obtained from the Chlamydomonas Resource Center.
Cells were grown in Tris-acetate-phosphate (TAP) medium (Gorman and Levine 1965) at 21° or 24° under constant illumination at 50–100 μmol photons m−2⋅s−1. LatB was purchased from Adipogen (AG-CN2-0031, lot no. A00143/G), and dilutions into TAP medium were made from a 10-mM stock in DMSO. Paromomycin (Sigma or EMD Millipore) and hygromycin (Sigma) were used at 10 µg/ml.
Plasmid construction and transformation by electroporation
Construction of plasmids pMO459 (PHSP70A/RBCS2:Lifeact-Venus-3FLAG:AphVIII:TRBCS2) and pMO448 (PHSP70A/RBCS2:APHVIII:TRBCS2) will be described elsewhere (M. Onishi and J. R. Pringle, unpublished results). Plasmid pMO524 (PHSP70A/RBCS2:AphVIII:TRBCS2) was constructed by PCR using pMO448 as template and primers MOP654 and MOP855 (primer sequences provided in Supporting Information, File S1), followed by a self-linearization of the PCR product by the one-step isothermal assembly method (Gibson et al. 2009) to remove the first exon and intron of RBCS2 from pMO448. Plasmids pMO529 (PHSP70A/RBCS2:NAP1:AphVIII:TRBCS2) and pMO531 (PTUB2:NAP1:AphVIII:TRBCS2) were constructed in two steps. First, two DNA fragments that together covered the complete NAP1 sequence were amplified from wild-type genomic DNA using primers MOP856 and MOP719 (fragment 1) and MOP726 and MOP857 (fragment 2) and assembled with HpaI/NdeI-digested pMO524. From the resulting plasmid, a DNA fragment containing NAP1:AphVIII was excised using HpaI and BamHI and then ligated into the same sites of pMO448 and pMO508 (M. Onishi and J. R. Pringle, unpublished results). Expression cassettes were excised with EcoRV; after heat inactivation of the enzyme (65°, 10 min), the digests were used for transformation by electroporation as described previously (Yamano et al. 2013), but using CHES buffer [10 mM N-cyclohexyl-2-aminoethanesulfonic acid, pH 9.25, 40 mM sucrose, 10 mM sorbitol]. A NEPA21 square-pulse electroporator (Bulldog Bio) was used, with two poring pulses of 250 V and 150 V for 8 ms each and five transfer pulses starting at 20 V with a decay rate of 40% for 50 ms each. Transformants were selected on TAP agar + 10 µg/ml paromomycin.
Isolation of mutants and genetic analysis
UV mutagenesis was performed essentially as described by Tulin and Cross (2014). Cells were spread on TAP-agar plates and irradiated with a germicidal lamp in a tissue-culture hood, with inocula and times of irradiation determined empirically to yield ∼300 viable colonies per plate. Irradiated plates were kept in the dark at room temperature for ∼18 h to prevent light-dependent repair of UV lesions, then incubated in the light at 21° for ∼1 week. Colonies were picked to a 16 × 24 array with a Hudson Robotics colony picker. Grown arrays were condensed to a 32 × 48 array using a Singer RotoR replica-plating robot, then replicated to TAP or TAP + LatB (10 µM), incubated in the light at 21° for 5–7 d, and photographed using an Olympus digital camera. The resulting JPEG images were processed with custom MATLAB software for background subtraction and alignment to a 32 × 48 grid. Differential pixel intensities with and without LatB were calculated and used for the initial selection of LatB-sensitive candidates, which were then repicked using the Singer RotoR (Stinger attachment) and retested for LatB sensitivity.
Complementation, linkage analysis, and genetic crosses for strain construction were performed essentially as described previously (Dutcher 1995; Tulin and Cross 2014). Segregant growth was recorded microscopically at 100× with an AmScope MD500 camera. In most complementation and linkage experiments, hygromycin-resistant and paromomycin-resistant strains were mated, so that mating and meiosis could be confirmed by recovery of doubly resistant diploid or haploid progeny. For bulk analysis of meiotic progeny (without tetrad dissection), unmated haploids and stable nonzygospore diploids were removed by scraping zygospore plates with a clean razor blade before inducing meiosis by incubation in the light.
Isolation of nap1-1 revertants
TAP agar plates with ∼107 nap1-1 mutant cells per plate were UV-irradiated for varying amounts of time, incubated in the dark for ∼18 h, and returned to the light as described above. Cells from the resulting lawns were resuspended, plated at ∼107 cells per plate on LatB-TAP agar (10 µM LatB), and incubated at 21° until colonies appeared (∼1 week). Colonies were picked and retested for LatB resistance, and the NAP1 and/or IDA5 transcripts were amplified from resistant colonies by RT-PCR and sequenced as described below.
Identification of mutations by Illumina sequencing of pooled mutants and bulked segregants
For identification of the causative mutations, cultures of the initial 14 mutants were combined in five overlapping pools of seven or eight mutants each, such that each mutant was present in a unique “pool pattern” of two or three pools. These pools were processed for Illumina library preparation using reagents from New England Biolabs and oligonucleotides from Illumina; the five libraries were then sequenced at the New York Genome Center to ∼20–25× coverage (∼3× coverage for each mutant in each pool; this coverage was low but proved to be sufficient for mutant detection in almost every case). The Illumina output files (.fastq) were aligned to the Chlamydomonas reference genome (Merchant et al. 2007; file Creinhardtii_281_v5.0.softmasked.fa.gz; http://genome.jgi.doe.gov). The Bowtie2.bam output was processed with Samtools mpileup (Li et al. 2009), and the mpileup output was processed with custom MATLAB software to compress the data into matrices (rows, genome positions; columns, read counts for each call). These matrices allow sensitive detection of low-frequency alternative (mutant) calls at specific positions by comparison of the five pools to libraries from parental strains. The putative mutations were filtered for quality (base-call quality at SNP position and across the read, number of mismatches compared to reference, and mapping quality; these quality features are semi-independent) to include only high-quality calls predicted to alter coding potential (either coding-sequence change or disruption of intronic dinucleotides essential for splicing) based on the Chlamydomonas genome annotation (Blaby et al. 2014) (Creinhardtii_281_v5.5: annotation_info.txt, defline.txt, description.txt, gene_exons.gff3, from http://genome.jgi.doe.gov).
These low-frequency mutations were then assigned to the original mutant strains based on pool profile (e.g., because mutant 16 contributed to pools 1, 3, and 5, and no other mutant had this pattern, a mutation found only in libraries 1, 3, and 5 was assigned to mutant 16). Because the mutagenesis conditions resulted in hundreds to thousands of mutations that could be assigned to each mutant strain, we then searched computationally for gene models mutated (according to pool pattern) in all members of each genetic complementation group (see Results). For all four complementation groups, exactly one such putatively causative gene model was found (with minor discrepancies due to suboptimal coverage; see Results). (Note that we use the term “gene model” to reflect the hypothetical status of the functional units in the genome annotation, many of which were assigned largely or entirely on the basis of computation; Blaby et al. 2014).
In some cases, the initial identification of the mutations putatively causing LatB sensitivity was made or confirmed by sequencing ≥10 bulked LatB-sensitive segregants from crosses of mutant to wild type, as described by Tulin and Cross (2014).
Sanger sequencing and allele-specific PCR
For the seven mutants assigned to the lat4 (i.e., nap1) complementation group, confirmation (two cases) or initial identification (five cases) of the putatively causal mutations was performed by PCR amplification of NAP1 using primers MOP874 and MOP721 followed by Sanger sequencing (Table 1). To search for mutations in NAP1 and IDA5 in the nap1-1 phenotypic revertants, total RNA was extracted from cells treated with 10 µl LatB for ∼2 hr, and the transcripts were amplified by RT-PCR using primers MOP874 and MOP721 for NAP1 and MOP875 and MOP876 for IDA5. The products were then purified and used for Sanger sequencing with the same primers.
To follow segregation of mutations in crosses, we used allele-specific PCR (Gaudet et al. 2009). For each mutation to be tested, three primers were designed (File S1). The MUT forward primer has a 3′ end that is complementary to the mutant sequence except for a single-nucleotide mismatch at −3 from the 3′ end; this primer also has 20 nucleotides of random sequence at its 5′ end. The WT forward primer has a 3′ end that is complementary to the wild-type sequence except for a mismatch at the same position as the MUT primer but to a different nucleotide. The common reverse primer is complementary to sequence ∼100 nucleotides from the mutant site. The three primers were used together in a single PCR reaction using either purified genomic DNA or crude cell extract as template, and the products were analyzed by electrophoresis using 3% agarose gels in sodium borate buffer (Brody and Kern 2004). Products of 120-bp and 100-bp are expected for the mutant and wild-type alleles, respectively (see Figure 9B). Purification of genomic DNA was done using the DNeasy Plant Mini kit (Qiagen) following the manufacturer’s instructions. Crude cell extracts were prepared by adding ∼2.5 µl of packed cells to 50 µl of 10 mM EDTA, pH 8.0, incubating at 105° for 15 min, and centrifuging at 12,000 × g for 1 min to remove cell debris; 1 µl was then used for the PCR reaction. Both methods worked equally well for the allele-specific PCR.
Analysis of gene expression
Total RNA was extracted from frozen cells by adding to the frozen pellet, in the following order and without mixing: ∼400 µl acid-washed glass beads, 500 µl of phenol-chloroform plus 500 µl of extraction buffer (10 mM Tris-Cl, pH 8.0, 1 mM EDTA, 0.2 mM NaCl, 0.2% SDS), followed by vigorous vortexing and centrifugation. RNA was then precipitated from the aqueous phase with 1 ml of ethanol, dissolved in 50 µl of distilled H2O, and further purified using the RNeasy Mini kit (Qiagen) following the manufacturer’s instructions. [This protocol, based on one developed for yeast RNA (Cross and Tinkelenberg 1991), gave RNA preparations of greater integrity than we obtained with Trizol extraction.] A total of 0.5 µg of total RNA was used for reverse transcription using the Maxima RT-PCR First Strand cDNA Synthesis kit (Thermo Fisher Scientific; for Figure 3) or the iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad; for Figure 6), following the manufacturers’ instructions.
For evaluation of transcript levels by electrophoresis, the cDNA was subjected to PCR using primer sets (File S1) specific for IDA5 (MOP875 and MOP876; from 100 nucleotides upstream of the start codon to 86 nucleotides downstream of the stop codon) or NAP1 (MOP874 and MOP721; from 100 nucleotides upstream of the start codon to 115 nucleotides downstream of the stop codon) followed by agarose-gel electrophoresis.
For qPCR, the same cDNA was used with gene-specific primers (File S1) and the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) in a CFX384 Touch Real-Time PCR Detection System (Bio-Rad) with the following parameters: an initial denaturation at 95° for 30 sec, then 40 cycles of 95° for 15 sec and 55° for 15 sec. Melting-curve analysis was performed using the instrument’s default setting to verify amplification of single DNA species. For each analysis, the expression levels of IDA1 and NAP1 were normalized first against an internal reference and then to the value obtained for wild-type cells in the absence of LatB treatment. All reactions were performed in triplicate with at least two biological replicates or repetitions. For a control gene, we used NUOB13 (Cre13.g568800.t1.2; NADH:ubiquinone oxidoreductase 18-kDa subunit), because NormFinder analysis (Andersen et al. 2004) of a preliminary RNA-seq dataset for wild-type cells treated with LatB indicated that it was expressed at an essentially constant level independent of LatB treatment (data not shown).
Live Lifeact-Venus-expressing cells were mounted on a thin pad of TAP medium containing 1–1.5% low-melting agarose (SeaPlaque; FMC Corporation) and sealed with a coverslip. The cells were observed using a Nikon Eclipse 600-FN microscope equipped with an Apochromat ×100/1.40 N.A. oil-immersion objective lens, an ORCA-2 cooled CCD camera (Hamamatsu Photonics), and Metamorph version 7.5 software (Molecular Devices). Images were postprocessed using ImageJ (National Institutes of Health) and Photoshop (Adobe) software. Images from a single experiment with a single strain were processed identically and can be compared directly. However, because of the variable expression levels of Lifeact-Venus among strains and cultures, the brightness of images cannot be directly compared across different strains or experiments.
Phylogenetic analysis and sequence comparisons
A phylogenetic tree for 13 actin proteins and human ARP1 was generated by a neighbor-joining algorithm and a bootstrap test with 10,000 iterations using a MUSCLE alignment file as input in the Geneious software version 7.0.3 (Kearse et al. 2012). The sequences used a wide phylogenetic spectrum: AAA51577 (Homo sapiens), XP_001704653 (Giardia lamblia), NP_116614 (Saccharomyces cerevisiae), XP_001699068 (C. reinhardtii IDA5), AAA34243 (Volvox carteri IDA5), XP_001703266 (C. reinhardtii NAP1), XP_002946759 (V. carteri NAP1), NP_187818 (Arabidopsis thaliana), AAA30152 (Trypanosoma brucei), P10992 (Tetrahymena thermophila), XP_002369663 (Toxoplasma gondii), P27132 (Naegleria fowleri), CBJ30601 (Ectocarpus siliculosus), and NP_005727 (H. sapiens ARP1).
Strains, plasmids, plasmid sequences, Illumina sequencing data, and MATLAB code are available upon request. File S1 contains sequences of all DNA oligonucleotides used in this study.
Upregulation of NAP1 expression and formation of latrunculin-resistant F-actin-like structures in LatB-treated cells
One possible explanation for the surprising absence of a growth phenotype in the ida5-1 null mutant is that the upregulated expression of NAP1 in the mutant (Kato-Minoura et al. 1998) allows this unconventional actin to provide any essential actin functions (see Introduction). In that case, it might be possible to investigate actin function by treating wild-type cells with LatB, a potent inhibitor of polymerization for most if not all actins that have been tested to date (see Introduction). In short-term experiments, LatB was effective in depolymerizing the Lifeact-detected filaments, allowing clarification of the role of actin in regulation of flagellar length (Avasthi et al. 2014; Figure 2A, 10 min). Surprisingly, however, Lifeact-detected structures reappeared after 30–60 min of drug treatment (Figure 2A), and these structures were resistant even to a 10-fold higher concentration of LatB (Figure 2B). Correspondingly, LatB had little if any effect on the proliferation of wild-type Chlamydomonas cells (Figure 2C). These observations might be explained if LatB treatment induces (over)production of one or more actin-binding proteins whose interaction with IDA5-based filaments stabilizes them against the effects of the drug. However, examination of the ida5-1 null mutant revealed the presence of F-actin-like structures that were also largely resistant even to a high concentration of LatB (Figure 2D).
To explain these results, we hypothesized that LatB-treated wild-type cells, like ida5-1 mutant cells, upregulate NAP1; NAP1 might then form LatB-resistant F-actin-like structures. Indeed, we found an ∼70-fold upregulation of NAP1 in LatB-treated wild-type cells, considerably higher than that observed in ida5-1 mutant cells (Figure 3, A and B; note log scale in B). Interestingly, the expression of IDA5 was also upregulated ∼5-fold in the LatB-treated wild-type cells (Figure 3, A and B), suggesting that Chlamydomonas has a regulatory system that attempts to compensate for damage to its F-actin cytoskeleton by upregulating the production of both its conventional and unconventional actins.
Isolation and initial genetic analysis of LatB-sensitive mutants
The observations described above suggested that actin functions can be carried out redundantly by IDA5 and NAP1, as speculated previously (Kato-Minoura et al. 1998; see Introduction). However, the absence of any means to inactivate NAP1 has prevented a direct test of this idea. If actin function is essential for Chlamydomonas, then a screen for LatB-sensitive mutants might yield mutations in genes specifically essential for NAP1 expression or function (including NAP1 itself). Because NAP1 is expressed little or not at all in wild-type cells (Figure 3A), a mutation eliminating its function should have little or no effect in cells not treated with LatB. Accordingly, we mutagenized wild-type Chlamydomonas cells with UV as described previously (Tulin and Cross 2014) and screened mutants for inability to proliferate on agar medium containing 10 µM LatB (see Materials and Methods). An initial screen yielded 14 strongly LatB-sensitive (“lat”) mutants. These mutants had no significant proliferation defect in the absence of LatB (Figure 4A, left) but failed to grow on 10 µM LatB (Figure 4A, right) and displayed at least partial sensitivity to 3 or 1 µM LatB, whereas wild-type cells could grow even on 100 µM LatB (data not shown). The mutants were insensitive to 0.1% DMSO (the LatB carrier) and showed little or no alteration in sensitivity to other drugs (anisomycin, rapamycin, amiprophos-methyl; data not shown), suggesting that the lat mutants have alterations specifically affecting LatB response.
We crossed the 14 mutants to wild type and analyzed tetrads. In each case, LatB sensitivity segregated 2:2, indicating that a single genetic locus was responsible for the phenotype. Complementation tests carried out on these mutants and others isolated subsequently indicated that they defined four complementation groups (Figure 4, B and C): lat1 (14 mutants), lat2 (7 mutants), lat3 (11 mutants), and lat4 (6 mutants). In most cases, we confirmed by testing bulk meiotic products that noncomplementing mutant pairs also yielded only LatB-sensitive haploid meiotic progeny, whereas complementing mutant pairs yielded a mix of LatB-resistant and sensitive haploid progeny. One additional LatB-sensitive mutant was strongly dominant (data not shown), precluding complementation analysis. The causative mutation was found to be closely linked to LAT4 based on a lack of LatB-resistant segregants after crosses to lat4-1 (mutant 16) and lat4-2 (mutant 18) testers, and sequencing subsequently showed that the strain indeed contained a LAT4 mutation (see below).
Tetrad analysis indicated that lat1, lat3, and lat4 are all unlinked to each other, whereas lat1 and lat2 are linked at a distance of ∼18 cM [9 parental ditype (PD):0 nonparental ditype (NPD):5 tetratype (T), significantly different from the expectation of PD = NPD for unlinked genes; P < 0.003 by chi-square], placing them on the same chromosome, but presumably separated by several megabases, given the genome average of 10 cM/Mb (Merchant et al. 2007). These crosses generated the double lat mutants (confirmed by complementation testing) at the expected frequencies. The double mutants had no significant growth defects in the absence of LatB, and LatB sensitivity was approximately similar for double and single mutants (data not shown). These results suggest that the four LAT genes contribute largely or entirely to a single LatB-resistance pathway and that this pathway is essential in the presence of LatB but dispensable under normal conditions.
Taken together, these data suggest that the screen was close to saturation, such that most or all genes and pathways required specifically for cell proliferation in the presence of 10 µM LatB have been identified.
Identification of NAP1 and three novel LAT genes as sites of the lat mutations
The mutagenesis conditions used result in up to 100 coding-sequence changes in each survivor (Tulin and Cross 2014). To identify the specific causative mutations leading to LatB sensitivity, we used Illumina sequencing on DNA from overlapping pools of the original 14 mutants, where each mutant was represented in a distinct set of two or three pools. All mutations specific to a given mutant should then be detected as low-frequency alternative calls in a unique pattern of pools (see Materials and Methods). Indeed, most such mutations did align with one of the predicted pool patterns (data not shown).
Of the many mutations aligning with a particular pool pattern, only one was expected to be causative, because LatB sensitivity was in each case due to a single locus (see above). To identify these causative mutations, we used the gene assignments obtained by complementation and linkage testing. A random, noncausative mutation should almost always fall into a gene model that was hit only in a single mutant strain, whereas a mutation responsible for the LatB-sensitive phenotype should fall into a gene model that was hit in each mutant assigned to a specific complementation group. Following this logic, we tentatively identified LAT1 as Cre10.g464550 [five of six lat1 mutants contained mutations that would change the coding sequence in this gene model, although in one case (mutant 8, Table 1), it was necessary to assume that one pool had failed to yield the alternative read due to stochastic error (see Discussion)], LAT2 as Cre10.g438250 (three of three mutants), LAT3 as Cre02.g076000 (two of three mutants), and LAT4 as NAP1/Cre03.g176833 (two of two mutants) (Table 1). The sixth putative lat1 mutant (mutant 35; lat1-6) did not show a predicted coding-sequence mutation in Cre10.g46455, but a single-nucleotide deletion was detected with the correct pool pattern in what is currently annotated as the 5′-untranslated region of this gene model (Table 1). It seems likely that this mutation is actually an N-terminal frameshift in the coding sequence, because there is an in-frame ATG without intervening stop codons in the 5′-untranslated region as currently annotated. However, we cannot yet rule out the possibility that the mutation is indeed in the 5′-untranslated region but has a deleterious effect on translation. In the third lat3 mutant (mutant 17; lat3-1), a mutation in Cre02.g076000 that was missed in the pool sequencing was identified subsequently by bulked-segregant sequencing (see Materials and Methods; Table 1). The ∼3.5-Mb separation of Cre10.g464550 and Cre10.g438250 on chromosome X is approximately consistent with the ∼18-cM genetic distance determined for lat1 and lat2 by tetrad analysis with a small number of tetrads (see above).
We used three different approaches to confirm the identification of the causative mutations (Table 1): bulked-segregant sequencing, PCR amplification and Sanger sequencing of the gene of interest, and an allele-specific PCR strategy (see Materials and Methods). Sanger sequencing was also used to show that all six mutants assigned to the LAT4/NAP1 complementation group indeed carried NAP1 mutations (as did the dominant NAP1-5 mutation) (Table 1). Bulked-segregant sequencing and allele-specific PCR on tetrads also confirmed that the putatively causative mutations cosegregated with LatB sensitivity (Table 1).
The mutations identified include at least one presumptive null allele for each gene (Figure 1; Figure 5; Table 1), including premature stop codons and/or mutations in the essential /GT….AG/ splice-site dinucleotides. Other mutations caused amino acid substitutions that were ranked as “severe” because of Blosum62 (Henikoff and Henikoff 1992) scores ≤ −2. For LAT1 and NAP1, the missense mutations also altered residues that are conserved between the predicted C. reinhardtii and A. thaliana proteins based on BLAST alignment. (We identified no missense mutations in LAT3, and the LAT2 product could not be aligned to any Arabidopsis protein.) In a broad range of Chlamydomonas Ts− lethal mutants, the causative mutations were up to 100 times more likely than random “passenger” mutations to be severe mutations in residues conserved in alignments to Arabidopsis (Tulin and Cross 2014); the findings with the LAT mutations confirm and extend these findings.
Mutations in the splice-site /GT…AG/ dinucleotides almost always eliminate use of the splice junction, but alternative splice junctions are often utilized instead (Brown 1996). Thus, as the splice-site mutation in nap1-2 was the strongest candidate for a null allele of NAP1 (Figure 1B; Table 1), we examined its protein-coding consequences by amplifying the nap1-2 transcript using RT-PCR with primers that anneal at the translation start site and in the 3′-UTR (165 nucleotides downstream of the normal stop codon). Sequencing demonstrated a 157-nucleotide deletion after exon 6, presumably by alternative splicing using the sequence ACGCAG/GGCAUC in exon 8 as an acceptor site. This alteration should cause a frameshift and synthesis of a 269-amino acid protein that lacks 120 amino acids from the normal C terminus of the 380-amino acid NAP1 (see Figure 1B).
Sequences of the LAT1, LAT2, and LAT3 proteins
The 1546-amino acid LAT1 has a protein-kinase domain near its C terminus (Figure 5A) that has relatively strong similarity to MAPKK kinases from a variety of organisms. Four of the six lat1 mutations specifically alter this domain, suggesting that it is important for LAT1 function. The 507-amino acid LAT3 also has a protein-kinase domain that is altered or truncated in all three lat3 mutants (Figure 5B). For both LAT1 and LAT3, BLAST searches using query sequences outside of the protein-kinase domains found no significant similarities to proteins from other organisms and thus provided no clues to their functions.
BLAST searches using either the entire 1065-amino acid LAT2 or portions of it as query sequences found similarities only to a small gene family in Chlamydomonas itself (five genes in tandem on chromosome X plus two on chromosome VI; Figure 5C) and in its close relative Volvox (two genes; Figure 5C). The single lat2 missense mutation that we characterized falls within the conserved domain, although the residue affected (N304) is not itself a conserved one. No function has previously been assigned to these genes.
Roles of LAT1, LAT2, and LAT3 in NAP1 induction in response to LatB
The observations that wild-type NAP1 is required for resistance to LatB and that NAP1 expression is strongly induced by LatB suggested the possibility that the other LAT proteins might be required for NAP1 upregulation. Indeed, presumed null mutations of LAT2 and LAT3 eliminated detectable NAP1 induction in response to LatB (Figure 6, A and B), and the lat1-5 mutation reduced it by ∼18-fold (Figure 6B). nap1 mutations allowed induction by LatB to a level comparable to that seen in wild type (Figure 6, A and B). As expected from the splicing defect in nap1-2 (see above and Table 1), the cDNA from this strain migrated faster than that from wild type, nap1-1, or nap1-3 strains (Figure 6A). IDA5 mRNA levels were also induced as in wild type in all lat mutants (Figure 6B). Thus, the defective NAP1 induction in the lat1, lat2, and lat3 mutants does not appear to be due to nonspecific transcriptional defects following loss of actin function.
If the LatB sensitivity of the lat1, lat2, and lat3 mutants results solely from their inability to induce NAP1, they should be rescued by expression of NAP1 from a constitutive promoter. To test this possibility, we made constructs with heterologous promoters expressing both NAP1 and APHVIII (paromomycin resistance) from a bicistronic mRNA (Figure 7A) (M. Onishi and J. R. Pringle, unpublished results). In contrast to nap1-2 cells transformed with a control construct that expresses only APHVIII (Figure 7B, pMO524), most paromomycin-resistant transformants obtained with the NAP1-expressing constructs were resistant to 5 µM LatB (Figure 7B, pMO529 and pMO531), confirming the function of the NAP1 transgene.
We then crossed a LatB-resistant nap1-2 transformant to lat1-5, lat2-1, and lat3-1 strains and analyzed tetrads for paromomycin and LatB resistance. All of the paromomycin-resistant (i.e., transgene-containing) segregants were resistant to LatB at concentrations up to 5 µM, and genotyping by allele-specific PCR showed that about half of these segregants contained the mutant lat alleles (Figure 7C; Table 2), indicating that ectopic expression of NAP1 can indeed largely suppress the lat1, lat2, and lat3 mutations. This result suggests that LAT1, LAT2, and LAT3 function in a common pathway leading to NAP1 expression, consistent with the failure to observe synergistic LatB sensitivity in lat double mutants (see above). Surprisingly, however, although the suppressed mutants grew nearly as well as wild type at lower concentrations of LatB (Figure 7C, top), they grew poorly or not at all on 10 µM LatB (Figure 7C, bottom). These observations suggest that LAT1, LAT2, and LAT3 must also have at least one function in addition to supporting the upregulation of NAP1 expression; for example, they might be important also for the expression (or upregulation) of some other gene(s) that is required for the development of full resistance to LatB (see Discussion).
Dependence of LatB-resistant F-actin-like structures on NAP1, LAT1, LAT2, and LAT3
The disappearance, then reappearance, of F-actin-like structures upon LatB treatment of wild-type cells (Figure 2A) could represent a depolymerization of LatB-sensitive F-IDA5 filaments followed by formation of LatB-resistant F-NAP1 filaments upon NAP1 upregulation. In this case, treatment of the nap1 and other lat mutants with LatB should result in a permanent disappearance of Lifeact-labeled F-actin-like structures. This was indeed observed (Figure 8). The nap1-1 mutant used here is a missense mutation (G371K) in a conserved residue near the C terminus (Figure 1; Table 1); the mutant protein may be deficient in polymerization or have lost its normal LatB resistance.
Synthetic lethality of nap1 and other lat mutants with ida5
If the primary effect of LatB in wild-type cells is indeed to depolymerize F-IDA5 filaments, and if upregulated NAP1 rescues the F-actin functions, then the nap1 and other lat mutants should also be sensitive to a genetic loss of IDA5. To test this, we crossed representative lat mutants to ida5-1 null-mutant strains and performed tetrad analysis. In all cases, the results indicated synthetic lethality between two loci (approximately one of four segregants inviable: Figure 9A; Table 3), and allele-specific PCR and tests of LatB sensitivity indicated that the two loci were ida5-1 and the lat mutation, as all viable spores were LatB-sensitive lat single mutants, LatB-resistant ida5-1 single mutants, or wild type at both loci (Figure 9B; Table 3). Interestingly, although the ida5-1 nap1-2 segregants grew little or not at all, the ida5-1 lat1-5, ida5-1 lat2-1, and ida5-1 lat3-1 segregants appeared to grow and divide normally for 2–3 d before ceasing growth and lysing during the next 24 hr (Figure 9A; data not shown). This delayed death may reflect the presence of parental LAT mRNAs or LAT proteins (or possibly of NAP1) that takes some days to be degraded and thus delays expression of the mutant phenotype.
Isolation of new IDA5 alleles as suppressors of nap1-1
We subjected a nap1-1 strain to UV mutagenesis (see Materials and Methods) and selected phenotypic revertants capable of growing on 10 µM LatB. We anticipated three possible types of mutation: true reversion or a compensatory second-site mutation in the NAP1 gene itself; mutation of IDA5 to yield an IDA5 protein that was latrunculin resistant; or mutations of other genes that somehow enhanced LatB resistance by some parallel pathway. For an initial set of 14 LatB-resistant isolates, we used RT-PCR and Sanger sequencing to sequence the NAP1 transcript from cells treated with 10 µM LatB for ∼2 hr to induce NAP1 expression. All strains still contained the original nap1-1 GG→AA mutation (Table 1), and no new mutations in NAP1 were found. We then used RT-PCR and Sanger sequencing to sequence the IDA5 transcript from these 14 strains plus an additional 15 LatB-resistant isolates. In 28 of the 29 cases, one or more new mutations were found in IDA5 (Figure 1; Table 4). In some cases, these mutations altered residues that are near the presumed LatB-binding site (based on observations on yeast and mammalian actins), either in the primary sequence (IDA5-10, -16, and -17: Figure 1B) or in the 3D structure [IDA5-3 through IDA5-9: Figure 1B and Protein Data Base (PDB) 1IJJ]. LatB resistance in all of the new IDA5 mutants was dominant (Table 4), suggesting that these mutations reduce or abolish the affinity of IDA5 for latrunculin but do not affect the protein’s intrinsic polymerization or depolymerization rate. In the 29th phenotypic revertant, no mutation in IDA5 was found, and a cross to wild type indicated that the strain contained a recessive suppressor unlinked to nap1-1 (Table 4); this suppressor gene has not yet been identified. Taken together, these data provide strong genetic evidence that IDA5 is the normal target of LatB (resulting in LatB sensitivity in nap1-1 mutants). The new mutations may reduce the affinity of IDA5 for LatB, thus rendering the F-IDA5 structures resistant to the drug.
Cellular response to simultaneous inactivation of IDA5 and NAP1
Wild-type or ida5-1 cells treated with LatB exhibited few or no persistent morphological abnormalities and proliferated with little or no delay (Figure 2; Figure 10, A and B). However, although the lat1, lat2, lat3, and nap1 mutant cells grew normally in the absence of LatB (Figure 4A; Figure 10, A and B), they exhibited an essentially immediate and complete block to cell proliferation upon treatment with LatB. We examined this behavior more closely for the presumed null nap1-2 allele.
The Chlamydomonas cell cycle is separated into a long growth phase, during which cells can increase in size by >10-fold, followed by a rapid series of cell divisions; entry into cell division, and the number of divisions carried out, is dependent on cell size (reviewed by Cross and Umen 2015). We obtained a partial synchronization of wild-type and nap1-2 cells by inoculation into low-nitrogen medium; upon depletion of the medium, such cultures consist mostly of small newborn cells. These cells were then refed with medium containing normal nitrogen levels, initiating cell growth. Cell sizes increased with little or no cell division for ≥12 hr, and in the subsequent 12 hr, almost all cells underwent multiple division cycles. nap1-2 cells plated on LatB early in the growth cycle (4 hr after refeeding, while cells were still small) showed little increase in cell size and no cell divisions over the succeeding 22 hr, whereas wild-type controls with or without LatB, or nap1-2 cells without LatB, all grew and carried out multiple division cycles in this time (Figure 10A). This defect could be due to a direct requirement for IDA5 or NAP1 for division; alternatively, the cells could fail division as a secondary consequence of failing to reach a critical cell size (Cross and Umen 2015). To test this, we examined wild-type and nap1-2 cells plated on LatB 12 hr after refeeding, when they were large and would normally divide within a few hours. Plating on LatB almost completely blocked division in nap1-2 cells but had no effect on wild type (Figure 10B). A small minority of nap1-2 cells did appear to enter the division cycle, but these divisions appeared morphologically abnormal and did not result in completion of division (data not shown). Flow cytometry suggested that little if any DNA replication occurred after LatB addition to nap1-2 cells (data not shown). Taken together, these results suggest that IDA5 and NAP1 are redundantly required for cell growth; there may also be a specific IDA5/NAP1 requirement for cell division.
Overlapping-pool sequencing and allele-specific PCR
Several methodological aspects of this study deserve comment. First, the novel overlapping-pool sequencing strategy that we used greatly facilitated the identification of the causative mutations in our collection of LatB-sensitive mutants. Other recent studies have used high-throughput, whole-genome sequencing either of individual mutants after prior meiotic mapping (Dutcher et al. 2012; Lin et al. 2013) or of bulked segregants after backcrossing (Alford et al. 2013; Tulin and Cross 2014) to localize the mutations of interest to genomic regions of 0.3–2 Mb, and in many cases to identify the actual causative mutation, a major improvement over methods that had been available previously. The overlapping-pool approach offers a substantial further improvement. It does not require prior meiotic mapping or backcrossing but only the definition of complementation groups (which we facilitated by the use of introduced hygromycin-resistance and paromomycin-resistance markers to allow easy selection of diploids). It makes highly efficient (and thus economical) use of a small number of library preparations and Illumina lanes: with just five libraries, we unequivocally identified the causative mutations (against a background of thousands of irrelevant mutations) in 12 of the 14 mutants included in the original analysis. We have recently also been successful in extending the approach to a set of 32 mutants sequenced in six pools (F. Cross, K. Lieberman, and M. Breker, unpublished results)—an ∼5× reduction in labor, reagents, and sequencing costs—and even further extensions may be possible.
For success with the overlapping-pool approach, high-quality sequence information (including PCR-free library preparation) and careful bioinformatics filtering are required, because the approach relies critically on interpretation of alternative reads that may occur only once or a few times in a given library. Indeed, one causative mutation (lat3-1) was initially missed in this study, whereas a second (lat1-1) was identified only tentatively on the assumption that the relevant mutation had been missed in one of the three relevant pools. Given that Illumina sequence information is typically considered reliable only at ≥10× coverage, these failures presumably occurred because the 20–25× sequencing coverage used (only ∼3× coverage per mutant strain) led to occasional stochastic failures to detect the alternative read. Recurrent mutations (i.e., cases in which two independently isolated mutants contain identical mutations) can also interfere with the overlapping-pool approach. Although such cases will often be deconvoluted successfully, this is not guaranteed; fortunately, each mutant in the present study had a mutation that was molecularly distinct from those in others of the same complementation group. Finally, although in the present study, we made central use of the prior establishment of complementation groups, we are currently developing related approaches for which this will not be needed.
In the present study, as in many others, it was important both to confirm the initial identifications of causative mutations and to track those mutations in the segregants from genetic crosses. For both purposes, we found that an adaptation of the allele-specific PCR method (Gaudet et al. 2009) was rapid, cheap, and effective.
Resolution of the ida5/NAP1 mysteries and implications for actin structure and function
The normal proliferation of the ida5-1 null mutant (Kato-Minoura et al. 1997), upregulation of the highly divergent actin NAP1 in that mutant (Kato-Minoura et al. 1998), and uncertainty both about the presence of F-actin in vegetative cells (Harper et al. 1992; Kato-Minoura et al. 1998) and about the polymerization ability of NAP1 (Hirono et al. 2003; Kato-Minoura 2011) raised questions that the isolation of the nap1 and lat mutants have now allowed us to answer, revealing some important and rather surprising aspects of actin structure and function. First, the inviability of the nap1, lat1, lat2, and lat3 mutants either in combination with ida5-1 or upon treatment with LatB appears to demonstrate both that actin function is essential in vegetative Chlamydomonas cells and that either the conventional actin IDA5 or the highly divergent actin NAP1 can provide the essential function(s). Moreover, in wild-type cells (where NAP1 is not expressed), the F-actin-specific probe Lifeact detects structures that disappear rapidly upon LatB treatment, whereas in ida5-1 cells, or in wild-type cells treated for longer times with LatB, it detects similar structures that are highly resistant to LatB and are absent both in nap1 mutants and in the lat mutants (which cannot upregulate NAP1). Thus, it seems clear that both IDA5 and NAP1 form F-actin filaments and that F-IDA5 is sensitive to LatB, whereas F-NAP1 is resistant. It was not clear a priori that an actin as divergent as NAP1 could either form filaments or provide normal actin functions, and we are not aware of other cases in which an F-actin has been shown to be resistant even to high doses of latrunculin. In addition, the isolation of IDA5 mutations as suppressors of the LatB sensitivity of a nap1 mutant provides a formal demonstration that F-IDA5 is the principal or exclusive target of LatB in wild-type cells. An interesting question that remains unanswered is whether IDA5 and NAP1 could/would assemble into common filaments if they were both expressed in the same cells.
It is important to note that although NAP1 can provide the essential F-actin function(s) in the absence of IDA5, F-NAP1 does not appear to be functionally identical to F-IDA5. ida5-1 cells have several subtle but significant defects, such as altered inner-arm dynein structure in their flagella and abnormal and inefficient fertilization-tubule formation (Kato et al. 1993; Kato-Minoura et al. 1997). Moreover, although the type-VIII myosin MYO2 showed a LatB-sensitive localization like that of F-actin in wild-type cells (Avasthi et al. 2014), it did not show a detectable localization in ida5-1 cells (our unpublished data), suggesting that F-NAP1 cannot serve as an efficient track for this motor. Although NAP1-like proteins have been found only in the close relatives of Chlamydomonas within the green algae, further study of their molecular functions (what they can and cannot do) should provide broader insights into actin structure and function.
Importantly, as the nap1 and lat1–lat3 mutations greatly reduce or eliminate NAP1 function, and LatB ablates IDA5 function, we can now effectively explore the cellular roles of actin in Chlamydomonas. Our limited results to date (Figure 10) suggest that actin has roles both in cell growth and in cell division; future studies should be able to clarify these roles. Such clarification should have significance far beyond the understanding of Chlamydomonas itself, as this alga can now serve as an experimentally tractable representative not only of the plant superkingdom (where genetic studies of actin function are challenging because of the multitude of actin genes) but also of at least some fraction (Rogozin et al. 2009) of the vast majority of the eukaryotic world that lies outside of the opisthokonts (animals, fungi, and their close relatives), where most studies of actin function have been done. For example, is the role of actin in endocytosis, well documented in the opisthokonts (Goode et al. 2015), universal (and thus presumably ancestral) in eukaryotes more generally? Does the formation of cleavage furrows (which is nearly universal in eukaryotes other than higher plants) always involve actin (Balasubramanian et al. 2012), even though type-II myosin (a component of the opisthokont actomyosin ring) is not present in groups other than the opisthokonts and their “close” relatives the amoebozoa (Richards and Cavalier-Smith 2005)?
The F-actin surveillance pathway, the roles of LAT1–LAT3, and the evolution of NAP1
When F-actin is lost either by ida5 mutation or by treatment of wild-type cells with LatB, Chlamydomonas cells respond by massively upregulating NAP1 (and, to a lesser extent, IDA5). This surveillance pathway and response are adaptive, because the ability of NAP1 to provide essential F-actin functions and its resistance to depolymerization by LatB allow the cells to survive and continue proliferating in the face of what would otherwise be a lethal insult. Regulatory responses to actin depolymerization have been observed in other systems. In yeast, actin perturbations trigger transient G2 arrest (the “morphogenesis checkpoint”) via a pathway involving activation of the MAP kinase Mpk1 and impinging on the Cdc28 phosphatase Mih1 (Harrison et al. 2001). Although it seems likely that changes in gene expression also result from such Mpk1 activation, this possibility does not appear to have been examined. In mammalian cells, both actin and actin-binding proteins participate in transcriptional and post-transcriptional regulation of a variety of genes in response to shocks to the actin cytoskeleton (Miralles and Visa 2006; Olson and Nordheim 2010; Aragona et al. 2013; Mana-Capelli et al. 2014). In the best-studied case, perturbations in F-actin integrity are sensed by altered abundance of free MRTF (a G-actin-binding protein), which in turn regulates the transcription factor SRF (Olson and Nordheim 2010). As in Chlamydomonas, these pathways regulate the expression of actin isoforms; they also regulate the expression of genes for actin-binding proteins, among others. In plant cells, disturbance of F-actin under attack by pathogens has been reported to induce genes involved in the innate-immune response (Day et al. 2011; Matoušková et al. 2014).
LatB treatment could induce NAP1 if F-actin represses NAP1, if G-actin–LatB complexes induce NAP1, or if normal (LatB-free) G-actin represses NAP1; these possibilities are not mutually exclusive. ida5-1 mutants express NAP1 (although at a lower level than wild-type cells treated with LatB); thus, activation of NAP1 by G-actin–LatB complexes cannot be the exclusive mechanism. LAT1, LAT2, and LAT3 are required for NAP1 upregulation in response to LatB, and the synthetic lethality of lat1, lat2, and lat3 with ida5-1 suggests, but does not prove, that the LAT proteins are also required for NAP1 induction in the genetic absence of IDA5. A conditional IDA5 mutant will be required to test this idea.
The degree to which the pathways for response to actin perturbations in other organisms resemble the surveillance pathway we have found in Chlamydomonas is not clear. The three LAT proteins do not have obvious orthologs in organisms beyond the closely related green algae, and their sequences have provided no clues to their function beyond the presence of serine-threonine-kinase domains in LAT1 and LAT3. The rescue of the lat1–lat3 mutations by ectopic expression of NAP1 indicates that the regulation of this gene is the most critical role of the LAT proteins. However, the rescue was only partial (rescue on 5 µM LatB, but not on 10 µM LatB), which suggests that the LAT proteins may have one or more other roles (e.g., in the regulation of other genes) that are necessary for full resistance to the drug. However, we have not yet ruled out the possibility that the levels of ectopic expression were simply not sufficient for full LatB resistance. Although our screen for lat mutants appeared to be nearly saturated, it seems unlikely that LAT1, LAT2, and LAT3 constitute the entire F-actin surveillance pathway. In particular, at least one transcription factor would appear to be required, because none of the three LAT proteins contains a recognizable DNA-binding motif, along with at least one actin-binding protein (which might conceivably be one of the three LAT proteins). Such genes might have been missed in our screen because they are essential even in the absence of LatB or are functionally redundant.
How NAP1 achieves latrunculin resistance and how it evolved remain unclear. One possibility is that NAP1 may have a very low intrinsic depolymerization rate. As seen with the yeast act1-159 mutant (Belmont and Drubin 1998), reduced filament turnover can reduce the efficacy of G-actin-sequestering drugs like latrunculins. This idea is attractive because such an alternative actin would display resistance to multiple antiactin drugs and therefore seemingly provide the cells with a major evolutionary advantage. However, it would be surprising if a poorly depolymerizing actin could fulfill all essential actin functions, given how tightly filament turnover is regulated in other systems (Doherty and McMahon 2008). Careful biochemical experiments using purified NAP1 and in vivo experiments using pharmacological and genetic manipulation of actin dynamics should be able to illuminate these possibilities.
Another possibility is that NAP1 has a low affinity for the drug, as do the latrunculin-resistant mutant actins in other organisms (Ayscough and Drubin 1996; Fujita et al. 2003) and (probably) the mutant IDA5 proteins identified here as suppressors of nap1-2. Indeed, one of the five amino acid residues directly involved in hydrogen bonding between rabbit-muscle actin and LatA (Morton et al. 2000) is altered in NAP1 (His for Tyr at position 73: Figure 1B). This simple model would be consistent with the near-normal phenotype of the ida5-1 mutant but would seem to require exposure of the ancestors of modern Chlamydomonas (and related Volvocales algae, the only other organisms in which close homologs of NAP1 have been found: Kato-Minoura et al. 2014) to a latrunculin-like toxin during evolution. Although the latrunculins themselves are the products of marine sponges (Spector et al. 1983), and Chlamydomonas lives in fresh water, such an exposure is certainly possible. Moreover, because the LatA-binding site in actin is located close to the ATP/ADP-binding pocket (Morton et al. 2000), it is possible that other, currently unknown natural toxins target the same region as do the latrunculins, thus promoting the evolution of resistant proteins with alterations in that region.
It may also be relevant to the evolution of NAP1 that modern Chlamydomonas also shows upregulation of IDA5 in response to the loss of F-actin during LatB treatment; this upregulation may be a vestige of an early step in the evolution of a surveillance pathway and then a toxin-resistant paralog. Controlled expression of the paralog could prevent its product from interfering with the function of normal actin in the absence of stress.
Examples of highly divergent actins can also be found in other organisms. For example, of the 10 Arabidopsis actins, ACT5 and ACT9 share only ∼70% identity with the other, more conventional actins. Because of this sequence divergence and a lack of evidence for expression, ACT5 and ACT9 are currently considered to be pseudogenes (McDowell et al. 1996), but it may be that the conditions that provoke their expression, and under which their products become useful, have just not been discovered. Accelerated actin evolution under environmental pressure might be another example of the evolutionary arms races that have been found to promote very rapid evolution in proteins that otherwise would be expected to be highly conserved (Malik and Henikoff 2001).
We thank Luke Mackinder, Martin Jonikas, and Arthur Grossman for helpful discussions; Craig Atkins for help with strain construction; and Kresti Pecani for expert preparation of Illumina sequencing libraries. This work was supported by National Science Foundation EAGER award 1548533 (to J.R.P.), the Stanford University Department of Genetics, National Institutes of Health grant 5R01GM078153-07 (to F.R.C.), and The Rockefeller University.
Communicating editor: D. Lew
Supporting information is available online at www.genetics.org/lookup/suppl/doi:10.1534/genetics.115.184663/-/DC1.
- Received November 13, 2015.
- Accepted December 28, 2015.
- Copyright © 2016 by the Genetics Society of America