Clustering and gene ontology enrichment analysis of genes significantly changing during C. cinerea meiosis reveal transcriptional waves and distinct temporally regulated processes
Gene transcripts significantly changing during the C. cinerea
meiotic time course were classified by clustering. Genes were K-means clustered using the Pearson correlation, which groups genes according to similarity in their temporal expression patterns, using a successive bifurcation strategy that removed user-choice from the resultant number of clusters (Materials and Methods
, Figure S1
). This strategy produced nine distinct gene transcript profiles (), exhibiting successive “waves” of transcription, as reported for other organisms 
C. cinerea meiotic gene clusters.
Gene ontology (GO) enrichment analysis of the nine clusters was used to identify over-represented gene functions within each of the nine C. cinerea
clusters (Table S3
). The nine clusters exhibit a clear difference in transcript profile between clusters 1–5 and 6–9, as indicated by the initial bifurcation (). This is supported by the distinct classes of genes enriched in these clusters, which represents a broad switch from expression of genes required for the meiotic prophase I activities in early clusters, to expression of gill maturation and sporulation-related genes in later clusters (Table S3
). This corresponds with a similarly dramatic transcriptional switch from “early” to “middle” gene expression, as observed in S. cerevisiae 
Pre-meiotic DNA replication in C. cinerea
occurs just prior to nuclear fusion 
. Many aspects of DNA replication are well-conserved 
, such as the origin-recognition complex (ORC) and MCM2–7 complex, and these genes are expressed primarily in early clusters 1–3 (Table S1
). Clusters 1–3 are enriched in functional categories of genes involved in early meiotic processes (Table S3
); DNA replication is reflected in categories such as nucleic acid binding. Cluster 2, which exhibits a more prolonged transcript presence than other clusters, is enriched for regulation and organization of the cytoskeleton. This is likely to be important for karyogamy, organization of the meiotic spindle, and segregation of chromosomes; these processes span the entire time course, explaining the prolonged requirement of these transcripts. RNA splicing functions are also enriched in cluster 2, which is notable because control of splicing has been implicated in meiotic regulation 
. All the genes encoding components of the cohesin complex (scc3
and the meiosis-associated factor rec8
) are present in cluster 3. Cohesin holds sister chromatids together during meiosis, and primarily loads early, during replication. The gene encoding Spo11, which initiates recombination through its formation of double-strand breaks 
is also in cluster 3. This suggests that cluster 3 may be a source of promising candidates for early-acting meiotic genes.
We noted a massive enrichment of genes involved in ribosome production, translation, protein catabolism, and ribosomal RNA processing in cluster 4. In S. cerevisiae
, ribosomal protein genes are repressed on entering meiosis, with a subsequent increase in expression during sporulation, reflecting the starvation conditions required to induce meiosis in this organism 
. In C. cinerea
, transcript levels of ribosomal protein genes are relatively high until karyogamy, after which a gradual decline is observed, with no subsequent increase of transcription. Ribosomal degradation prior to meiosis and subsequent resynthesis during meiosis or sporulation have been previously noted in S. cerevisiae
and Chlamydomonas reinhardtii
(reviewed in 
), and ribosomal turnover is implicated in regulation of cell growth and proliferation in Xenopus laevis
and D. melanogaster
(reviewed in 
). Enhanced expression of ribosomal genes in C. cinerea
at K-3 and K may be in preparation for meiosis and for the massive, rapid cellular expansion in the gills and fruit body over the timescale examined in this time course.
Cluster 5 genes exhibit intermediate transcript levels prior to karyogamy, with increased levels of expression during nuclear fusion and leptotene/zygotene, after which transcripts decrease rapidly. Cluster 5 is highly enriched for genes known to be involved in meiotic processes such as damaged-DNA binding, mismatch repair, and DNA modification (Table S3
). Characterized genes in this cluster include those critical for key meiotic events such as strand exchange (dmc1
), axial element formation and synapsis (hop1
), and crossover formation (msh5
). Several of the genes expressed in cluster 5 play key roles in meiosis in other organisms (as summarized in 
), making this cluster a rich source for exploration of meiotic gene candidates.
Clusters 6–9 are enriched in genes required for spore formation. We observed a progressive shift from expression of biosynthetic genes, which may play a role in gill expansion due to carbohydrate acquisition and vacuolation (e.g. fatty acid synthesis in cluster 6, sugar and energy reserve synthesis in cluster 7) to those involved in formation of spore structure and spore packaging (e.g. cell wall biogenesis in cluster 8 and extracellular polysaccharide and carbohydrate transport in cluster 9) as well as preparation for spore germination (spore germination associated genes in cluster 8, and those involved in perception of external stimuli in cluster 9). A comparative analysis of spore formation, although potentially of great interest, is beyond the scope of this study.
Meiotic function genes are more conserved in their induction and expression patterns than genes not known to be meiotic
Previously, meiotic genes in S. cerevisiae
and S. pombe
were found to be more likely to be co-induced than a control set of genes with orthologs in S. pombe
that were induced in one S. cerevisiae
strain but not another 
. We wished to ask if meiotic genes are also more likely to be co-induced than non-meiotic genes in comparisons among the two yeasts and C. cinerea
To determine which genes are induced upon entry into meiosis in C. cinerea
, we compared gene expression during vegetative dikaryotic growth to expression in meiotic gill tissue at K-3. We observed 886 genes to be expressed only in gill tissue, with a further 3,621 genes expressed in vegetative tissue but significantly induced upon entry into meiosis. To ask whether genes meiotically induced in C. cinerea
are also induced upon meiotic entry in S. cerevisiae
and S. pombe
, we identified single copy, unambiguous, putative orthologs (henceforth referred to as “orthologs”; see Materials and Methods
) and compared their patterns of induction. Transcript level changes upon meiotic entry in S. cerevisiae
and S. pombe
were determined from previously published microarray data 
, and induction of the meiosis-associated gene spo11
was used as a control indicator of the transition between non-meiotic and meiotic cells. Transcript level changes in S. cerevisiae
and S. pombe
were compared with one another and with the significant changes between vegetative tissue and K-3 in C. cinerea
Orthologs and expression data were available in all three fungal species for 2,006 genes. Genes were assigned to “meiotic function” (MF) or “no known meiotic function” (NKMF) categories as defined by the Saccharomyces
Genome Database 
and the Gene Ontology 
). Of the 2,006 pertinent genes, significant induction on entry to meiosis was observed for 1,046 C. cinerea
genes. In the yeasts, 829 S. cerevisiae
genes and 869 S. pombe
genes were induced. We considered the 829 genes induced in S. cerevisiae
, as this is the maximum number of genes with the potential to be induced in all three species, and asked how many were induced upon entry into meiosis in all three species for the MF and NKMF classes. We observed that 50 of the 119 MF genes were induced in all three fungal species, while only 169 of the 710 NKMF genes were co-induced (p<1×10−
4, Fisher's Exact test). The names and putative functions of genes in the MF_co-induced, MF_not_co-induced, and NKMF_co-induced categories are listed in separate tabs in Table S1
The 50 commonly induced MF gene set contains a number of genes known to be crucial for meiosis, such as all three genes of the Mre11 complex (mre11
), genes encoding strand invasion proteins (dmc1
), and genes encoding meiosis-associated proteins (spo11
, hop1 and dmc1
). This suggests that coordinate induction of genes across multiple species may prove to be an indicator of meiotic function; the inclusion of C. cinerea
as a comparator clarifies those genes that are likely evolutionarily conserved in their meiotic behaviour. Several of the genes that are coordinately induced in all three species but currently have no known meiotic function are involved in spindle formation, chromosome segregation or DNA-metabolic processes, and may yet prove to be important in meiosis (Table S1
Comparison of genes coordinately induced on entry into meiosis is necessarily a binary approach, asking “on/off” questions that do not query the changes in transcript level through a time course. A complementary approach is to compare the temporal transcript profiles of genes during meiosis. Comparative studies of mammalian gametogenesis asked whether genes were conserved in their relative expression patterns in distinct pre-meiotic, meiotic, and post-meiotic tissues in rat and mouse (~23 million years divergent 
), and found that correlated genes were enriched for reproductive function 
. The availability of time course data describing meiosis in three different fungal species affords us the opportunity to ask if conservation of transcript profile can also be observed within meiotic cells in these more diverged organisms. This may also highlight similarities and differences in meiotic process not observed by examining coordinate induction.
We examined the 2,721 genes with significantly changing transcript levels in C. cinerea
, and found S. cerevisiae
and S. pombe
orthologs and corresponding expression data for 743 genes 
. Meiotic progression in the three fungi differs with respect to the overall time required for completion of meiosis, and the duration of certain stages within the meiotic program. Thus, in order to compare meiotic transcript profiles in the three species, we aligned expression data according to previously described meiotic landmarks and defined an eight-point time course (). Data were unavailable for all three species at every stage defined; in these cases, expression data were interpolated by averaging the expression from flanking time points.
Timing of meiotic events in C. cinerea, S. pombe, and S. cerevisiae.
The 743 orthologs were again divided into MF genes (81) and NKMF genes (662). For each orthologous gene, the transcript profiles were compared for each of the three possible interspecies pair-wise combinations (i.e., Ccin vs. Scer
, Ccin vs. Spom
, Scer vs. Spom
), and correlation coefficients (r
) were generated. In all comparisons, more transcript profiles are well-correlated (r
>0.5) for MF genes than NKMF genes (Ccin/Spom
, 44% vs. 28%; Ccin
, 54% vs. 30%; Spom
, 32% vs. 29%), and the correlation value distributions () were significantly different when MF genes were compared with NKMF genes (Mann-Whitney-Wilcoxon test: Ccin
<0.0112). Thus, the transcript profiles of MF genes are more highly conserved than those of NKMF genes. Fifty-two genes of the NKMF class are well-correlated in all three pair-wise comparisons (and six of these genes are also coinduced in all three species); these subsets (Table S1
) provide an interesting pool of candidates that may have additional, as yet uncharacterized, meiotic functions.
Distribution of gene expression profile correlation coefficients.
Given that MF genes are enriched both for coordinate induction on entry to meiosis and transcript profile correlation though meiosis, we noted some surprising differences in induction and correlation in some meiotic genes, highlighting the complementary value of both these types of analysis. Of genes coordinately induced in the three species, several do not exhibit well-correlated transcript profiles. Of the 50 coordinately induced MF genes, only 41 are well correlated between at least 2 of the three species, with only a single gene (rad50) coordinately induced and well-correlated in all three species (), indicating that transcript profile conservation reveals additional information about meiotic regulation; coordinate induction does not predict transcript profile correlation or vice versa. Meiotic genes may be expected to be all induced upon meiotic entry, but their subsequent expression behavior may be able to inform us about the different ways meiosis is achieved in different organisms.
rad50 expression is well-correlated in C. cinerea, S. pombe, and S. cerevisiae.
Interestingly, despite their well-characterized meiotic roles, and although they are induced upon entry into meiosis in all three fungal species, the transcript profiles of spo11
, which encodes a meiosis-associated cohesin subunit 
, are well-correlated only between C. cinerea
and S. pombe
, with S. cerevisiae
expression peaking late in meiosis, just before the first meiotic division, later than the timing of the corresponding protein activity (). Other genes essential for meiosis, such as hop1
, also exhibit a similar late transcript peak in S. cerevisiae
. This unexpected lack of correlation may indicate additional or differing functions for some meiotic proteins; for example, Spo11 forms meiotic double-strand breaks independently of the Mre11 complex in C. cinerea
and S. pombe
, but in an Mre11-dependent manner in S. cerevisiae 
. Alternatively or additionally, post-transcriptional regulation might be more prevalent in core meiotic genes in S. cerevisiae
. For example, alternative splicing of introns is disproportionately involved in meiotic gene regulation when compared to other biological processes in S. cerevisiae 
Gene expression in spo11 and rec8 are well-correlated only between C. cinerea and S. pombe.
Other genes exhibit the correlated transcript profiles expected given the roles of their proteins. For example, the genes encoding three members of the cohesin complex, Smc1, Smc3, and Scc3, are all well-correlated between C. cinerea and S. cerevisiae, and somewhat correlated between S. cerevisiae and S. pombe (). Single genes encode the cohesin Scc1 (S. cerevisiae)/Rad21 (S. pombe) in the yeasts, whereas C. cinerea has two homologous genes encoding this protein. For one of these, rad21.2, the transcript profile is well-correlated with those of its S. cerevisiae and S. pombe homologs. The Rad21.2 protein is found exclusively in meiotic tissue (Palmerini et al., in preparation). In contrast, rad21.1 displays a very different transcript profile (). The Rad21.1 protein is the only mitotic kleisin in C. cinerea (Palmerini et al., in preparation) and its RNA abundance spike late in meiosis may reflect transcription in preparation for the first post-meiotic mitosis.
Gene expression profiles of cohesin subunits are well correlated between C. cinerea and S. cerevisiae, and between C. cinerea and S. pombe.
We also noted an unusual expression pattern of some MCM complex genes. This complex, composed of MCM2–7, is involved in replication, and thus one would expect these genes to be expressed early and coordinately. Most of the genes are indeed expressed in such a manner, with the exception of S. pombe mcm6
and S. cerevisiae mcm5
, which have very similar, late-peaking, expression profiles (). C. cinerea mcm2
also exhibited this late expression profile, but the change through meiosis was not statistically significant. In Drosophila melanogaster
mutants are defective in the resolution of meiotic double-strand breaks to crossovers 
raising the possibility of a similar meiotic role for mcm5
in S. cerevisiae
. The similar expression patterns of S. pombe mcm6
and C. cinerea mcm2
suggest this additional role may not be confined to a specific MCM subunit.
MCM complex gene expression declines through meiosis, with key exceptions.
Previous studies have shown both conservation and apparent divergence in proteins required for meiosis 
. Our work shows that, for proteins whose primary sequence is conserved enough for homology to be recognized, gene expression profiles throughout meiosis are significantly conserved. That this conservation occurs across >500 million years of evolution suggests that meiosis is more conserved than hitherto recognized. Based on our data and existing criteria, we propose an expanded inventory of genes involved in meiosis (Table S4
We also predict that additional conservation of meiotic genes will be found as the algorithms for detecting homology become more sophisticated. This has practical implications, in that previously uncharacterized genes with meiotic roles could be identified both by similarity of expression profile to known meiotic genes within an organism (i.e., ) and by conservation of expression profile across organisms. For example, ubc9
, which is induced upon entry into meiosis in the three fungi, has well correlated expression profiles in a comparison between C. cinerea
and S. pombe
, and a late peaking expression profile similar to that of spo11
in S. cerevisiae
. Ubc9 has no known meiotic role in the yeasts but is involved in sumoylation during meiosis in C. cinerea 
. Genes with orthologs in the C. cinerea
meiotic cluster 5 with no currently identified meiotic function (53) are also compelling candidates for study. Of these genes, four, including those encoding two transcription factors (Hir1 and Tfb2) are coordinately induced in all three fungal species. An additional six genes are well-correlated in pair-wise comparisons in all three fungi. In addition, given the sequence divergence of many meiotic genes, such as those encoding synaptonemal complex components, a proportion of the genes in cluster 5 with no currently apparent orthologs are likely to have meiotic roles. This is illustrated by bad42
, which has a critical meiotic role in C. cinerea
meiosis but has no known orthologs 
It is logical that meiotic processes must be tightly controlled to avoid deleterious effects of renegade gene expression (e.g. 
). The broad conservation of meiosis opens up interesting possibilities for the study of this process in different organisms. Thus, protein function can be inferred from studies in different species and the exploitation of the benefits of various study organisms, such as the elegant cytology and uncoupled recombination and SC formation in C. elegans
and D. melanogaster
; the tractability of S. cerevisiae
and S. pombe
; and the synchronous meiosis, facile screening, and prolonged prophase in C. cinerea
. In addition, the striking differences in expression pattern between species for necessarily tightly controlled genes such as spo11
indicate differences in meiotic gene regulation, highlighting the value of different types of analysis.