Theme and variation in the evolution of insect sex determination

Abstract

The development of dimorphic adult sexes is a critical process for most animals, one that is subject to intense selection. Work in vertebrate and insect model species has revealed that sex determination mechanisms vary widely among animal groups. However, this variation is not uniform, with a limited number of conserved factors. Therefore, sex determination offers an excellent context to consider themes and variations in gene network evolution. Here we review the literature describing sex determination in diverse insects. We have screened public genomic sequence databases for orthologs and duplicates of 25 genes involved in insect sex determination, identifying patterns of presence and absence. These genes and a 3.5 reference set of 43 others were used to infer phylogenies and compared to accepted organismal relationships to examine patterns of congruence and divergence. The function of candidate genes for roles in sex determination (virilizer, female‐lethal‐2‐d, transformer‐2) and sex chromosome dosage compensation (male specific lethal‐1, msl‐2, msl‐3) were tested using RNA interference in the milkweed bug, Oncopeltus fasciatus. None of these candidate genes exhibited conserved roles in these processes. Amidst this variation we wish to highlight the following themes for the evolution of sex determination: (1) Unique features within taxa influence network evolution. (2) Their position in the network influences a component's evolution. Our analyses also suggest an inverse association of protein sequence conservation with functional conservation.

Graphical abstract

Sex determination and dimorphic somatic differentiation are common in insects, such as the milkweed bugs (Oncopeltus fasciatus) shown here. Variation in the functions and structure of genes implicated in insect sex determination is not uniformly distributed. This article reviews sex determination mechanisms in insects and examines relationships between genes involved in sex determination to identify apparent patterns in their evolution.

RESEARCH HIGHLIGHTS

• We review sex determination, a well‐studied process in Drosophila and one increasingly explored in diverse insects.

• Structural conservation in sex determination candidate genes masks mechanistic variation in this process.

• More functionally conserved genes in the sex determination network are paradoxically more divergent in their protein sequence.

• Knockdown of several candidate genes in Oncopeltus does not support their function in sex determination or dosage compensation in this species.

1. INTRODUCTION

A gene network (GN) is an interconnected set of genes that specifically regulates the development of a structure in a multicellular organism (McQueen & Rebeiz, 2020; Peter & Davidson, 2011). Gene regulatory networks (GRNs, a subtype of GN encompassing only regulatory factors and not effector proteins), and their evolution have been the particular focus of many recent studies on the evolution of development. Over evolutionary time GNs can be co‐opted to regulate the development of a novel structure or modified, resulting in phenotypic change to the same structure (Hansen, 2003; McQueen & Rebeiz, 2020; Peter & Davidson, 2011). Understanding how the components of existing GNs evolve is critical for understanding the interplay of developmental processes in individuals and the long arch of phenotypic evolution. Evolutionary change has proven to be uneven across GNs (Stern & Orgogozo 2009), and the challenge now is to identify widespread themes and to appreciate unique variations underlying phenotypic diversity.

Here we review the existing theory and experimental reports from the sex determination GN of diverse insects. We also present novel findings from the examination of sex determination in the milkweed bug Oncopeltus fasciatus, a phylogenetically informative insect species. Based on this information, we identify a set of themes for the evolution of sex determination, including the observations that (1) there may be unique influences on GN evolution within certain taxa and (2) a component's position in the GN influences its evolution.

The genes that constitute a GN encode diverse proteins. Genetic networks are composed of cell signaling ligands, receptors and intracellular signal transducers, transcription factors, splicing regulators, and proteins that affect differentiation, such as enzymes, cytoskeletal elements and other structural proteins. In principle, phenotypic evolution could occur through changes in any of these components, but in actuality differences are not randomly or uniformly distributed. One reason for this pattern is pleiotropy. For example, consider that changes in the DNA sequence of a transcription factor binding site may affect affinity at the target sequence and the resulting degree of expression for the target gene (Ramos & Barolo, 2013; Rogers & Bulyk, 2018). Changes in the protein‐coding sequence of the same transcription factor might have a similar effect on the target gene's expression, but while the first change would be limited to one interaction, the second would likely be pleiotropic, affecting all downstream genes regulated by the transcription factor (reviewed by Caro & Mallarino, 2020; Hoekstra, 2006). Therefore, a protein with multiple critical roles in development is more likely to incur lethal or deleterious mutations in its protein‐coding sequence. Gene duplication illustrates this tendency, and one evolutionary resolution to pleiotropy (Hughes, 1994; Lynch & Conery, 2000). One copy of a duplicated gene can serve its ancestral function, making the second dispensable in that regard and freeing it to accumulate new mutations that may alter its function dramatically, or allow it to be co‐opted into another GN entirely (Peter & Davidson, 2011). Alternatively, mutations in regulatory elements allow a transcription factor to evolve in its interactions with different downstream GN components, producing different temporal or spatial expression patterns (Hughes et al., 2021; Stern & Frankel, 2013).

According to Waddington (1959) the differentiation of a cell during development is analogous to a ball rolling down a landscape: it will follow the path of least resistance determined by both environmental factors and genetic underpinnings. Here, proteins upstream in the developmental cascade have a more dramatic effect on the final phenotype than proteins downstream in the pathway. In this conceptual model, downstream components are thought to control subtler phenotypes and might be more likely to sustain evolutionary changes. Similarly, we propose that the evolution of GN components is biased by their placement within the GN. As we compare developmental mechanisms in extant species, this bias results in patterns of conservation and divergence, at either the structural or functional levels, among GN components. These patterns suggest another analogy: the motifs of theme and variation seen in Western classical music (Angelini & Smith, 2019). Here we will explore the themes and variations in the evolution of insect sex determination.

The regulation and development of sex are central questions in developmental biology. The biological importance of sex might suggest conservation among sex determination mechanisms; however, developmental and genetic studies have revealed great diversity in the ways animals determine sexes and develop sexual dimorphisms (Bopp et al., 2014; Saccone et al., 2002; Schütt & Nöthiger, 2000). Insects are the most numerous and species‐rich group of animals on the planet, and although insects are widely studied, their systems of sex determination and differentiation have primarily been examined in Holometabola, the clade with complete metamorphosis (Gempe & Beye 2011). Even within Holometabola, high levels of diversity have been observed in the sex determination mechanisms of different species (Bopp et al., 2014; Saccone et al., 2002; Schütt & Nöthiger, 2000).

In species where the number of one or more chromosomes correlates with sex, mechanisms exist to normalize gene expression, compensating for different genetic dosage of these sex‐associated chromosomes. Therefore, dosage compensation is an important downstream phase of sex determination. In several model species, dosage compensation is initiated by a component of the sex determination network, but the mechanism by which different species equalize gene expression in females and males differs at multiple levels. Different species link sex determination to dosage compensation via the activity of different genes (Kiuchi et al., 2019; Krzywinska et al., 2021; Qi et al., 2019). While some insects upregulate the heterogametic chromosome (Y or W), others downregulate the homogametic chromosome (X or Z). Moreover, epigenetic controls appear to play an important role in regulating sex chromosome gene expression (Samata & Akhtar, 2018).

Due to this high mechanistic diversity, and the tractability of many insects in the laboratory, insect sex determination offers an excellent model to study how GNs may change over evolutionary time. Below we review what is known about this process in different insect groups.

1.1. Sex determination in Diptera

Sex determination is best understood in Drosophila melanogaster, where it occurs autonomously in somatic cells (Figure 1; Robinett et al., 2010; Schütt & Nöthiger, 2000). During mammalian development most somatic sexual dimorphisms occur in response to widely circulating endocrine signals, such as the steroids estrogen and testosterone, which are produced by the gonads. However, insects are not known to have an analogous endocrine regulation of sexual differentiation. Each individual cell of a fruit fly has a chromosome‐level sexual identity that influences differentiation in structures that exhibit sexual dimorphism. The relative dosage of X chromosomes ultimately determines sex in the fly, at a stage before any dosage compensation has occurred. The level of transcription of X‐chromosome linked signal element (XSE) genes (runt, sisA, scute, and unpaired) determines whether an early promoter is activated at Sex‐lethal (Sxl; ​​Erickson & Quintero, 2007). deadpan (dpn) is an important autosomal co‐factor of the XSE proteins (Barbash & Cline, 1995; Salz & Erickson, 2010; Younger‐Shepherd et al., 1992).

Figure 1.

The sex determination mechanism as described from Drosophila melanogaster. Proteins named in black are functional. Proteins named in gray are present, but not functionally relevant. Proteins named in a gray box with white text are not produced. Superscripts indicate sex‐specific isoforms

Production of XSE proteins from two X chromosomes promotes early Sxl expression, which is then maintained by an auto‐regulatory loop that leads to female developmental fates. Another X‐linked gene, splicing necessary factor/sans fille (snf), and three autosomal genes, virilizer (vir), female‐specific‐lethal‐2‐d (fl(2)d), and splicing factor 45 (spf45), encode proteins that help promote the auto‐regulatory splicing of Sxl in females from a later‐acting maintenance promoter (Niessen et al., 2001; Salz & Erickson, 2010). In presumptive males, the single X chromosome does not produce the same threshold level of XSE proteins, and Sxl is spliced into an isoform that does not produce a functional protein.

vir and fl(2)d encode nuclear proteins necessary for the auto‐regulatory splicing of Sxl (Niessen et al., 2001) and for female‐specific splicing of transformer (tra) and repression of dosage compensation in females (Hilfiker et al., 1995). Some vir mutations are lethal in both sexes before pupation, implying that vir is involved in other critical processes. vir is predicted to encode a nuclear envelope protein involved in messenger RNA (mRNA) transport (Niessen et al., 2001). These proteins are likely general splicing factors, since the primary transcript of the homeotic gene Ultrabithorax is another known target of vir and fl(2)d (Burnette et al., 1999).

Sex‐specific expression and splicing of downstream genes control sexually dimorphic anatomy, dosage compensation, and behavior (Serna et al., 2004). Sxl regulates splicing of tra, so that functional Tra protein, another RNA‐splicing regulator, is only expressed in females (Boggs et al., 1987; Inoue et al., 1990). transformer‐2 (tra‐2) is expressed in both sexes and encodes a protein with an RNA‐binding domain and an arginine/serine‐(RS)‐rich region, suggesting it also functions in RNA splicing (Mattox et al., 1990). The Tra/Tra‐2 complex is necessary for female‐specific splicing of doublesex (dsx; Martín et al., 2011). Tra‐2 is also required for spermatogenesis (Belote & Baker, 1983; Mattox et al., 1990) and for expression of fruitless (fru) in males (Salz & Erickson, 2010).

doublesex is a critical gene for sex determination, which encodes a transcription factor, spliced to produce different isoforms in each sex (Salz & Erickson, 2010). Dsx isoforms promote female or male differentiation by repressing and activating different target genes. The female isoform, Dsx^F, together with the products of intersex (ix) and hermaphrodite (her), directly regulates the transcription of terminal sex differentiation proteins and represses male development (Burtis & Baker, 1989; ​​Garrett‐Engele et al., 2002). Conversely, Dsx^M represses female development (Kopp, 2012). The default protein product of fru is male‐specific, while splicing of fru mRNA in females by the Tra/Tra2 heterodimer results in a nonfunctional protein and default female development (Heinrichs et al., 1998).

Somatic cells support the germline in both female and male Drosophila melanogaster. Here, not only does somatic sex need to be determined during development, but also maintained throughout reproductive life. A transcription factor, chronologically inappropriate morphogenesis (chinmo), is necessary for maintaining male identity of cyst stem cells in the testes; mutations in chinmo result in cyst stem cells becoming follicle‐like in morphology and producing follicle markers (Ma et al. 2016). chinmo maintains male identity by repressing production of tra and the splicing factors vir and fl(2)d in male cyst cells (Grmai et al., 2018). Expression of Sxl, tra, tra2, and dsx in the soma is also important for correct initial germline sex determination (Nöthiger et al., 1989). Two additional genes, ovo and ovarian tumor (otu), are also required for germline sex determination in D. melanogaster. The expression of ovo, which encodes a transcription factor, and ovarian tumor (otu), a cytoplasmic protein, is female‐specific, and both genes are required for sex‐specific splicing of Sxl (Bopp et al., 1993).

Across different insect model species not all components of the sex determination GRN have been described (Figure 1). However, it is clear even from this incomplete picture that major differences exist. In fact, the XSE system of D. melanogaster does not appear widespread among Diptera, where other species rely on a dominant male‐specifying factor to direct alternative splicing of tra and dsx. Among Tephritidae, such as the medfly Ceratitis capitata, the Y‐linked gene MoY specifies male development, via regulation of tra splicing (Meccariello et al., 2019), although this protein has not been described at the molecular level and does not have obvious homology to known splicing factors. Similarly, the house fly Musca domestica has a dominant male determiner, Mdmd, which directs male‐specific splicing of tra and dsx in that species (Sharma et al., 2017). Mdmd originated by duplication of a conserved spliceosomal factor encoded by CWC22/nucampholin. In mosquitos, various dominant male‐inducing splicing factors exist. In Aedes aegypti and Anopheles gambiae, Nix and Yob, respectively induce male‐specific splicing of dsx (Hall et al., 2015; Krzywinski et al., 2016). Nix appears to be derived from a duplication of tra‐2 within Aedes. In Anopheles species, expression of Yob in An. gambiae or its ortholog Guy1 in An. stephensi, also causes embryonic death when misexpressed in otherwise genetically female embryos (Criscione et al., 2016; Krzywinski et al., 2016). Yob/Guy1 in Anopheles have no obvious orthologs in other insects.

Diptera have provided valuable genetic models to describe sex determination. Across the order, the genes tra and dsx appear to be conserved points of integration for upstream chromosomal signals of sex. However, even among Diptera the nature of those primary signals varies widely.

1.2. Sex determination in Lepidoptera

At greater evolutionary distances, the Drosophila sex determination model appears less conserved. In most species of Lepidoptera, sex is determined by a ZW chromosome system, where heterogametic (ZW) individuals develop as female and homogametic (ZZ) individuals are male (Traut, et al. 2006). In some species the W chromosome acts dominantly to induce female development, while in others the occurrence of Z0 females suggest a Z‐dosage mechanism determines sex. In the silkmoth Bombyx mori the female‐specific W chromosome contains mostly repetitive elements (Katsuma et al., 2018). However, some of these have been mapped to a locus, called Fem, which appears to produce a PIWI‐interacting RNA (piRNA) essential for female development (Kiuchi et al., 2014). The Fem piRNA directs degradation of mRNAs from a gene called Masculinizer (Masc), which are transcribed in all individuals but only persist in males due to the absence of Fem piRNAs (reviewed by Yang, et al., 2021). Masc encodes a zinc‐finger protein required for male development and dosage compensation in Bombyx mori (Kiuchi et al.,2019). Although the next components in this GN are still being investigated, a strong candidate linking Masc activity to male development is the P‐element somatic inhibitor (PSI), which is required for normal male development (Yang et al. 2021) and for male‐specific splicing of dsx (Suzuki et al., 2008). Lepidoptera appear to lack a tra ortholog (Figure S1), but the B. mori tra‐2 ortholog retains an RNA‐binding domain (Niu et al., 2005). However, B. mori dsx transcripts lack a binding site comparable to that of Tra/Tra‐2 in D. melanogaster (Suzuki, et al., 2001). Although an ix ortholog is also present in the B. mori genome, Ix and Dsx failed to interact in a yeast two‐hybrid experiment (Fujii & Shimada, 2007). Compared to the fruit fly, the silkmoth presents a very different paradigm for sex determination (Yang et al., 2021). The Fem piRNA appears to be limited to B. mori, but Masc has orthologs in other Lepidoptera and PSI is conserved across the insects (Figure 2).

Figure 2.

Patterns of orthology across the insects for genes involved in Drosophila melanogaster sex determination. Dark boxes represent a single ortholog of the D. melanogaster candidate gene in all surveyed species in that taxon (order or superfamily). Variations in the ortholog number in the taxon are given by the text. Light gray boxes indicate the absence of identifiable orthologs. The only noninsect sequence that has highest BLAST similarity to fl(2)d is Varroa destructor XP_02266598. However, this sequence has greater similarity to the pre‐mRNA‐splicing regulator WTAP from Apis mellifera (XP_006566329)

From this review, it is clear that Lepidoptera have features of the sex determination GN unique to this taxon, such as the role of Masc. Most genes identified in D. melanogaster sex determination have structurally conserved orthologs in Lepidoptera (Figure 2), but the extent of their functional conservation remains unclear in most instances.

1.3. Sex determination in Hymenoptera

Hymenoptera are characterized by haplodiploidy, in which females are diploid and males are typically haploid. Unlike many other animals, they do not have sex chromosomes. Instead, sex is determined by a system called complementary sex determination, which relies on heterozygosity at a single locus to initiate female development (Cook, 1993; Schmieder et al., 2012). In the honeybee Apis mellifera, the initial sex determination locus is termed complementary sex determiner (csd; Beye et al., 2003; Cook, 1993). Heterozygotes at csd develop as females while haploid, hemizygous embryos develop as males. In situations where local genetic diversity is low, diploid individuals homozygous at the csd locus also develop as males (Cook & Crozier, 1995; Zayed & Packer 2005).

csd controls sex by regulating the splicing of feminizer (fem), which is also a splicing regulator. Only the female‐specific fem isoform produces a functional protein (Hasselmann et al., 2008). csd originated as a recent duplication of fem and occurs only within the genus Apis. It is not yet known what ultimate mechanisms might regulate fem in other Hymenoptera. However, the occurrence of diploid males in in‐bred populations suggests that some form of complementary sex determination is still involved (Cook, 1993; Cook & Crozier, 1995). Interestingly, independent duplications of fem appear in several different lineages of Hymenoptera (Koch et al., 2014; Privman et al., 2013), although the functional significance of these paralogs is still unknown.

Not all Hymenoptera exhibit complementary sex determination. In the wasp Nasonia vitripennis eggs are produced with a maternal deposition of mRNA encoding orthologs of fem and tra2, which are necessary for female development (Geuverink et al., 2017; Verhulst & Beukeboom, 2010). Translation from these transcripts requires expression of wasp overruler of masculinization (wom). This locus is silenced by epigenetic imprinting on maternally derived chromosomes. Therefore, female development requires fertilization and a paternal copy of wom capable of expression (Zou et al., 2020). Haploid embryos, which are not fertilized, have only the silenced maternal copy of wom, leaving tra transcripts in the oocyte untranslated, and resulting in male development. It is unlikely the N. vitripennis sex determination mechanism is widespread, however, since wom orthologs do not appear in other insect genomes (Figure 1).

As seen with Lepidoptera, unique features of the biology of Hymenoptera, such as haplodiploidy, are correlated with unique aspects of sex determination. Hymenopteran fem has been suggested as a tra ortholog based on the presence of shared arginine/serine (RS)‐ and proline‐rich domains (Hasselmann et al., 2008), making it a point of functional conservation in the sex determination GN. Most other D. melanogaster sex determination candidate genes have orthologs in Hymenoptera, but functional roles have not yet been described for them.

1.4. Sex determination in other insect groups

Studies in Diptera, Lepidoptera and Hymenoptera provide informative models of sex determination. However, relatively few functional experiments from other insect groups exist to provide more evolutionary context. One obstacle has been low sequence conservation among key genes, such as tra, which is described from only a few insect groups. Putative tra orthologs from the beetles Tribolium castaneum (Shukla & Palli, 2012) and Trypoxylus dichotomus (Morita et al., 2019) exhibit sex‐specific isoforms, regulate dsx splicing, and are required for sex‐specific development in those species. Genes with domain‐level homology to tra have also been described from the louse Pediculus humanus, the hemipteran Rhodnius prolixus, and the cockroach Blattella germanica (Wexler et al., 2019). Of these species, only R. prolixus exhibited sexual dimorphism in tra splicing. Tests of tra gene function have not yet been reported from P. humanus or R. prolixus, but RNA interference (RNAi) targeting tra in juvenile female B. germanica disrupted genital development and led to typically male behaviors (Wexler et al., 2019).

dsx has been described more broadly among insects, and may likely be the most conserved component of sex determination in animals in its protein structure and in its developmental function (Geuverink & Beukeboom, 2014; Herpin & Schartl, 2015; Schütt & Nöthiger, 2000; Shukla & Nagaraju 2010). In the beetle T. castaneum (Shukla & Palli, 2012), Onthophagus ssp. (Beckers et al., 2017; Kijimoto et al., 2012), Trypoxylus dichotomus (Ito et al., 2013) and Cyclommatus metallifer (Gotoh et al., 2014), dsx orthologs are spliced in sex‐specific patterns and required for sexually dimorphic adult traits in both females and males. In Auchenorrhyncha, such as the brown planthopper Nilaparvata lugens, dsx has undergone duplication, although functions in sex determination appear retained by only one paralog (Zhuo et al., 2018). At the functional level, dsx may only be required for male development in certain insect lineages (Wexler et al., 2019). In one hemipteran, the milkweed Oncopeltus fasciatus, dsx activity is required for multiple sexually dimorphic traits in both sexes (Just et al., 2021), similar to its requirement in Holometabola. However, in N. lugens, only males require dsx activity (Zhuo et al., 2018). Similarly, dsx is only required for male development in the cockroach B. germanica (Wexler et al., 2019).

fruitless orthologs have also been identified across the insects (Figure 2), with conserved patterns of splicing (Salvemini et al., 2010). Similar to its role in D. melanogaster, fru is required for male abdominal musculature and courtship behavior in mosquitos (Meier et al., 2013). RNAi experiments have revealed conserved roles for fru in male‐specific behaviors in the silkmoth Bombyx mori (Yang et al. 2021) and in cockroach Blattella germanica (Clynen et al., 2011). However, in the hemipteran O. fasciatus RNAi revealed requirements for fru activity in the development of male and female genitalia, as well as female‐specific abdominal sternite shape (Just et al., 2021). This finding suggests that while fru is structurally conserved, and consistently involved in sex determination, the extent of its developmental role may be diverse.

1.5. Dosage compensation

Dosage compensation is the process that equalizes gene expression for both sexes in species with heterogametic sex chromosomes. Mammals, where females are the homogametic sex, randomly inactivate most genes on one of the X chromosomes (Disteche, 2012). Hypotranscription of both X chromosomes in C. elegans hermaphrodites equalizes expression of X‐linked genes to the level present in X0 males (Hsu & Meyer, 1993). Heteromorphic sex chromosomes and dosage compensation appear wide‐spread, and have been documented in representatives of diverse insect lineages, including several Diptera and Lepidoptera, which are detailed below, as well as the beetle Tribolium castaneum (Mahajan & Bachtrog, 2015; Prince et al., 2010), the strepsipteran Xenos vesparum (Mahajan & Bachtrog, 2015), the cricket Gryllotalpa fossor (Rao & Padmaja, 1992), and the aphid Acyrthosiphon pisum (Richard et al., 2017). The component of the sex determination GN that initiates dosage compensation varies among model species: Sxl in D. melanogaster (Penalva & Sánchez, 2003), Masc in Bombyx mori (Kiuchi et al., 2019), Guy1 in Anopheles stephensi males (Qi et al., 2019) and femaleless in An. stephensi females (Krzywinska et al., 2021).

Insect dosage compensation has been best studied in D. melanogaster, where males only have a single X chromosome (Gelbart & Kuroda, 2009). Expression of X‐linked loci in male fruit flies are doubled to approximate expression from the two X chromosomes in females. Five proteins make up the dosage compensation complex: MSL1, MSL2, MSL3 (named for their male‐specific lethal loss‐of‐function phenotypes), MOF (Males‐absent‐on‐the‐first) and MLE (Maleless) (Penalva & Sánchez, 2003; Gelbart & Kuroda 2009). These five proteins interact with two noncoding RNAs on the X chromosome (Meller & Rattner, 2002). Male‐specific‐lethal‐1 (msl‐1), msl‐2, and msl‐3 are transcribed in both sexes, however only msl‐2 is necessary and sufficient to assemble the complex (Kelley et al., 1995). In female D. melanogaster, the splicing factor Sxl represses translation of msl‐2 (Bashaw & Baker, 1996), preventing X upregulation in females. Histone modification ultimately adjusts chromosome‐level gene expression, as MSL3 and MOF acetylate histone H4 to increase chromatin accessibility on the male X chromosome (Keller & Akhtar, 2015; Smith et al., 2000).

While orthologs of the msl genes are found in a wide variety of insects and vertebrates (Figure 2; Marín & Baker 2000; Liu et al., 2008), their functional conservation in dosage compensation is questionable. Mosquitos in the subfamily Culicinae, such as Aedes and Culex, have a sex‐determining locus present on an otherwise monomorphic chromosome (Gilchrist & Haldane, 1947; Jiang et al., 2015). Such species should have no need for dosage compensation, and Culicoidea genomes do not contain orthologs of msl1, although they retain other dosage compensation genes (Figure 2). Despite the absence of msl1, an X chromosome and dosage compensation have evolved independently in Anopheles (Jiang et al., 2015; Rose et al., 2016). Sex chromosomes and dosage compensation have evolved multiple times independently in Diptera (Vicoso & Bachtrog 2015). During this process, a new sex‐determining chromosome has supplanted a previous sex‐determining system. Therefore, the proteins responsible for dosage compensation in D. melanogaster may not be involved in other Diptera with X chromosomes. In the nematoceran Sciara ocellaris, antibody staining for MSL1, MSL3, MOF, and MLE is comparable on autosomes and the X chromosome (Ruiz et al., 2000).

Complete dosage compensation is known from several Lepidoptera, including Papilio species and Plodia interpunctella (Huylmans et al., 2017), but the role of MSL proteins in these species is unclear. In the cotton bollworm Helicoverpa armigera the msl‐1 homolog is located on the Z chromosome. RNAi targeting msl‐1 caused the upregulation of several Z‐linked genes in males, but no effect in females, suggesting a role is dosage compensation (Zhang, et al., 2019). In the silkmoth Bombyx mori, dosage compensation is promoted by the protein encoded by Masculinizer (Masc; reviewed by Katsuma et al., 2018). Orthologs of all five msl genes have been identified in this species, but it is not clear if their proteins interact with Masc or any other proteins involved in sex determination (Liu et al., 2008).

Here, we use comparative genomic and experimental developmental data from insect sex determination to examine what particular rules may underlie GN evolution and how different rules might apply to different types of pathway components and hierarchical levels. We previously investigated orthologs of doublesex, intersex and fruitless during the development of somatic sexual dimorphisms in the large milkweed bug Oncopeltus fasciatus (Just et al., 2021); here we present additional functional data for orthologs of female‐lethal‐2‐d, virilizer, tra‐2, msl‐1, msl‐2, and msl‐3.

2. MATERIALS AND METHODS

2.1. Sequence orthology

Putative orthologs were initially identified using NCBI BLAST searches using candidate protein sequences from Drosophila melanogaster or other model species (Table S1) against specific insect taxa listed in Figure 2. In ambiguous cases, putative orthologs were used as queries in reciprocal BLASTp searches of the D. melanogaster proteome. Sequences were assumed to be orthologs only if reciprocal BLAST searches produced hits with an E‐value less than 10⁻²⁰. Orthologs from Oncopeltus fasciatus were screened using local BLAST databases constructed from the available gene set (GCA_000696205.1). Orthologs from Boisea trivittata, Jadera haematoloma were obtained from de novo transcriptome assemblies of pooled whole adult bodies (GenBank accessions OL614968 ‐ OL614973). Orthology was confirmed using phylogenetic inference.

By these criteria tra orthologs described outside Diptera were not recovered. Therefore, supplemental searches were performed using the amino acid sequences from Tribolium castaneum Tra (AFQ62109), Rhodnius prolixus Tra (QGB21091), Blattella germanica Tra (QGB21093) and Apis mellifera Fem (NP_00112830), and this strategy identified several additional orthologs (Figure S1).

Alignments of amino acid sequences were performed by ClustalOmega (version 1.2.4) using the default parameters. Alignments included all amino acids. doublesex paralogs were also trimmed to include only the conserved DNA‐binding OD1 domain and the OD2 or DMRTA domain. Alignment of Tra sequences was also performed after trimming to include only the sex‐determining peptide region.

Phylogenies were inferred using RAxML version 8.2.11 on a high‐performance computing cluster. Mutation rates were modeled using the BLOSUM62 matrix with a gamma distribution. The program was called as “raxml‐mpi ‐d ‐f a ‐x 123 ‐p 123 ‐# autoMRE ‐m PROTCATBLOSUM62 ‐s X.phy ‐n X” where X stands in for the alignment name. Trees were not rooted, but included outgroup sequences from Crustacea and Chelicerata when they could be identified. Node supports were determined by bootstrapping. Phylogenies generated in this study are available at https://github.com/aphanotus/tree.comparison.

2.2. Phylogenetic analyses

Phylogenetic analyses were conducted in R (R Core Team, 2020) using the packages phytools (Revell, 2012), phyloSignal (Keck et al., 2016), ips (Heibl, 2008) and TreeDist (Smith 2020) and custom code available at https://github.com/aphanotus/tree.comparison. To examine phylogenetic signal in gene duplication and loss, these events were mapped onto the organismal tree (Figure S2). Taxa lacking well‐annotated genomes were excluded from this analysis, including Zygentoma, Orthoptera, Gerroidea, Reduvioidea, Miroidea, Coreoidea, and Platypezoidea. All branch lengths were set to one step. Signal in duplication and loss was tested using local Moran's I, as calculated with patristic distance in 10,000 permutations by the function lipaMoran in the R package phyloSignal (Keck et al., 2016). To facilitate the comparison of gene trees with organismal relationships, we employed a generalization of the Robinson‐Foulds distance (Smith 2020), which examines tree topology without considering branch lengths. Our analysis focused on the ordinal level, and at the superfamily level for well‐sampled orders, including Hemiptera, Hymenoptera, Coleoptera, Lepidoptera and Diptera (Figure 2 and Figure S3). Relationships of these taxa were assumed from several recent phylogenetic studies among and within insects and major orders using whole genome datasets (Johnson et al., 2018; Kawahara et al., 2019; McKenna et al., 2019; Misof et al., 2014; Peters et al., 2017; Wiegmann et al. 2011). For each gene tree, tips consisted of protein sequences from individual species. These tips were collapsed to combine sequences from species in the same higher‐level taxa (orders and superfamilies) used in the organismal tree. Gene trees were excluded if they contained fewer than 10 tips at this taxonomic level. This filter removed genes with orthologs limited to one or a few taxa (e.g., her, sis‐a, and upd). Internal nodes with bootstrap support less than 50% were collapsed into polytomies using the function ips::collapseUnsupportedEdges. The extent of polytomy in an unrooted tree was quantified as 1 minus the proportion of edges (internal nodes) −1 divided by the maximum possible number of edges in the tree −3. This metric of polytomy ranges from 0 (bifurcation at every internal node) to 1 (a completely unresolved “star” tree). The extent of paraphyly was quantified after condensing taxa to the analysis level as the number of duplicated tips divided by half the number of unique taxon names in the tree. Gene trees were then compared to an organismal tree containing the same high‐level taxa using TreeDist::ClusteringInfoDistance(normalize=TRUE). Trees that agree completely have a normalized clustering information distance (nCID) of 0; those that share no common bifurcations have an nCID of 1 (Smith, 2020). Tree distance calculation requires that two trees possess the same tips. Therefore, in cases of paraphyly, taxa that were duplicated at the tips were randomly pruned down to one occurrence before calculating nCID. This process was permuted 10,000 times and the mean of all nCID samples was used in the subsequent analysis.

Most gene trees do not precisely reflect true organismal relationships. This tree distance may be due to limited information content in the amino acid alignment, or histories of neutral genetic drift or directional selection that is incongruent with other evolutionary signals. We took two approaches to provide context for tree comparisons. First, we applied the same analysis to 43 reference genes involved in basic cellular processes (Table S1), which have been used to infer the relationships among insect taxa (Angelini & Jockusch, 2008; Che et al 2017; Fang et al., 1997; Ishiwata et al., 2011; McKenna & Farrell, 2010; Pistone et al., 2016; Wiegmann et al. 2009). Second, for each comparison of gene and organismal trees, we performed 10,000 boot‐strap permutations, randomly assigning taxa (without replacement) to tips of the gene tree topology (preserving polytomies where they occur), and recalculating nCID. The lower 5th percentile of this bootstrap distribution was taken as a 95% confidence interval, below which gene trees are unlikely to match the organismal tree by chance alone. Metrics from each gene's alignment (sequence identity, portion of parsimony‐informative sites), phylogeny (polytomy and paraphyly) and comparison to the organismal tree (nCID and nCID/sequence identity) are reported in Figure S4. Figure S5 plots the relationships between several of these metrics, noting strong correlations among some. Figures S6 and S7 showe metrics from repeats of the analysis using bootstrap cutoffs of 20% and 80% to collapse polytomies. Figure S8 presents metrics for candidate genes excluded from the main analysis (see Results Section 3.4). For trees containing paraphyly, the distribution of sampled nCID values is reported for bootstrap cutoffs at 50%, 20%, and 80% in Figures S9, S10 and S11. Figure S12 reports the nCID sampling distributions for the miscellaneous gene set shown in Figure S8.

2.3. RNA interference

Loss‐of‐function phenotypes from RNAi were used to investigate gene function in the milkweed bug Oncopeltus fasciatus as previously described (Aspiras et al., 2011; Just et al., 2021). Insect stocks were obtained from Carolina Biological Supply Company and maintained as described by Liu and Kaufman (2009). Knockdown was verified by qRT‐PCR. Double‐strand RNA (dsRNA) encoding green fluorescent protein (GFP) was used as a negative control for dsRNA toxicity and injection mortality. A synthetic gene fragment (IDT gBlock) or cloned gene fragment was used as template for the synthesis of dsRNA. Primers were designed with an approximately 20‐bp exact match to the target mRNA, avoiding sequences that occur in paralogs (Table S2). A 5′ T7 polymerase promoter sequence was added to both primers. Following PCR, dsRNA was transcribed using the T7 MegaScript Transcription Kit (Fisher Scientific). Template DNA was removed by treatment with DNase I. Slow cooling to room temperature allowed dsRNA annealing. Samples were cleaned by precipitation in cold ethanol and ammonium acetate. The product was resuspended in nuclease‐free water and stored at −20°C. Structure was confirmed by a coherent band in an agarose electrophoretic gel, and dsRNA concentrations were measured using a microliter‐volume spectrophotometer. dsRNA solutions were diluted in saline buffer (0.01 mM NaPO₄, 5 mM KCl, green food coloring) and stored at −20°C. Fourth instar O. fasciatus are not yet sexually dimorphic. Bugs at this stage were anesthetized using a CO₂ pad and injected with dsRNA solutions using a borosilicate glass microcapillary, pulled on a Sutter Instruments pipette puller into an injection needle. Per bug, approximately 0.5 µl of dsRNA solution entered the left side of the abdomen.

The effectiveness of RNAi was validated independently in female and male first‐day adult whole bodies using qRT‐PCR. Gene expression was measured in control (GFP dsRNA) specimens and RNAi treatments targeting the genes of interest. Primers used for validation (Table S3) did not overlap with dsRNA sequences. Total RNA was isolated from adult O. fasciatus according to the PureLink RNA Mini Kit protocol (Invitrogen). Isolated RNA was stored at −80°C. At least three biological replicates were included for each sex. RNA concentrations from each biological replicate were determined by duplicate measures on a nanoscale spectrophotometer and diluted to 50 ng/μl. Total RNA was used as a template in reverse transcription SYBR Green real‐time PCR reactions (Invitrogen Superscript III). For each gene, exact primers were designed using the Primer3 algorithm (Rozen & Skaletsky, 2000), avoiding conserved functional domains and dsRNA regions. Dissociation curves for each reaction were used to verify that only single products were amplified. To produce quantitative template standards, clones were linearized and transcribed in vitro from T7 promoters to produce single‐stranded RNA. This RNA was treated with DNase I to remove template DNA and purified by precipitation in ammonium acetate and ethanol. Before quantitative polymerase chain reaction assays, the RNA concentration was measured in triplicate and the molar quantity was calculated based on the size of the RNA. Dilution series were then prepared fresh for each plate at concentrations of 10⁴, 10⁵, 10⁷, and 10⁹ RNA molecules per microliter to serve as a standard curve (Pfaffl, 2004). Welch's t test was used to determine the significance of gene expression knockdown in comparison to nonspecific control treatments using dsRNA with the exogenous sequence of GFP. Knockdown was significant and ranged between 15% and 88% (Table 1).

Table 1.

Summary Results for O. fasciatus RNA Interference

Stage	Target	Scored	Sex	Phenotypic penetrance		Knockdown
4th instar	GFP	18	n.d.		(0)
		14	♂		(0)
		20	♀		(0)
	ix	106	n.d.	74%	(79)		n.d.
		6	♂	100%	(6)	78.4%	±2.6%	*
		16	♀	100%	(14)	90.9%	±0.8%	*
	fl(2)d	32	♂		(0)	49.2%	±16.1%
		27	♀		(0)	71.3%	±5.3%	*
	vir region 1	13	♂		(0)	27.1%	±48.9%
		25	♀		(0)	15.1%	±3.9%	*
	vir region 2	20	♂		(0)	88.3%	±3.4%	*
		9	♀		(0)	68.8%	±3.5%	*
	vir region 3	16	♂		(0)	72.5%	±3.4%	*
		25	♀		(0)	62.0%	±8.3%	*
	tra‐2a region 1	22	♂		(0)	20.6%	±3.4%	*
		49	♀		(0)	17.7%	±2.7%	*
	tra‐2a region 2	19	n.d.		(0)		n.d.
	tra‐2a‐like	6	♂		(0)		n.d.
		16	♀		(0)		n.d.
	tra‐2b region 1	14	♂		(0)	86.0%	±5.8%	*
		34	♀		(0)	60.7%	±2.7%	*
	tra2b region 2	7	♂		(0)		n.d.
		6	♀		(0)		n.d.
	tra‐2a, tra‐2b	5	♂		(0)		n.d.
		15	♀		(0)		n.d.
	msl‐1	2	♂		(0)		n.d.
		8	♀		(0)		n.d.
	msl‐2	17	♂		(0)	39.3%	±0.9%	*
		35	♀		(0)	80.3%	±25.5%	*
	msl‐3	6	♂		(0)	61.8%	±2.5%	*
		9	♀		(0)	58.5%	±1.1%	*
maternal	ix	19	♀	100%	(19)		n.d.
	tra‐2a	4	♀		(0)		n.d.
	tra‐2b	11	♀		(0)		n.d.

Note: Knockdown was quantified by quantitative polymerase chain reaction in three biological replicates of each sex. Standard errors and significance of knockdown are given for these samples. Significance was determined using Welch's t test compared to nonspecific GFP dsRNA controls; *p < 0.05. n.d., not determined.

Abbreviations: dsRNA, double‐strand RNA; GFP, green fluorescent protein.

3. RESULTS

3.1. A majority of D. melanogaster sex determination genes are conserved in other insects

We selected 29 genes from the sex determination and dosage compensation GN of Drosophila melanogaster (Table S1) and screened the genomes of other insects to identity putative orthologs. Candidate sex determination genes included Sxl and its regulators encoded by sisterless‐a, scute, upd, runt, deadpan, snf and spf45. We included tra, its regulators fl(2)d and vir, tra2, dsx, its transcriptional cofactors encoded by her and ix, and the transcription factor encoded by fru. We also included chinmo and nclb, transcription factors required for identity of the male somatic gonad and germline, respectively, as well as ovo, otu and stil, which are required upstream of Sxl in the female germline. Surveying the available genomic data in GenBank, most insect taxa have unambiguous single orthologs of these genes (Figure 2). However, several exceptions exist and their patterns are instructive. First, there are some insect taxa, including Zygentoma, Orthoptera, Gerroidea, Reduvioidea, Miroidea and Platypezoidea where numerous genes could not be detected. This result likely reflects limitations of the available genome sequences. The absence of orthologs across more well‐sampled clades can be more definitive. Two X‐linked activators of D. melanogaster Sxl, sis‐a and upd, only have orthologs in the Schizophora or Schizophora + Sciaroidea, clades with a relatively recent radiation (Wiegmann et al., 2011). Other regulators of Sxl have orthologs in most sampled taxa. Similarly, orthologs of the female germline determinant stil appear limited to Schizophora. The gene her is only found in some species of Drosophila, in the clade containing the melanogaster and obscura species groups. Other sex determination genes are represented by a single ortholog in most well‐sampled taxa, although gene duplications appear in a few lineages. Sxl has several paralogs in noninsect arthropods, but the only insect genome with an apparent duplicate is Drosophila melanogaster. fru paralogs appear in all butterfly genomes (superfamily Papilionoidea) and in the staphylinoid beetle Nicrophorus vespilloides. Besides their high degree of sequence divergence, tra orthologs also seem prone to duplication, with multiple copies in Apis (where they are known as fem and csd), the bumble bee Bombus terrestris, several ant species (Privman et al., 2013), and O. fasciatus (Figure S1).

3.2. Low structural conservation of tra

Identification of tra orthologs has been challenging (Figure S1). Protein sequence alignment of the available tra orthologs has among the lowest sequence identity in this study: 16.1% (Figure S4). Schizophora includes unambiguous orthologs of D. melanogaster tra (Figure 2 and Figure S1). Functional studies of putative tra orthologs in beetles and in the cockroach support claims of orthology from these diverse insect groups, initially based on modest domain‐level similarity (Morita et al., 2019; Shukla & Palli, 2012; Wexler et al., 2019). Using tra query sequences from these species, we recovered potential tra orthologs from nine additional insect lineages (Figure S1), including Phasmatodea (Timema ssp.), Fulgoroidea (N. lugens), several groups of Heteroptera (e.g., O. fasciatus) all sampled Hymenopteran groups except Tenthredinoidea, and two beetle superfamilies, Elateroidea (Photinus pyralis) and Cucujoidea (Aethina tumida). However, no tra orthologs were identified from several well‐studied lineages with high‐quality genomic resources, such as the aphids, Lepidoptera or mosquitos. Sex‐specific regulation of dsx by tra in the cockroach suggests an ancestral role for tra as a regulator of somatic sex (Verhulst & van de Zande, 2010; Wexler et al., 2019). However, there is large sequence divergence of tra across insect species outside of the RNA‐binding domain and large divergence between the functional domains of tra even within Drosophila (O'Neil & Belote, 1992), indicating that while it may be functionally conserved tra evolves rapidly at the sequence level. Given the extensive conservation of other focal genes, the scant recovery of tra orthologs suggests that, compared to other genes involved in sex determination, tra may be subject to different evolutionary pressures.

3.3. Several sex determination genes are absent or duplicated in Hemiptera

Gene duplications are nonrandomly distributed among the lineages we sampled (Figure 2). The occurrence of gene duplication displays significant phylogenetic signal in all Hemipteran taxa where it was observed (Local Indicator of Phylogenetic Association, p < 0.05; Figure S1). While Sxl is conserved across other insects (Cline et al., 2010), it is absent from Heteroptera (Figure 2). Two of the genes that regulate Sxl in D. melanogaster, runt and spf45, are duplicated in psyllids, and spf45 is also duplicated in the fulgoroid Nilaparvata lugens. tra2 is duplicated in the pentatomoid Halyomorpha halys, and in several lygaeoid species, including O. fasciatus. dsx appears to have undergone duplication within Auchenorrhyncha, with paralogs in N. lugens and Homalodisca vitripennis.

3.4. Most dosage compensation genes in D. melanogaster have orthologs across the insects

The five genes surveyed from D. melanogaster dosage compensation have orthologs in most insect taxa and in noninsect arthropods (Figure 2). However, msl1 appears to be missing from several insect groups, including Siphonaptera, all Coleoptera and Hymenoptera, Thysanoptera and Orthoptera. Orthologs of msl1 are present in most Diptera. Lepidoptera, except the early‐branching Yponomeutoidea, have orthologs of msl1. However, msl1 is present in the well‐sampled superfamilies of Hemiptera, including O. fasciatus where it has been duplicated. Other hemipteran lineages have experienced duplications of dosage compensation genes, with paralogs of mle in the aphid Acyrthosiphon pisum and two copies of msl2 in the psyllid Diaphorina citri. 3.5 Comparisons of gene and organismal phylogenies

3.5. Relationships of sex determination genes exhibit lower than expected congruence with organismal relationships

While the presence or absence of genes is a simple level on which to examine potential GN conservation, sequence differences among taxa can reveal the extent of structural conservation or divergence. Molecular phylogenetic inference of organismal relationships is based on the observation that purifying selection tends to conserve the sequence of many proteins, especially those with basic functions in cell physiology. Orthologs diverge in their sequence over time, accumulating differences that may reflect patterns of species divergence. However, many genes do not follow these patterns, and phylogenies based on individual genes often do not reflect organismal relationships. Rather than examining evolutionary rates, such as dN/dS, which require an identified outgroup and are calculated pairwise, we sought to generalize gene tree comparisons that are often made subjectively to an accepted tree of organismal relationships.

Therefore, we used the disagreement of phylogenies based on single proteins with a consensus phylogeny of insect relationships (based on Johnson et al., 2018; Kawahara et al., 2019; McKenna et al., 2019; Misof et al., 2014; Peters et al., 2017; Wiegmann et al. 2011) as a metric of evolution in these genes (Figure 3). Comparisons were based on tree distance (Smith's, 2020, nCID). Phylogenies may disagree simply due to an inability to resolve taxa. As expected, we found that the portion of parsimony informative sites declined significantly with greater sequence identity (Figure S5a) and nCID was positively correlated with polytomy (nodes with less than 50% bootstrap support were collapsed; Figure S5c). Therefore, we excluded genes from consideration if they had greater than 90% sequence identity, such as beta‐actin and calmodulin. However, tree distance (nCID) was not strongly correlated with the portion of parsimony‐informative sites in the alignments (Figure S5d), suggesting that low information content alone was not responsible for disagreement in phylogenies. For each protein tree, bootstrapping was used to define a 95% confidence threshold reflecting values that might be produced by random relationships across the topology. All genes included in the analysis had nCID values below this threshold (Figure S4).

Figure 3.

Comparisons of candidate gene trees to the consensus phylogeny of insect groups. Genes are grouped by their putative function, as described in D. melanogaster. (a) Genes involved in sex determination and dosage compensation have relatively high information content, reflected by a high proportion of parsimony‐informative sites, displayed on the horizontal axis. Tree distance reflects disagreement between protein‐based phylogenies and the consensus relationships of insect taxa (Figure 1), calculated as on normalized clustering information difference (nCID; Smith 2020). Tree distance is high for many sex determination genes, particular dsx, otu and msl1, which represent downstream regulatory points in the network. (b) Normalized tree distance is significantly greater for genes associated with sex determination compared to the reference gene set (two‐tailed Wilcoxon rank sum test, W = 83, p = 5.05 × 10⁻⁷). Points represent individual genes. Tukey‐style box‐whisker plots indicate the median, quartiles and range

Genes involved in sex determination had significantly higher tree distances from the organismal phylogeny than did a reference gene set (Figure 3; Wilcoxon rank sum test, W = 83, p = 5.05 × 10⁻⁷). Individual gene trees varied widely in the degree to which they diverged from organismal topology (Figure 3b), genes such as snf, spf45, vir, and Sxl had tree distances within the interquartile range of values from the reference gene set. However, tra2, runt, fru, dpn, tra, otu, and dsx had tree distances greater than the upper extremes of the references. dsx showed the greatest deviation from the organismal tree. The other three DMRT genes had distances that were within the range of the reference gene set or just above this value. Genes implicated in dosage compensation had a similar range to other sex determination genes, with mle and mof being within the interquartile range of the reference set. msl1 was the most divergent gene in this group, with a tree distance second only to dsx.

Because the metric of tree distance (nCID) was correlated with polytomy when 50% was used as the threshold for resolving nodes (Figure S5c), we repeated the analysis using 20% and 80% as permissive and conservative thresholds, respectively. These alternative analyses produced similar results (Figures S6 and S7), in which sex determination genes showed elevated tree distances compared to the reference gene set. The relative rank of genes was also very similar to the main analysis presented above. Using the more conservative 80% bootstrap cutoff resulted in trees with more polytomy (Figure S7c), but sex determination genes still had significantly greater tree distances (Figure S7a; WRST, W = 57, p = 1.65 × 10⁻⁸).

A small number of other genes was analyzed separately (Figure S8). This group included sisa, upd, stil and Masc, which are restricted to specific insect orders. PSI, which is involved in Bombyx mori sex determination, and Mdmd/CWC22, which regulates sex in Musca domestica, were also included because of the strong likely that these conserved proteins have sex determination roles that are lineage‐specific. Estimates of nCID for sisa, upd, stil and Masc fell at or above their confidence intervals (Figure S8a), making interpretation of those results moot. Values for PSI and Mdmd/CWC22 passed this filter criterion. After adjusting for sequence identity, the tree distances of these genes appeared within the range of the reference gene set (compare Figure 3a and Figure S8b). Because the trees include orthologs from across the insects, this result reflects the genes' evolutionary history across that entire group. Although it is noteworthy that relationships among Lepidoptera in the PSI tree are mostly unresolved, and that the branch leading to M. domestica Mdmd in the CWC22 tree is twice the length of other Diptera (tree files are available at https://github.com/aphanotus/tree.comparison). These observations are consistent with increased evolutionary rates and novel evolutionary pressures following cooption of these genes into sex determination.

3.6. RNAi of candidate genes in Oncopeltus

The occurrence of orthologous genes often carries the assumption of conserved biological function (Nehrt et al., 2011; Stamboulian et al., 2020). However, this hypothesis is not always true, especially in GNs with a dynamic evolutionary history. We performed RNAi in O. fasciatus for a subset of candidate gene orthologs, including fl(2)d, vir, tra‐2, msl‐1, msl‐2, and msl‐3 to determine if these genes have conserved roles in sex determination or were required for viability during adult development, which would be consistent with a role in dosage compensation. Our lab has previously used similar methods to describe requirements for dsx, ix and fru in the development of O. fasciatus somatic sexual dimorphisms (Just et al., 2021). These experiments were conducted alongside those reported here, and data from ix RNAi are reported in Table 1 for comparison to other target genes.

Knockdown of fl(2)d, vir, two paralogs of tra‐2, msl‐1, msl‐2, or msl‐3 failed to produce phenotypes that differed from nonspecific GFP dsRNA control treatments in external genital morphology or internal gonad morphology (Table 1). Multiple regions were targeted by dsRNAs for vir, tra2‐a and tra2‐b. Simultaneous knockdown of tra2‐a and tra2‐b was also performed.

While these genes did not produce sex development defects or strong lethality, knockdown of fl(2)d produced a defect in the wing hinge in 37.5% (21 of 56) specimens (Figure 4). This defect was present in both females (Figure 4a) and males (Figure 4b). Although sexually dimorphic structures were normal in fl(2)d specimens, these bugs were never observed mating and produced no viable eggs. fl(2)d encodes a component of the WMM complex, which increases the efficiency of mRNA splicing via adenosine methylation (Lence et al., 2016), and the protein is also associated with heterochromatin in D. melanogaster (Swenson et al., 2016). In D. melanogaster xio encodes a protein that interacts with vir and fl(2)d in Drosophila to mediate Sxl mRNA splicing (Guo et al., 2018). However, in addition to female‐to‐male transformation, xio loss of function also produces a “held‐out” wing defect, suggesting pleiotropic roles for these splicing regulators. It is interesting to speculate that fl(2)d or its cofactors may have ancestral roles in wing development.

Figure 4.

Phenotypes of fl(2)d RNAi females (a) and males (b), showing nonsex‐specific defects in wing orientation

We also performed maternal RNAi for tra‐2. Adult virgin females were injected with dsRNA targeting both paralogs of tra‐2. All injected females were able to mate and each produced several clutches (~4–6) of viable offspring, totaling >100 individuals. As adults these offspring of maternal tra‐2 RNAi displayed normal internal reproductive structures and external genitalia and were able to mate and produced viable offspring of their own that also developed into adults (>100 individuals). In D. melanogaster tra‐2 is required for fertility in both sexes, in the female soma and in the male germline (Mattox et al., 1990). The ability of O. fasciatus individuals from tra‐2 RNAi treatments to display normal mating behavior and produce viable, fertile offspring suggests that tra‐2 is not involved in sex‐specific developmental processes in the milkweed bug. Since dsx transcripts appear to lack a Tra/Tra‐2 splicing site in B. mori (Suzuki et al., 2001), it is likely that this interaction evolved in the lineage leading to Drosophila sometime after its divergence from Lepidoptera.

4. DISCUSSION

4.1. Structural conservation masks mechanistic variation

Among the genes examined in this study, most have orthologs in the genomes of diverse insect groups (Figure 2). Conservation of orthologous genes suggests the potential for functional conservation, but this hypothesis is not always found to be true (Nehrt et al., 2011; Stamboulian et al., 2020), and it should be considered skeptically. While the GN effecting sex determination and dosage compensation has been well‐described in D. melanogaster, considerable differences in sex determination mechanisms are found in other insect species. The GRNs described from model species in the Lepidoptera (Kiuchi et al., 2019; Yang et al. 2021) and Hymenoptera (Hasselmann et al., 2008) present significantly different paradigms. Our results show that the O. fasciatus orthologs of several candidate sex determination genes from the D. melanogaster model are not involved in this process in the milkweed bug (Table 1). dsx, ix, and fru are required for somatic sexual dimorphisms in O. fasciatus, but their sex‐specificity and anatomical context are different from D. melanogaster and other insects (Just et al., 2021). We have also shown that genes involved in sex determination and dosage compensation have significantly different evolutionary trajectories from genes involved in other cellular processes or the species they compose (Figure 3). Thus, these findings are consistent with previous observations that sex determination mechanisms evolve rapidly (Gempe & Beye 2011; Hopkins & Kopp, 2021). However, we suggest that from these variations, certain patterns or themes can be identified. Before discussing these themes in Section 4.5 below, we first discuss other notable results and summarize theories of network evolution.

4.2. Gene duplication is frequent in Hemiptera

Gene duplication can introduce evolutionary potential because of the possibility for sub‐ and neofunctionalization of paralogs (Lynch & Conery, 2000). doublesex is an important example, as it is one of four members of the DMRT family, which radiated at the base of the insects. These genes have been functionally tested during adult development in the planthopper N. lugens (Zhuo et al., 2018) and in the milkweed bug O. fasciatus (Just et al., 2021). In both species only dsx is required for sexually dimorphic development; although the other DMRT genes exhibit sexually dimorphic gene expression, no RNAi phenotypes have been described, and their functions remain unknown. Because DMRT‐related genes are involved in sex determination in other animal groups, the duplication events that occurred early in the lineage leading to insects is an example of duplication and divergence: one copy has maintained the ancestral function in sex‐specific development, allowing the others to either obtain novel functions or to lose their functionality altogether. Other genes, including runt, spf45, tra2, mle, msl1, and msl2, have been duplicated in different hemipteran lineages. This compares to the observation of only four duplications of sex determination genes in other insect groups. While two potential tra orthologs were found in O. fasciatus (Figure S1), tra is also prone to frequent duplications in Hymenoptera (Hasselmann et al., 2008; Privman et al., 2013).

4.3. Functional conservation of genes in dosage compensation remains unclear

Based on the diversity of sex chromosome systems in insects, it appears that multiple solutions to the challenge of dosage compensation have evolved. Therefore, a reasonable prediction is that insects will exhibit different mechanisms to equalize expression of genes on sex chromosomes. However, functional data to evaluate this hypothesis are limited and mixed. Knockdown of msl‐1 in the cotton bollworm Helicoverpa armigera caused upregulation of Z‐linked gene expression in males, consistent with a conserved role for this gene in dosage compensation, despite the ZW sex chromosome system (Zhang, et al., 2019). However, similar density of MSL proteins on the X chromosome and autosomes in both sexes of the gnat Sciara ocellaris does not support a role for these proteins in dosage compensation in that clade of Diptera (Ruiz et al., 2000). Knockdown of msl‐1, msl‐2, and msl‐3 in the milkweed bug O. fasciatus did not produce developmental defects or lethality in either sex (Table 1), also suggesting that these genes, while structurally conserved, were not required for dosage compensation. Given the phylogenetic distribution of these species, it is not possible to conclude whether MSL proteins have an ancestral role in dosage compensation. However, the sequence of these proteins predicts a molecular function in chromatin modulation (Li & Dou, 2010), and this may represent a process easily recruited into dosage compensation in independent insect lineages.

4.4. The extent of network evolution in sex determination

Experimental evidence suggests that sex determination networks are flexible and can evolve via compensatory adaptation (Chandler et al., 2012). When the sex‐determining pathway is altered to bias the development of one sex, there will be a strong advantage to new alleles or standing genetic variation that restores an equal sex ratio. This evolution may occur via changes in genes not previously associated with the sex‐determination pathway. Pomiankowski et al. (2004) hypothesize that the Drosophila sex determination GRN could have evolved in this fashion, with selection favoring mutations that reduce expression of sex‐specific transcripts in the wrong sex.

Wilkins (1995) has suggested that sex determination hierarchies evolve via sex ratio selection from the bottom up, by recruiting new upstream control elements. The most downstream components of the GRN, such as dsx, are thus predicted to be the most ancient and under the most selective pressure to retain their function in the network. This kind of downstream conservation is present in other sex‐determining networks: DMRT genes in vertebrates (orthologs of insect dsx) promote male development throughout the group, while the male determinant SRY is found only within mammalian taxa (Kopp, 2012; Wallis et al., 2008).

Genomic sexual conflict is another possible cause for the speed of GN evolution in sex determination mechanisms compared to other processes (Werren & Beukeboom, 1998). In diploid sexual species, the paternal genome is predicted to prevent female development. In support of this hypothesis are the observations that in diploid insects with a male‐determining Y chromosome, such as Ceratitis capitata and Musca domestica, the male determinant actively blocks expression of tra and thus prevents the establishment of the auto‐regulatory tra loop required for female development (Hediger et al., 2010; Pane et al., 2002). Autosomal genes in D. melanogaster, which spend equal time in both sexes, prevent the Sxl expression required for female development. In Lepidoptera, where females are the heterogametic sex and possess a female‐determining factor, the role of the paternal genome in tra activity is not as clear. Pomiankowski et al. (2004) suggest how a homozygous tra genotype that promotes male development, such as the complementary sex determination system in Hymenoptera, could have evolved to eliminate this genomic conflict. Therefore, genetic models of sex determination are compatible with theoretical predictions of genomic sexual conflict. Theoretically and empirically, this influence is most relevant to points the sex determination network integrating primary signals, such as tra. This additional evolutionary pressure may help explain the extent of sequence divergence in tra despite its conserved functional role.

4.5. Themes and variations in the evolution of insect sex determination GNs

While functional tests of genes in development are still only available from a handful of species, the increasing availability of genomic data allows us to determine the presence and absence of orthologous genes from a much wider sampling of the diversity of life (Figure 2). Together functional, cytological, and genomic data provide clues to broad phylogenetic patterns of network evolution. Studies of insect sex determination and dosage compensation demonstrate that these genetic pathways vary between different insect groups. However, the patterns in these variations may be instructive. Information on sex determination and dosage compensation from diverse insects suggest themes for the evolution of this genetic network: (1) Unique features within taxa influence network evolution. (2) Position in the network influences a component's evolution.

4.5.1. Taxon‐level influences on network evolution

Rather than a consistent mechanism of sex determination, variation is itself a key theme. However, more critically several local variations are suggested by the phylogenetic distribution of orthologs (Figure 2). An obvious example is the loss of a gene, which imposes a hard constraint on its involvement in a GRN. The absence of otu from most Coleoptera and of msl1 from Coleoptera and Hymenoptera would preclude the possibilities of their involvement in germline sex determination and dosage compensation in those groups. The apparent absence of tra from many insect lineages is a challenge to the central role this gene plays in our present model of insect sex determination. Such lineages may well have novel sex determination mechanisms, and investigating these systems is likely to be illuminating. Conversely gene duplications are unevenly distributed across the insect phylogeny, and the appearance of paralogs may allow for unique innovations within those lineages. The prevalence of ZW chromosome systems and orthologs of Masc and PSI in Lepidoptera suggests that this group may have a sex determination mechanism that is well‐represented by the model species Bombyx mori. Sex determination in many Diptera is marked by the role of dominant Y‐linked male inducing factors. However, nonhomologous genes appear to perform this function in different dipteran groups. While broader sampling is required within the order, it may appear that dominant‐Y sex determination is a lineage‐specific theme, with variations on the specific male‐inducing factor.

4.5.2. Network position influences functional evolution

Waddington (1959) suggested that the landscape of genetic interactions might influence phenotypic evolution, and that mutations at some positions within that landscape might exert more phenotypic influences than others. This idea has been explored in recent decades by evo‐devo (Stern & Orgogozo, 2009), where pleiotropy is often emphasized as a factor limiting the potential evolution of a gene's function. Generally, upstream regulatory components in a GRN are assumed to be more pleiotropic, because they have greater indirect influence across the network. Therefore, upstream components have been suggested as generally more conserved. In contrast, Wilkins (1995) argued for the evolution of sex determination GRNs “upward”, meaning that upstream network components would tend to be more evolutionarily labile, because of a selective advantage to genes equalizing skewed sex ratios.

As reviewed here, different components of sex determination networks exhibit very different degrees of conservation and divergence among insect models. Neither prediction of conservation at the “top” or “bottom” of the GRN fit the observed data very well. From limited functional studies, tra and dsx appear to be the most functionally conserved components of the sex determination network. (Although this may be due to a bias to report confirmatory results in diverse species.) Instead, these genes may represent a sort of “cold spot” in the evolution of gene interactions in the middle of the GRN. It is noteworthy that tra, a key regulator of sex determination across diverse insect groups, has exceptionally low sequence conservation. This observation leads to a critical aspect of this theme that we wish to highlight.

4.5.3. Network position influences structural evolution

While tra and dsx are functionally conserved, their sequence evolution (Figure 3) is highly dynamic. In contrast, genes with roles early in the D. melanogaster sex determination GRN, such as the X‐signal elements (runt, sisA, scute, upd), the primary signal of sex (Sxl) or its regulators (vir, fl2d, spf45, snf), all appear absent outside Schizophora, either from a role in sex determination or absent entirely from the genome. However, in species that retain orthologs of these genes (Figure 2), their evolutionary history is relatively congruent with their taxonomic relationships (Figure 3). Other relatively downstream components of the D. melanogaster sex determination mechanism, ix and fru, also have roles in this process in O. fasciatus and B. mori, albeit with interesting differences. Relationships among these orthologs of these genes also have relatively high deviations from their organismal relationships (Figure 3). This pattern is counter‐intuitive: more functionally conserved genes are more divergent in their sequences.

While a gene or protein may diverge in its sequence, this may not influence its function within a network. It is possible that dsx and tra evolve rapidly in sequence due to frequent duplication (Ohno, 1970). Retaining a functioning paralog could allow another to diverge in sequence without antagonistic fitness effects from pleiotropy (Lynch & Conery, 2000). Upstream components of the GRN that seem less functionally conserved, but are more conserved in protein sequence, must evolve different connections to this or other GRNs via mutations affecting regulatory DNA sequences.

Determining a null hypothesis for GN evolution has been a challenge (Church & Extavour, 2020), as we are still in need of perspectives from a broader range of species. Further studies of fundamental developmental pathways and their differences between animals will broaden our perspective of how GNs bias evolutionary outcomes and how evolutionary processes lead to the alteration of those genetic networks. We hope that future studies of sex determination will consider whether the themes outlined here continue to apply.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

Mara Laslo and Josefine Just: designed and conducted most experiments and contributed equally to data analysis, creation of draft figures, and writing of the main text. David R. Angelini: designed and supervised all experiments, secured funding, prepared final figures, and contributed to data analysis and to the writing of the main text.

1. PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1002/jez.b.23125

Supporting information

Supporting information.

Laslo, M. , Just, J. , & Angelini, D. R. (2023). Theme and variation in the evolution of insect sex determination. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, 340, 162–181. 10.1002/jez.b.23125

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are freely available at https://github.com/aphanotus/tree.comparison.