Analysis of the Drosophila melanogaster Testes Transcriptome Reveals Coordinate Regulation of Paralogous Genes

Abstract

Gene duplications have been broadly implicated in the generation of testis-specific genes. To perform a comprehensive analysis of paralogous testis-biased genes, we characterized the testes transcriptome of Drosophila melanogaster by comparing gene expression in testes vs. ovaries, heads, and gonadectomized males. A number of the identified 399 testis-biased genes code for the known components of mature sperm. Among the detected 69 genes downregulated in testes, a large fraction is required for viability. By analyzing paralogs of testis-biased genes, we identified “co-regulated” paralogous pairs in which both genes are testis biased, “anti-regulated” pairs in which one paralog is testis biased and the other downregulated in testes, and “neutral” pairs in which one paralog is testis biased and the other constitutively expressed. The numbers of identified co-regulated and anti-regulated pairs were higher than expected by chance. Testis-biased genes included in these pairs show decreased frequency of lethal mutations, suggesting their specific role in male reproduction. These genes also show exceptionally high interspecific variability of expression in comparison between D. melanogaster and the closely related D. simulans. Further, interspecific changes in testis bias of expression are generally correlated within the co-regulated pairs and are anti-correlated within the anti-regulated pairs, suggesting coordinated regulation within both types of paralogous gene pairs.

RECENT analyses of sequenced eukaryotic genomes suggest that they are composed of two sets of genes, including a highly conserved set of “housekeeping” genes (Koonin et al. 2004) and a set of lineage-specific genes. The more sophisticated tissue organization correlates with the increased genome complexity and therefore with the increase in the number of lineage-specific genes, as exemplified by comparison of mammals vs. invertebrates. Gene duplications that result in expansion of gene families have been considered the major mechanism for generation of lineage-specific genes (Lespinet et al. 2002; Copley et al. 2003). Duplicated genes may acquire new expression patterns and functions (Ohno et al. 1968) and thus contribute to diversification of tissues during development.

In particular, in Drosophila a number of testis-specific genes have been generated by gene duplications. For example, the testis-specific gene Sdic evolved from the broadly expressed gene Cdic (Nurminsky et al. 1998). The gene Dntf-2r is a retroposed copy of Dntf-2, which is specifically expressed in the male reproductive system while the parental gene is ubiquitous (Betran and Long 2003). A number of the TFIID subunits have testis-specific paralogs (Hiller et al. 2004). Similarly, specific subunits of the proteasome complex are expressed in testis and are required for male fertility (Yuan et al. 1996; Zhong and Belote 2007). The genes e(y)2 and e(y)2b provide a contrary case in which the parental gene becomes testis specific while the duplicated copy maintains ubiquitous expression (Krasnov et al. 2005). The examples of testis specificity acquired by a duplicated gene have also been described in mammals (McCarrey and Thomas 1987; Ivanov et al. 2000).

Creation of a new testis-specific gene in evolution suggests acquisition of a novel function by this gene. However, the above examples illustrate that such events occur in the presence of a broadly expressed paralog that is potentially capable of performing a similar role. Studies of several paralogous pairs have suggested that this apparent redundancy may be resolved in two distinct ways:

1. The broadly expressed paralog may be expressed at the same level (or even may be upregulated) in testes along with the testis-specific paralog; this pattern implies that the testis-specific function supplements the ubiquitous function. For example, the ubiquitously expressed TFIID subunits (TAFs) are present in testes and are sufficient for expression of the broadly expressed genes, whereas the testis-specific TAFs are required for expression of the testis-specific genes (Hiller et al. 2004).

2. Alternatively, the broadly expressed paralog may be downregulated in testes, suggesting that the testis-specific function replaces the ubiquitous function. For instance, studies of the proteasome subunit genes indicate that the ubiquitous proteasome machinery is replaced by its testes-specific counterpart during spermatogenesis (Yuan et al. 1996; Ma et al. 2002; Zhong and Belote 2007). Similarly, the ubiquitous translation factor eIF4G is downregulated in testes and is replaced by the paralogous factor ofs that is essential for germline development (Baker and Fuller 2007; Franklin-Dumont et al. 2007).

Thus, expression patterns of the paralogous genes in testes provide valuable indications for the functional relationships between these genes. In addition, as we show below, analysis of these patterns in related Drosophila species provides insights into regulation of paralogous genes. Here we report a comprehensive microarray-based study of the Drosophila testes transcriptome, focused on the analysis of the paralogous gene pairs.

MATERIALS AND METHODS

RNA preparation, amplification, and sample labeling:

Total RNA was isolated from adult Drosophila (age 1–5 days) tissues with Trizol reagent (Invitrogen, San Diego) and purified with RNeasy kit (QIAGEN, Valencia, CA). Poly(A)⁺ RNA was selectively amplified from 0.5 to 1 μg of total RNA samples with the BD SMART cDNA synthesis kit and the BD Advantage 2 PCR enzyme system (CLONTECH), according to the manufacturer's protocol. PCR products were cleaned with the Wizard SV Gel and PCR Clean-Up system (Promega, Madison, WI). cDNA (3.5–4.5 μg) was labeled with the ULYSIS Alexa Fluor 546 or Alexa Fluor 647 (Molecular Probes, Eugene, OR) dyes according to the manufacturer's protocol. Pairs of samples labeled with different dyes were mixed together and used for competitive hybridization with microarrays.

Hybridization of samples to microarrays:

Drosophila oligonucleotide microarray set (Qiagen-Operon) was printed on the aminosilane-coated slides at Tufts-New England Medical Center Expression Array Core facility. The set contains 14,593 70-mer oligonucleotides representing 13,664 genes, which cover most of genes in the release 3.2 of the Drosophila genome. Additional information on the Drosophila oligonucleotide set can be found at the manufacturer's website (http://www.operon.com). Labeled pairs of samples were hybridized in 1× hybridization buffer (Amersham, Piscataway, NJ), 20% formamide, and 0.025% each of Ficoll, polyvinylpyrrolidone, Na pyrophosphate, and heparin. Samples were hybridized with microarrays for 44 hr at 37°. After hybridization slides were washed, dried by centrifugation, and scanned on the ScanArray 4000 scanner (Perkin-Elmer, Norwalk, CT).

Data acquisition and analysis:

Fluorescence intensities of individual spots were acquired from the array images with ImaGene software (BioDiscovery). Subsequent normalization and statistical analysis were performed with package limma (Smyth and Speed 2003; Smyth 2004), part of the BioConductor project (Gentleman et al. 2004). Briefly, the individual background subtracted arrays were normalized using the print-tip loess method and then scale normalized between arrays, followed by linear model fitting with testes samples used as a common reference and the Bayesian method for statistical analysis. Selection of differentially expressed genes was treated as a classification problem across three contrasts of interest, with P < 0.01 for the classification statistic F used as a threshold. q-Values for individual contrasts represent P-values adjusted for multiple testing according to Benjamini and Hochberg (1995).

Analysis of gene ontology terms associated with groups of differentially expressed genes was performed with the web-based tool GeneMerge (http://genemerge.bioteam.net; Castillo-Davis and Hartl 2003).

BLAST and selection of paralogs:

Paralogs for different gene groups (described in the results) were selected with a BLAST tool as implemented in EBI/ENSEMBL Blastview (http://www.ensembl.org/Multi/blastview). Only the genes with at least 50% amino acid sequence identity to the target and homology regions of at least 50 amino acids were selected as paralogs.

The expression patterns of identified paralogs were assessed from microarray data. The count data were generated using original groups of genes for which the paralogs were found. A relevant gene was scored once for each category regardless of how many other paralogs fell into this category. The statistics for the count data were acquired with the Fisher's exact test for count data, as implemented in R (R Development Core Team 2006).

Real-time RT–PCR:

Total RNA was isolated from dissected adult testes and from gonadectomized males (age 1–5 days) with the Trizol reagent (Invitrogen). Reverse transcription reactions were performed with 1-μg samples of total RNA using the Superscript II enzyme (Invitrogen) and oligo(T) adaptor. A total of 0.5% of the reverse transcription reactions were used as templates in 20-μl real-time PCR reactions. Reactions were run in triplicates in the ABI 5700 Sequence Detector, using SYBR Green chemistry (Applied Biosystems, Foster City, CA). After the real-time PCR run, sizes of the amplified fragments were verified by the agarose gel electrophoresis.

RESULTS AND DISCUSSION

Characterization of the Drosophila testes transcriptome reveals upregulation of structural constituents of the sperm and of the genes involved in basic cell biological processes:

The gene expression pattern in Drosophila tissues was studied using the competitive hybridizations of whole-genome long-oligonucleotide-based gene microarrays with testes-derived cDNA vs. cDNAs obtained from heads, ovaries, and whole gonadectomized males. Hybridization data are publicly available under ArrayExpress accession no. E-TABM-273 (http://www.ebi.ac.uk/microarray-as/aer/entry). Analysis of these data identified a relatively large set of genes (399) upregulated in testes in all three comparisons (vs. ovaries, heads, and gonadless males) (P < 0.01) (supplemental Table 1). The obtained data, to our knowledge, represent the first microarray-based comprehensive analysis of testis transcriptome in Drosophila performed by comparisons of gene expression in testes vs. diverse tissue samples.

To evaluate our results, we compared our data to previously published data sets obtained in the relevant studies by Andrews et al. (2000) (GEO accession nos. GPL5 and GSM3–GSM10), and Parisi et al. (2003) (GEO accession nos. GPL20 and GSM2456–GSM2469). Owing to different approaches to data analysis, direct comparisons of the processed data are not feasible. Instead, we used a twofold change threshold to select differentially expressed genes from the published data. Analysis of a limited number of genes (1681 ESTs) by Andrews et al. (2000) identified 153 genes that show at least twofold overexpression in testes as compared to both ovaries and gonadectomized males. Among these, 111 genes are present in our data set, and 41 (36%) show upregulation in testes in the relevant comparisons with q < 0.05. The genomewide study by Parisi et al. (2003) identified 378 genes that show at least twofold overexpression in testes as compared to the ovaries. Among these genes, 337 are present in our data set, and 108 (32%) are upregulated in testes vs. ovaries with q < 0.05. We also used an additional indirect approach to evaluate our set of testis-biased genes by comparing this set to the list of testis-specific genes generated by the analysis of the EST databases (Boutanaev et al. 2002). Among the 399 genes in our microarray-based studies that showed upregulation in testes vs. heads, ovaries, and gonadectomized males, 48% have also been identified as testis specific by the EST data mining. Taking into account the differences among the objectives of the studies, experimental design, data analysis, and diverse microarray platforms, we consider the correlation between our results and the published data quite satisfactory.

Biological roles of the identified testis-biased genes were inferred using the Drosophila genome annotations in FlyBase (Crosby et al. 2007), BLAST searches, and surveys of the published research in Medline. The results of this analysis, presented in supplemental Table 1, are summarized in the Table 1. Among the 399 testis-biased genes, about half (218) were not annotated in FlyBase and their functions could not be inferred from the BLAST homology searches or from the literature. We suggest that a number of these “unknown” genes code for the highly specialized sperm proteins. Indeed, 13% of the proteins encoded by these genes have been identified as the components of mature sperm (Dorus et al. 2006). Further identification of the biological roles of such proteins, in particular in male reproduction, is expected to provide novel insights into molecular mechanisms of spermatogenesis.

TABLE 1.

Inferred biological functions, presence of paralogs, and mutant phenotypes for genes up- or downregulated in testes

	Total		Sperm proteome		Constitutively expressed paralogs		Testis- biased paralogs		Downregulated paralogs		Male sterile		Lethal
Gene category	U	D	U	D	U	D	U	D	U	D	U	D	U	D
Microtubule and associated proteins	11		2		2		2				1
Outer dense fiber of sperm tail	5		2				1
Mitochondrion components	28	17	14	6	5		3		7			1		6
Structural nuclear components	5		1				4
Nuclear transport	3										2
Chromosome maintenance	3	2												2
Transcriptional regulators
Transcription factors	11	5			1		2		1				3	2
Other DNA-binding proteins	4
DNA methylation	1
RNA processing/metabolism	5						2		1		1		1
Other RNA-binding factors	3
Translational regulators
Ribosome proteins	0	12								1				5
Other translation factors	6	1			2								2
Signaling
Receptors and associated proteins	11	2			1		2				1		1	1
Intracellular signaling	4	4	1			1		1		1			1	1
Protein folding	8	1			4		2		2		1
Protein modifications
Phosphorylation	7		2		1						1
Other modifications	1												1
Protein degradation
Ubiquitin dependent	9				2		1						1
Other proteases	6	2	4		2		4
Actin and associated proteins	6	2	1			1		1		1	1		1	1
Transmembrane transport	7	2							2
Peroxisome	4		1						1
Vesicular trafficing and exocytosis	7										1		2
Extracellular matrix	7	3						1	1	1			3
Sugar metabolism
Glycolysis	7	1	5	1	1		2	1	2	1
Other	4	2	3
Amino acid metabolism	1
Nucleotide metabolism	1	2
Lipid metabolism	3	3							1
Detoxification	3	3	1		1			2	1	1		1		1
Genes with “unknown” function	218	5	29		4		7		1
Total	399	69	66	7	26	2	32	6	20	6	9	2	16	19
% of up- or downregulated genesa			17	10	7	3	8	9	5	9	5	3	9	30
Difference between up- and downregulated genes; P-value			0.20		0.42		0.97		0.29		0.83		<0.01

Genes upregulated (U) or downregulated (D) in testes as compared to ovaries, heads, and whole gonadectomized males.

^aProportion of annotated genes (“unknown” genes excluded).

However, biological roles could be inferred for the rest (181) of the testis-biased genes. About 20% of the proteins encoded by these genes have been identified as the sperm components by mass spectrometry of the total sperm proteins (Dorus et al. 2006). We observed that certain categories of testis-biased genes are specifically enriched with the sperm protein-coding genes. These groups include (i) a large group of 44 known or putative components of sperm tail structures such as the microtubules, outer dense fiber, and mitochondria; (ii) the group of 11 genes involved in sugar metabolism; and (iii) the group of six peptidases. In aggregate, 49% of the proteins encoded by these three groups of genes have been found by Dorus et al. (2006) in mature sperm.

Intriguingly, about half of the proteins included in the above three categories of testis-biased genes were not identified as constituents of mature sperm in the broad studies of the sperm proteome (Dorus et al. 2006), although a number of these genes code for the putative sperm components (e.g., the outer dense fiber of sperm tail proteins). It is possible that such proteins will be identified as the sperm components by future targeted assays or, alternatively, that some testis-biased genes may code for proteins that contribute to spermatogenesis without serving as the sperm components. These include the groups of testis-biased genes that code for the proteins generally not present in sperm such as transcriptional and translational regulators, proteins involved in signaling, transmembrane transport, assembly and regulation of actin cytoskeleton, modifications and targeted degradation of proteins, protein folding, vesicular trafficking and exocytosis, extracellular proteins, peroxisome components, and proteins involved in lipid metabolism and detoxification. Upregulation of these diverse genes in testes suggests that multiple and very basic cell biological processes are modified during spermatogenesis.

To obtain further insight into the roles of identified testis-biased genes in male fertility as well as in general viability, we analyzed information available on the mutant phenotypes. Intriguingly, in the three groups of testis-biased genes that prominently contribute to sperm proteome, only 1 gene (of 61) shows male sterility of the mutants. However, among the groups of testis-biased genes that generally do not code for the sperm components, disruption of 8 genes (of 120) causes male sterility, and mutations in 16 genes are associated with lethality (Table 1, supplemental Table 1). Surprisingly, these observations revealed a marked paucity of male-sterile mutant phenotypes associated with the genes coding for the sperm proteins. Taking into consideration that the mutant data are provided by the studies of individual genes rather than by a systematic screening, it is quite possible that this paucity merely results from a relatively low number of studies aimed at analysis of the sperm protein-coding genes. For the group of genes that generally do not code for the sperm proteins but rather contribute to the diverse basic cell biological processes, the mutant data are more comprehensive. They show that while a number of these genes have essential functions beyond spermatogenesis, some are specifically required for male fertility, for example, the regulators of nuclear import tim (Beaver et al. 2002) and Pen (Mason et al. 2003), steroid binding protein CG15623 and calmodulin-like CG15179 (Benson et al. 2006), protein-folding factor CG7235 (Sarkar and Lakhotia 2005), cdk inhibitor rux (Avedisov et al. 2000), and vesicular regulator of protein-trafficking fan (Renner et al. 2003).

Downregulation of essential genes in testes:

In addition to testis-biased genes, we also identified a set of 69 genes that are downregulated in testes (P < 0.01) as compared to ovaries, heads, and gonadectomized males. Biological roles of these genes were inferred using the database searches as described above for the testis-biased genes. The results of this analysis are presented in supplemental Table 2 and are included in Table 1. Downregulation affects genes coding for the proteins involved in diverse aspects of cellular biology. Specifically, we observed two major groups of downregulated genes, including a group of 17 genes that code for known or putative mitochondrion components, and a group of 12 ribosomal protein genes. Analysis of the mutant phenotype data shows that many (30%) downregulated genes are essential for viability, despite the fact that spermatogenesis is associated with active biosynthesis, cell proliferation, and differentiation. One possible explanation is that transcriptional downregulation of genes does not necessarily lead to ablation of the proteins encoded by these genes, if these proteins are stable enough. This is apparently true for the 7 downregulated genes coding for the proteins present in mature sperm (Table 1) and may also explain why the broad downregulation of ribosomal protein genes in testes does not disrupt translation. Alternatively, downregulation of the essential genes in testes may be compensated by upregulation of the genes with similar function, in particular the paralogous genes, as has been strongly suggested for the proteasome components (Zhong and Belote 2007).

Testis-biased genes frequently have co-regulated and anti-regulated paralogs:

To better understand regulation of genes that have similar functions, we searched for the paralogs of the testis-biased genes and of the genes downregulated in testes using BLASTp (as described in materials and methods). Expression patterns of the identified paralogs in testes vs. other analyzed tissues were determined from the microarray data. This analysis (Table 1, supplemental Tables 1 and 2) showed that both testis-biased and downregulated genes have paralogs with the matching expression pattern (co-regulated paralogs) as frequently as the paralogs with the opposite expression pattern (anti-regulated paralogs). For example, 8% of the testis-biased genes have co-regulated paralogs that also show upregulation in testes, and, similarly, 5% of the testis-biased genes have anti-regulated paralogs that show downregulation in testes. The downregulated genes show exactly the same frequency (9%) of co- and anti-regulated paralogs. In addition, we found that similar numbers of testis-biased and downregulated genes have paralogs that show no differential expression in testes as compared to other tissues (i.e., were constitutively expressed). These results were not expected because paralogs are often created by gene duplications that may include regulatory sequences. Hence, the simple model would predict predominant co-regulation of paralogs.

To increase the comprehensiveness of the analysis of paralog expression, we relaxed the stringency of selection of the testis-biased and downregulated genes from the microarray data. We focused on the differences in expression between testes and somatic tissues, using the reference sample obtained from the gonadectomized males, which includes all adult tissues (except the gonads) and thus reflects the averaged expression levels in soma. The results of such comparison between testes and gonadectomized males yielded 814 genes that are testis biased and 834 genes that are downregulated in testes, at q < 0.05. Validation of these microarray-based predictions by real-time RT–PCR confirmed the expression pattern for 86% of genes that show a >1.3-fold expression difference in microarrays (n = 29) and for 72% of genes that show a 1.2- to 1.3-fold difference (n = 38) (data not shown). Next, paralogs of the testis-biased genes were determined using BLASTp. We found that 175 of the testis-biased genes have one or more paralogs. Of these, 45 testis-biased genes have testis-biased paralogs (which are therefore co-regulated), 27 testis-biased genes have downregulated paralogs (which are anti-regulated), and 93 testis-biased genes have constitutively expressed paralogs. In this analysis, the number of co-regulated paralog pairs was significantly (P = 0.042) higher than the number of anti-regulated paralog pairs.

To determine whether the numbers of the testis-biased genes that have co- or anti-regulated paralogs differ from the numbers expected by chance, we performed a random simulation. For this, we created two randomized nonoverlapping sets of genes. The sizes of these sets (870 and 824 genes) were comparable to the sizes of the testis-biased (814) and downregulated (834) gene sets. Paralogs for the genes in the first randomized set were identified as described above. In this set, the number of genes that have paralogs (184) was similar to the number observed for the set of testis-biased genes (175). Further, we simulated expression status of the identified paralog pairs. In this analysis, the presence of the paralog in the same (first) randomized set corresponds to co-regulation, and the presence of the paralog in the other randomized set corresponds to anti-regulation. We found that the numbers of such simulated pairs of both mock co-regulated (21 pair) and mock anti-regulated (15 pairs) paralogs were significantly smaller than the numbers of co-regulated or anti-regulated paralog pairs observed for the testis-biased genes (Figure 1). This analysis showed that paralogs of the testis-biased genes show a significantly elevated tendency to be differentially expressed between testes and somatic tissues. As mentioned above, elevated frequency of co-regulated paralogous pairs may result from the duplication of regulatory sequences along with the coding sequences. In contrast, overrepresentation of the anti-regulated paralog pairs, in which one paralog is upregulated in testes but the other is downregulated, does not have a trivial mechanistic explanation. Instead, this observation suggests that the testis-biased genes frequently substitute for the similar broadly expressed genes, probably leading to the adjustment of the cellular mechanisms to the peculiar requirements of male gametogenesis.

Figure 1.—

Increased frequency of the co-regulated or anti-regulated paralogs found for testis-biased genes. Proportions of testis-biased genes that have testis-biased (co-regulated) paralogs, or paralogs that are downregulated (anti-regulated) in testes, are shown (solid bars). They are significantly (*P < 0.05) or highly significantly (**P < 0.01) different from the results of random simulation (shaded bars).

Testis-biased genes that have differentially expressed paralogs show decreased frequency of known lethal mutations:

To characterize the sets of co-regulated and anti-regulated paralogs, as well as the paralog pairs in which one member is testis biased and the other is constitutively expressed in testes and in somatic tissues, we inferred the biological roles of identified genes using the FlyBase annotations (Crosby et al. 2007), BLAST homology searches, and analysis of published reports. The results are presented in supplemental Table 3, a–c, and summarized in Table 2 and Figure 2. To increase the sample sizes for this analysis, we expanded the sets of paralogs by lowering the stringency of assignment of expression status by increasing the q-value threshold for paralogs to 0.2. This analysis revealed that the co-regulated paralogs, and the paralog pairs in which one member is testis biased and the other is constitutively expressed, similarly contribute to the diverse categories of genes classified by their biological function (correlation coefficient r = 0.77). Specifically, the co-regulated paralogs, and/or paralog pairs in which one member is constitutively expressed, predominantly contribute to the categories of genes coding for the microtubule components and associated proteins, outer dense fiber of the sperm tail, structural components of the nucleus (e.g., protamines), peroxisome proteins, transcriptional regulators, and extracellular proteins. About 15% of these genes code for the components of mature sperm (Dorus et al. 2006) (Table 2). Some of the constitutively expressed paralogs (∼7%) also code for the constituents of mature sperm, consistent with the analysis by Dorus et al. (2006) indicating that a significant proportion of sperm proteome is encoded by the genes that do not show testis-biased expression.

TABLE 2.

Inferred biological functions of testis-biased genes and their paralogs

Gene category	Testis-biased genes that have testis- biased paralogs	Testis-biased genes that have downregulated paralogs	Downregulated paralogs	Testis-biased genes that have constitutively expressed paralogs	Constitutively expressed paralogs
Microtubule and associated proteins	6			8 (2)	10 (2)
Outer dense fiber of sperm tail	2
Mitochondrion components	8 (2)	13 (4)	12 (3)	7 (2)	7 (2)
Structural nuclear components	4 (1)
Nuclear transport
Chromosome maintenance				1	1
Transcriptional regulators	16	2	2	15	20
Translational regulators		2	2	3	3
Signaling	4	5	8	7	8
Protein folding	8	5	2	13	16
Protein modifications	2	1	1	6	10 (1)
Protein degradation	10 (5)	2 (1)	6	13 (3)	17 (3)
Actin and associated proteins	4 (1)	2 (1)	6	1	1
Transmembrane transport		3	4	2	3
Peroxisome	2 (1)
Vesicular trafficking and exocytosis		1	1
Extracellular matrix	2			2	12
Sugar metabolism	2 (1)	3 (3)	4 (1)	2 (2)	2
Amino acid metabolism				1 (1)	1
Nucleotide metabolism
Lipid metabolism		2	4	1	1
Detoxification		4 (1)	3 (1)	2	4
Genes with “unknown” function	15 (3)	3 (2)	3	6 (2)	7 (1)
Total	85	48	53	90	123
Mutants lethal	8	3	12	17	21
% of essential genes a	11	7	24	20	18
Mutant males sterile				5	8

No. of identified sperm proteins (according to Dorus et al. 2006) are in parentheses.

^aProportion of annotated genes (“unknown” genes excluded).

Figure 2.—

The spectrum of genes encoded by the anti-regulated paralogous pairs (open bars) is different from the spectra of genes encoded by the co-regulated, testis-biased paralogous pairs (solid bars) or by the paralogous pairs in which one gene is testis biased and the other constitutively expressed (shaded bars). Standard errors are indicated. Gene categories include genes with an inferred role: MT, microtubule components and associated proteins; OF, outer dense fiber; MC, mitochondria; NC, structural components of the nucleus; NT, nuclear transport; CM, chromosome maintenance; TR, transcriptional regulation; TL, translational regulation; SG, signaling; PF, protein folding; PM, protein modifications; PD, protein degradation; AC, actin and associated proteins; TT, transmembrane transport; PX, peroxisome; VE, vesicular trafficking and exocytosis; EM, extracellular matrix; SM, sugar metabolism; AA, amino acid metabolism; NT, nucleotide metabolism; LP, lipid metabolism; DT, detoxification; UN; unknown.

In contrast, the spectrum of biological functions of the anti-regulated paralogs is generally distinct from that of the co-regulated paralogs (r = 0.33) or of the paralog pairs in which one member is constitutively expressed (r = 0.32). The anti-regulated paralogs predominantly contribute to the categories of genes coding for mitochondrial proteins, components of actin cytoskeleton, and proteins involved in signaling and transmembrane transport, metabolism of sugars and lipids, and detoxification. This identified spectrum of biological functions of anti-regulated paralogs suggests that in testes a number of rather basic cell biological mechanisms are replaced with their modified counterparts during germline differentiation. One well-documented example of such replacement is the substitution of the broadly expressed proteasome components with their testis-biased homologs (Yuan et al. 1996; Ma et al. 2002; Zhong and Belote 2007). In agreement with these reports, we observed a downregulation of the broadly expressed Pros28.1 and an upregulation of the testis-biased paralog Pros28.1B. Identification of other numerous anti-regulated paralog pairs in our analysis provides the basis for future studies of the roles of these paralogs in the known diverse events of germline differentiation. For example, the observed change in repertoire of the mitochondrion proteins probably supports extensive remodeling of mitochondria in maturing spermatids, which leads to the formation of a peculiar fused mitochondrion derivative—nebenkern (Tokuyasu 1975). Changes in the expression of actin genes may underlie the vast remodeling of the actin cytoskeleton in spermatogenesis, in particular during sperm individualization (Noguchi and Miller 2003). “Replacement” of the genes involved in metabolism of sugars and lipids is probably related to the fundamental differences in metabolism that have been observed between the male germline and somatic tissues (Geer et al. 1972).

To further characterize the roles of the paralog-encoded proteins in testes, the sets of identified paralogs were compared to the list of proteins identified in mature sperm (Dorus et al. 2006). The results of this analysis are presented in supplemental Table 3, a–c, and are summarized in Table 2. The categories of paralogs that prominently contribute to sperm proteome included the genes coding for the microtubule components and associated proteins, mitochondrial proteins, proteases, and enzymes involved in sugar metabolism. As expected, this spectrum matches the spectrum of the testis-biased genes that code for the sperm components, as described above. Further analysis of the individual paralog pairs showed that in the 27 pairs in which only one paralog codes for a sperm protein, this paralog was almost always (in 24 pairs) testis biased, rarely (in 3 pairs) constitutively expressed, and never downregulated. In contrast, in the 17 pairs where both paralogs (one of which is testis biased) code for the sperm proteins, the other paralog was equally likely to be testis biased (in 6 pairs), downregulated in testes (in 5 pairs), or constitutively expressed (in 6 pairs). These observations suggest that when the functions of paralogs are distinct and only one paralog codes for the sperm component, high expression of this paralog during spermatogenesis is required for reproduction. However, when both paralogs encode sperm proteins, the functions of these proteins may be redundant and therefore the requirement for upregulation of both paralogs in testes is relaxed, as long as at least one paralog is testis biased.

Finally, we analyzed the available data on the phenotypes caused by mutations in the identified paralogs. This survey showed that the proportion of the essential genes (i.e., genes with known lethal mutations) among the testis-biased genes that have differentially expressed paralogs—either testis biased or downregulated in testes—is two times lower than among the other classes of genes included in the analysis (Table 2). Taking into consideration that the mutant data are far from comprehensive, this observation may reflect the bias in the researchers' attention to the different types of genes. The other explanation, which we favor, is that the testis-biased genes that have differentially expressed paralogs are specifically involved in male reproduction, rather than in other essential aspects of Drosophila development.

High interspecific variability of expression patterns of the testis-specific paralogs:

A number of studies have shown that genes related to reproduction evolve more rapidly than other genes (Jagadeeshan and Singh 2005). In particular, rapid evolution has been suggested for the male-biased genes (Meiklejohn et al. 2003) as well as for male reproductive proteins (Swanson et al. 2001). We hypothesized that testis-biased paralogs may evolve especially rapidly, because gene duplications that result in creation of the paralogs produce redundancy in gene functions and therefore relax the constraints on the expression levels. To test this hypothesis, we analyzed the pattern of gene expression in the testes of Drosophila simulans, as compared to the gonadectomized males, using microarrays. We focused our analysis on the differentially expressed paralogs that were detected in Drosophila melanogaster. The results are presented in supplemental Table 3, a–c, along with the D. melanogaster data.

We estimated conservation of gene regulation between the two species for the entire transcriptome represented on our microarrays for testis-biased genes only (supplemental Table1) and for the different groups of paralogs (supplemental Table 3, a–c). In addition, as a reference, we selected a group of constitutively expressed genes that do not have differentially expressed paralogs. The genes that showed ambiguous expression data due to their low level of expression or to high variability between the replicates were excluded from this analysis. Changes in regulation were detected as transition between the core expression patterns, which included “testis biased” (upregulated in testes; q < 0.2), “downregulated” (q < 0.2), and “constitutively expressed” (<1.2-fold signal difference between testes and gonadectomized males). The results are summarized in Tables 3 and 4.

TABLE 3.

Proportion of genes changing regulation in testis between D. melanogaster and D. simulans within different groups of genes

Group	No. of genes with same regulation	No. of genes with changed regulation	Total	Proportion of genes with changed regulation
Entire gene set	6275	2183	8458	0.26
Testis-biased genes	113	57	170	0.34*
Testis-biased paralogs	50	62	112	0.55**
Downregulated paralogs	30	17	47	0.36
Constitutively expressed paralogs	75	30	105	0.29
Constitutively expressed genes with no differentially expressed paralogs	1990	433	2423	0.18**

^*Significant (P < 0.05) and ^**highly significant (P < 0.01) difference from the entire gene set.

TABLE 4.

P-values of Fisher's exact test for the differences in proportion of genes changing regulation in testis between D. melanogaster and D. simulans

Group	Entire gene set	Testis-biased genes	Constitutively expressed genes with no differentially expressed paralogs	Testis-biased paralogs	Downregulated paralogs
Testis-biased genes	P = 0.027
Constitutively expressed genes with no differentially expressed paralogs	P = 2.2e-16	P = 2.35e-6
Testis-biased paralogs	P = 5.96e-11	P = 0.0003	P = 2.2e-16
Downregulated paralogs	P = 0.131	P = 0.73	P = 0.003	P = 0.037
Constitutively expressed paralogs	P = 0.5028	P = 0.425	P = 0.009	P = 6.95e-5	P = 0.35

As predicted, the testis-biased genes have a higher proportion of genes that change regulation (34%) as compared to the entire set of genes on the microarray (26%). Although this difference is not large, it is statistically significant (Table 4). Interestingly, the testis-biased paralogs show an even higher proportion of genes with changed expression (55%); in fact, it is the greatest difference among all analyzed groups of genes, and the difference is highly significant (Table 4). In contrast, genes that are constitutively expressed and do not have differentially expressed paralogs in D. melanogaster show the most conserved expression pattern (only 18% of such genes are differentially expressed between testes and gonadectomized males in D. simulans).

Our data support and extend the previous observations (Meiklejohn et al. 2003) that testis-biased genes show more variability in expression level than other genes. We demonstrated that the co-regulated testis-biased paralogs are most variable among the testis-biased genes. To expand this analysis to another Drosophila species, we performed real-time RT–PCR for a selection of 34 genes in D. melanogaster, D. simulans, and in their more distant relative, D. erecta (supplemental Table 4). Nineteen genes did not show any differences in expression pattern between the three species whereas 15 genes showed altered regulation in D. simulans, D. erecta, or both. The majority of the genes examined belong to the group of the testis-biased paralogs. This likely explains why this set is also significantly enriched for the genes that show change in regulation between species (P = 0.028) as compared to the whole transcriptome, thus supporting the microarray data. The same number of changes (10) was observed between D. melanogaster and D. erecta as between D. melanogaster and D. simulans despite expected greater differences between D. melanogaster and D. erecta on the basis of phylogeny. This result is likely due to the relatively small sample size.

Interspecific differences in expression of paralogs are correlated with their regulation in testes:

Interspecific differences in expression of the testes-biased genes, described above, suggest that expression of the paralogs of these genes may be also altered. To address this hypothesis, we compared the bias of gene expression toward testes in D. melanogaster and in D. simulans, focusing on the sets of (i) the paralogous pairs in which both members are testis biased in D. melanogaster (the co-regulated paralogs), (ii) the pairs of paralogs in which one member is testis biased and the other is downregulated in testes of D. melanogaster (the anti-regulated paralogs), and (iii) the pairs of paralogs in which one member is testis biased and the other is constitutively expressed in D. melanogaster testes and gonadectomized males. To perform these comparisons, the logarithm of the fold change in microarray signal between testes and gonadectomized males in D. melanogaster (logFC Mel in supplemental Table 3, a–c) for each gene was subtracted from the corresponding parameter observed in D. simulans (logFC Sim in supplemental Table 3, a–c). Positive values indicate an increase in testis bias of expression in D. simulans, as compared to D. melanogaster. Conversely, the negative values indicate a decrease in the testis bias in D. simulans.

The results of this analysis showed that, in general, the genes that are testis biased in D. melanogaster show a smaller bias toward testes in D. simulans (Figure 3). Among the paralogous pairs that are co-regulated in D. melanogaster, 75% (standard error 5%) show correlated changes in the testis bias between D. melanogaster and D. simulans, whereas only 25% demonstrate the anti-correlated changes when one gene becomes more testis biased and the other less testis biased (Figure 3A). In contrast, the majority (74%, standard error 5%) of the paralogous pairs that are anti-regulated in D. melanogaster show anti-correlated changes in the testis bias between D. melanogaster and D. simulans, whereas only 26% of the paralogous pairs in this group exhibit correlated changes in expression bias (Figure 3B). Finally, the paralogous pairs in which one member is testis biased and the other is constitutively expressed in D. melanogaster showed equal proportions of the correlated (45%) and anti-correlated (55%, standard error 4%) changes in the testes bias in D. simulans (Figure 3C).

Figure 3.—

Correlated changes in expression log ratio between D. melanogaster and D. simulans for three groups of paralogs. Each data point represents one paralogous pair.

The observed results are consistent with the hypothesis that functional roles of the paralogous genes are at least partially redundant. In this case, the balance in expression of the anti-regulated paralogs may be essential for male gametogenesis, and indeed the interspecific changes in expression of one such paralog in testes are usually offset by the opposite change in expression of the other paralog. Further, the observed correlation between expression patterns of paralogous genes and interspecific changes in their expression suggests that regulation of these genes in testes is coordinated. A simple mechanistic model that explains this phenomenon suggests that for each paralogous pair, the same mechanism (for example, the same transcription factor) is often responsible for regulation of both genes in testes. The activity of such mechanisms may vary among species, leading to coordinated interspecific changes in the expression of the paralogs. Surprisingly, this coordination was found not only for the pairs of paralogous genes that exhibit upregulation in testes, but also for the pairs of anti-regulated paralogs, suggesting that the same mechanism(s) may control both up- and downregulation of the related genes in testes. Further identification and dissection of such mechanisms will significantly advance understanding of the gene expression program that controls male gametogenesis.