An Unexpected Regulatory Cascade Governs a Core Function of the Drosophila PRC1 Chromatin Protein Su(z)2

Abstract

Polycomb group (PcG) proteins are major chromatin-bound factors that can read and modify chromatin states to maintain gene silencing throughout development. Here we focus on a close homolog of the PcG protein Posterior sex combs to better understand how these proteins affect regulation. This homolog, called Suppressor 2 of zeste [Su(z)2] is composed of two regions: the N-terminal homology region (HR), which serves as a hub for protein interactions, and the C-terminal region (CTR), which is believed to harbor the core activity of compacting chromatin. Here, we describe our classical genetic studies to dissect the structure of Su(z)2. Surprisingly, we found that the CTR is dispensable for viability. Furthermore, the core activity of Su(z)2 seems to reside in the HR instead of the CTR. Remarkably, our data also suggest a regulatory cascade between CTR and HR of Su(z)2, which, in turn, may help prioritize the myriad of PcG interactions that occur with the HR.

THE Polycomb group (PcG) proteins are epigenetic modifiers whose contribution to stable silencing belies the dynamic nature and diversity of their function. One set of PcG proteins forms the Polycomb Repressive Complex 1 (PRC1), which can compact chromatin in vitro into a state that is refractory to remodeling and transcription (King et al. 2002, 2005; Francis et al. 2004; Grau et al. 2011; Beh et al. 2012; Simon and Kingston 2013). PcG proteins are also believed to effect silencing via the post-translational modification of histones, such as the methylation of histone H3 on K27 as well as both the ubiquitylation and deubiquitylation of histone H2A on K118/9 (Cao et al. 2002; Wang et al. 2004; Lagarou et al. 2008; Scheuermann et al. 2010; Schwartz and Pirrotta. 2013; Grossniklaus and Paro 2014). While maintaining the silenced state of targeted genes, some PcG proteins have residency times on chromatin that last less than a single cell cycle, binding and dissociating from different loci at different rates (Ficz et al. 2005; Ren et al. 2008; Fonseca et al. 2012; Vandenbunder et al. 2014). These observations attest to the dynamic equilibria of PcG proteins and the maintenance of their different, and sometimes opposing, chromatin states.

Here we describe our use of classical genetics to better understand how the disparate functions of PcG proteins might be coordinated. In particular, we have focused on the genetic dissection of a protein domain, called the homology region (HR), which was originally defined for its conservation among the PcG Drosophila Suppressor 2 of zeste [Su(z)2] gene, its paralogous Posterior sex combs (Psc) gene, and their mammalian homologs Bmi-1 and Mel-18 (Brunk et al. 1991; van Lohuizen et al. 1991). The HR, located at the N-terminal end of Su(z)2 and Psc, is believed to house the protein–protein interaction subdomains responsible for interactions with the other subunits of the PcG complexes PRC1 and dRAF, the H2AK118/9 ubiquitylation complex (Kyba and Brock 1998; King et al. 2005; Lagarou et al. 2008; Lo et al. 2009; Schwartz and Pirrotta 2013; Simon and Kingston 2013; Kang et al. 2015; Wani et al. 2016). The C-terminal regions (CTRs) of Su(z)2 and Psc have been considered the core of these proteins as they are the main contributors of PRC1-mediated silencing; they can, on their own, recapitulate the inhibition of chromatin remodeling, compaction of chromatin, and transcriptional repression exhibited by PRC1 in vitro (King et al. 2002, 2005; Francis et al. 2004; Lo et al. 2009; Beh et al. 2012). This is consistent with the fact that, while the mammalian orthologs do not have the CTR, the function of the CTR has been retained in mammals in a stretch of positively charged residues found in the Pc subunit of the PRC1 (Grau et al. 2011). Thus, it is believed that the HR serves as the hub of protein–protein interactions which, in part, help direct the CTR to genomic regions requiring regulation by PcG proteins. While characterization of the CTR has been driven by the genetics of both Su(z)2 (Wu and Howe 1995; Emmons et al. 2009; Lo et al. 2009) and Psc (Francis et al. 2004; King et al. 2005), the Su(z)2 locus has been more amenable to the genetic dissection of the HR (Wu and Howe 1995; Emmons et al. 2009; Lo et al. 2009). Thus, we have been continuing our study of Su(z)2 to characterize and analyze the functions of HR.

One allele of Su(z)2, named Su(z)2¹, is caused by a truncation mutation that deletes nearly all of the CTR while keeping the HR intact (Wu et al. 1989; Wu and Howe 1995; Emmons et al. 2009). As expected, this allele is recessive lethal, which is in line with in vitro evidence suggesting that the CTR is essential, providing the compaction mechanism that is necessary for transcriptional silencing (Wu and Howe 1995; King et al. 2005; Lo et al. 2009). Despite lacking the CTR, however, Su(z)2¹ exhibits several gain of function phenotypes (Wu and Howe 1995; Emmons et al. 2009), suggesting that some of the functions of Su(z)2 are compartmentalized within the HR alone. Thus, Su(z)2¹ provides us with the opportunity to isolate and study the HR.

An interesting illustration of the gain-of-function nature of Su(z)2¹ is its capacity to interfere with the viability of a class of hypomorphic alleles that are hemizygous viable (Emmons et al. 2009). This is apparent with four missense mutations of the HR (Figure 1A, e26, s36, s20, and s21), which show varying levels of viability when heterozygous with a deletion of Su(z)2, but are inviable when heterozygous with Su(z)2¹ (Emmons et al. 2009). These findings suggested that Su(z)2¹ is more detrimental than a complete deletion of Su(z)2 and, therefore, may interfere with essential functions provided by the HR alleles. For example, Su(z)2¹ might interact directly with mutant HR proteins or compete with them such that they are no longer functional, perhaps by the HR of Su(z)2¹ acting as a “sink” and outcompeting mutant HR proteins for interactions with other factors (Figure 1B). This interpretation is particularly intriguing because the HR contains domains implicated in protein–protein or other types of interactions. These domains include a ring finger (RF) (residues 35–73) (Brunk et al. 1991; van Lohuizen et al. 1991), the core sequence of the conserved subregion 1 (CSR1), henceforth called the conserved patch (CP) (residues 97–103) (Buchwald et al. 2006; Emmons et al. 2009; Bentley et al. 2011), and the ubiquitin-like domain (UL) (residues 137–228) (Figure 1A) (Sanchez-Pulido et al. 2008; Bezsonova et al. 2009; Wang et al. 2010).

Figure 1.

(A) Structure of Su(z)2 and previously extant alleles (Wu and Howe 1995; Emmons et al. 2009). Su(z)2 is divided into two main regions: the HR and the CTR. The HR is composed of three domains, the RF (red), the CP (gray), and the UL domains (blue). The alleles differ from wild-type either by point mutation (○), truncation, and/or frameshift mutation (white region denotes the novel amino acid sequence generated by the frameshift). (B) The competition hypothesis between Su(z)2¹ and s36. Su(z)2¹ outcompetes s36 for essential protein interactions because the HR of s36 is compromised (○), while mutation of the HR of Su(z)2¹ (♦) restores essential interactions with s36.

Here, we describe our genetic studies clarifying how the specific domains within the HR contribute to Su(z)2 function. Specifically, we conducted a screen for mutations that would suppress the lethality of Su(z)2¹ when it is heterozygous with an HR allele called s36, which disrupts the RF (Figure 1B). Remarkably, all of the isolated lines had mutations in the RF, CP, and UL domains of Su(z)2. Furthermore, they demonstrated that contrary to expectations, the CTR may not be essential and that the HR, alone, can support viability, arguing that the core activity of Su(z)2 may actually reside in the HR. Interestingly, our data reveal a potential cascade of intramolecular regulation between different domains of Su(z)2. We discuss how our findings may explain the potential coordination of disparate PcG functions that occur through the HR interface of Su(z)2.

Materials and Methods

Drosophila stocks and culture conditions

Stocks were maintained and crosses were conducted at 25° at 80% humidity on standard Drosophila cornmeal, yeast, sugar, and agar media with p-hydroxybenzoic acid methyl ester added as a mold inhibitor. Crosses were carried out with approximately four males and four females and brooded daily to prevent crowding. All Su(z)2 alleles were maintained over CyO-19 (Flybase ID FBba0000315), a GFP-bearing balancer. The Su(z)2 deficiency (Df), generously provided by Jun-yuan Ji, was generated via transposase-mediated deletion using PBac{RB}Su(z)2[e00448] and P{XP}[d02486].

Mutagenic screen for the generation of Su(z)2¹ derivatives

Male w⁻; Su(z)2¹/w⁺, CyO-19 flies were fed EMS as previously described (Wu and Howe 1995) and crossed to w⁻/ w⁻; Su(z)2^s36 /w⁺, CyO-19 females. A total of 57,906 F₁ progeny were scored for white eyes and normal wings (Su(z)2¹*/s36), with 100 exceptional progeny isolated. These flies were backcrossed to the s36 stock to test for transmission of mutation. The 12 lines that showed transmission were then isogenized for chromosome 2.

Molecular characterization of Su(z)2¹ derivatives

Embryos homozygous for the second chromosome were collected by setting up an overnight egg lay and selecting GFP⁻ embryos. These embryos were subjected to proteinase-K digestion to release DNA (Gloor et al. 1993) to be used for sequencing (Emmons et al. 2009). Exonic regions of Su(z)2 were amplified by PCR, gel purified with the QIAQuick Gel Extraction kit (QIAGEN, Valencia, CA; no. 28704), and sequenced at the Harvard Medical School Biopolymers Facility. Primers for PCR and sequencing are listed in (Supplemental Material, Table S2) and ordered from Integrated DNA Technologies (Coralville, IA).

Inter se crosses and calculation of viability

Inter se crosses were performed by crossing the parental lines (Su(z)2 mutant allele/w⁺, CyO-19) to the Su(z)2¹ derivative lines. Percent viability was calculated assuming equal transmission of each allele and lethality for homozygosity of the CyO balancer, in which case 33% of progeny would be expected to be heterozygous for both Su(z)2 alleles. We calculated a viability score as the observed number of non-CyO progeny/total progeny, normalized to the expected frequency of 33%. We report the average viability among the replicate crosses. SPSS software (IBM; Armonk, NY) was used for one-way ANOVA calculation and post hoc Dunnett’s test to compare all crosses with the control cross with Df.

Data availability

The authors state that all data necessary for confirming the conclusions presented in the article are represented fully within the article. Strains are available upon request.

Results and Discussion

A screen to generate point mutations in domains within Su(z)2

Based on our model of competition between Su(z)2¹ and Su(z)2^s36 (Figure 1B), we predicted that reversion of the dominant negativity of Su(z)2¹ would reveal residues or domains within Su(z)2 (Figure 1B) or perhaps even within other genes that are important for Su(z)2 function. To this end, we mutagenized males carrying Su(z)2¹ with EMS and then crossed them to females carrying Su(z)2^s36 to identify mutations that would suppress the lethality of the Su(z)2¹/Su(z)2^s36 progeny. We screened a total of 57,906 progeny in three independent mutageneses and recovered 100 Su(z)2¹/Su(z)2^s36 flies. These exceptional progeny were crossed back to flies carrying Su(z)2^s36 to test for fertility and transmission of the mutation, and 12 lines survived this process. All mutations were mapped to chromosome 2, on which Su(z)2 is located, and sequencing of the exons of Su(z)2 confirmed that each line retained the original Su(z)2¹ mutation while gaining either a missense or nonsense mutation within the HR (Figure 2A). Henceforth, we will refer to these alleles collectively as “Su(z)2¹ derivatives.” In brief, our screen to suppress the lethality between Su(z)2¹ and Su(z)2^s36 recovered only mutations in the HR of Su(z)2¹, which houses the protein interaction domains of Su(z)2.

Figure 2.

(A) The structure of Su(z)2¹ derivatives. (B) The viability of progeny resulting from crosses between Su(z)2¹ derivatives, averaged between reciprocal crosses and grouped according to the affected domains (RF, RF–CP, and UL) and either s36, e26, s20, or s21. Average viabilities of individual derivatives are provided in Table S1. Viability of Df and Su(z)2¹ are also shown. The R93Q allele (^†) and the R228GΔ allele (^‡) show lethality when paternally contributed and crossed with e26 and s21, respectively (see Table S1).

Point mutations define functional domains of Su(z)2

Two mutations, D49N and I60V, affect single residues within the RF of Su(z)2¹ (Figure 2A). As these derivatives are viable with s36 (66–87%, Figure 2B, see Materials and Methods for calculation of viability), they demonstrate that disruption of the RF is sufficient to revert the dominant negativity of Su(z)2¹. In addition, three mutations were found downstream of the RF and close to the CP, wherein two are missense mutations (R93Q and I96N) and one is a nonsense mutation that truncates the protein just before the CP and UL domains (R93*) (Figure 2A). The remaining seven mutations affect the UL, of which four are missense and three are nonsense: Q105*¹, Q105*², S141F, Q151*, C165Y, C214Y, and R228GΔ (Figure 2A); although Q105*¹ and Q105*² contain the same Q105* change and are phenotypically similar to each other, they are considered distinct alleles, as they do not behave identically and may therefore be associated with alterations in the unsequenced regions of Su(z)2 or changes elsewhere in the genome. Note that the recovery of mutations not only in the RF domain but also within the CP and UL domains suggest that these latter two domains may indirectly impact or cooperate with the functions of the RF. Indeed, the RF and the CP have been shown to engage in the same interaction with the PcG protein Ring1b in vitro (Buchwald et al. 2006; Bentley et al. 2011).

Suppression of dominant negativity is not allele specific

To determine whether the suppression of dominant negativity by the Su(z)2¹ derivatives is general or, instead, limited to the specific amino acid change in the RF of the s36 mutation, we crossed the derivatives to another RF mutation, called e26 (Figure 1A), which disrupts a conserved histidine and is also completely lethal when heterozygous with Su(z)2¹ (Figure 2B). Like s36, e26 is also hemizygous viable, although its viability is significantly less than that of s36 (52% vs. 93%, P < 0.01, Student’s t-test, respectively, Figure 2B). Ten of the 12 derivatives are viable when heterozygous with e26, confirming that the capacity of the derivatives to suppress the dominant negativity of Su(z)2¹ is not allele specific with respect to s36. The inviability of two derivatives (D49N and I60V) when heterozygous with e26 was not surprising, given the phenotypic severity of e26 (Figure 2, A and B). Interestingly, the viabilities of the derivatives when heterozygous with e26 was more variable than that observed with s36 (0–103% for e26 vs. 66–105% for s36, Figure 2B and Table S1), revealing potential functional differences between the Su(z)2¹ derivatives. Overall, these data show that the derivatives generally suppress the dominant negativity of Su(z)2¹ with one other RF mutation.

We further assessed whether the dominant negativity of Su(z)2¹ could also be suppressed when the derivatives are in trans to mutations other than those in the RF by crossing the derivatives to the s20 and s21 alleles, which are located within or near the CP and UL domains, respectively, and show strong or complete lethality with Su(z)2¹ (Figure 2B) (Emmons et al. 2009). Only one of the derivatives, UL mutation R228GΔ, maintained lethality with s21, and only when the s21 allele was contributed maternally (Table S1). This suggests a possible involvement of maternal effects, although we have yet to investigate this further. The remaining 11 of the 12 Su(z)2¹ derivatives were robustly viable when heterozygous to either s20 or s21 (Figure 2), further arguing the significance of their impact on Su(z)2.

Zygotic contribution of the CTR is not essential

Over the course of our studies, we made some unexpected observations of the CTR, leading us to question the presumed essential nature of the CTR. Here, we took advantage of two alleles of Su(z)2, called s84 and h29 (Figure 1A), both of which harbor nonsense mutations that remove the CTR as well as the UL (Emmons et al. 2009) and are lethal when hemizygous (Figure 3A). As these alleles are extreme hypomorphs or nulls, we expected that heterozygosity of either with any allele also lacking the CTR would not be viable. Surprisingly, however, two of the Su(z)2¹ derivatives, C165Y or R228GΔ, were viable when heterozygous with s84 or h29. While the viability ranged from as low as 4–55%, it was observed across all our replicate crosses (Figure 3A). These observations suggest that the CTR is not necessary for viability when there are compensatory mutations within the UL domain. An alternative explanation for the viability we observed suggests read through of the nonsense mutations of these alleles. However, h29 results from a frameshift mutation wherein read through would only produce novel sequence followed by additional nonsense mutations, and the sequence signatures of the nonsense mutations of s84, C165Y, and R228GΔ (stop codon identity and the downstream base) are the least associated with read through (Chao et al. 2003; Jungreis et al. 2011). Of the 10 remaining Su(z)2¹ derivatives, 4 were inconsistently and weakly viable, and 6 were completely lethal when heterozygous with s84 or h29 (Figure 3A and Table S1). To distinguish the two exceptional alleles, C165Y or R228GΔ, from the other derivatives of Su(z)2¹, we have placed them into their own category, which we call class I.

Figure 3.

(A) The range of viabilities of the different classes of Su(z)2¹ derivatives when heterozygous with s84, h29, or e26. The crosses with e26 are separated based on the allele coming from the mother (♀) or father (♂). Average viabilities of individual derivatives are found in Table S1. The green and red shading highlights the viabilities that were used to help classify the Su(z)2¹ derivatives. While all Su(z)2¹ derivatives are hemizygous inviable, class I alleles are defined by their strong viability with s84 and h29, and class II and III alleles are distinguished by their different viabilities with e26. (B) A model proposing a cascade of regulation. The HR is proposed to harbor a CTR-independent function that requires the RF and CP (arrow 1), but is suppressed by the UL domain (arrow 2). Suppression of the UL domain by the CTR (arrow 3) allows the CTR-independent function to persist.

Our novel findings reveal a cascade of regulation between the domains of Su(z)2 that govern its core function (Figure 3B). In our model, viability is dependent on the RF and CP domains, as observed with s84, h29, and the class I derivatives (Figure 3B, arrow 1). An intact UL domain, such as the one in Su(z)2¹, may suppress the RF and CP domains (Figure 3B, arrow 2), unless the UL domain is, in turn, suppressed by the CTR (Figure 3B, arrow 3) or mutations within the UL itself, as exhibited by the class I derivatives. Thus, this model proposes that, rather than the CTR encoding the core function of Su(z)2, the CTR acts to release the core function which is encoded, instead, in the HR. Note, not all UL mutations display CTR-independent viability, possibly because they destabilize Su(z)2 or do not fully disrupt the suppression by the UL domain. In this light, class I mutations may be highlighting new subdomains within the UL responsible for this suppression (Figure 3B).

It is possible that our observations of CTR-independent viability reflect rescue by the Su(z)2 homolog, Psc, and/or a maternal effect by the heterozygous mothers carrying a wild-type Su(z)2. This interpretation must be tempered, however, by the fact that neither Psc nor a maternally contributed Su(z)2 function are sufficient for viability (Figure 3A) (Wu and Howe 1995; Emmons et al. 2009). Furthermore, we have evidence suggesting that the maternal contribution of the CTR of Su(z)2 is not essential, as females that are heterozygous for h29 and the class I derivative R228GΔ can produce progeny; females that are genotypically null for the CTR of Su(z)2 were crossed to males from either parental strain, and while the yield of progeny was low (<15 progeny per cross, 66 progeny total scored), we obtained viable progeny, suggesting that maternal contribution of the CTR is not essential. Interestingly, of the 64 progeny scored, 2 were also genotypically null for the CTR, suggesting that both zygotic and maternal contributions of the CTR may be dispensable for viability. In summary, our data show that at the very least, neither maternal contribution nor somatic contribution of the CTR of Su(z)2 are essential for viability.

Release of an essential function by mutations in the HR

The viability of the class I derivatives in spite of their lack of a CTR prompted us to determine whether these derivatives can contribute to viability in other contexts. To this end, we compared them to a simple deficiency (Df) of Su(z)2. In particular, we compared their viability when heterozygous with e26 (class I/e26) to that of Df/e26 (Figure 3A, gray shading) noting separately the viabilities associated with either maternal or paternal inheritance of the e26 allele. While the Su(z)2¹ derivatives are hemizygous inviable, the viability of class I/e26 is similar to or significantly greater than that of Df/e26 (Figure 3A). C165Y shows a slight, though significant, increase in viability when paternally inherited (82% vs. 46%, P = 0.017), whereas R228GΔ shows a significant increase in viability when inherited either maternally or paternally (103% vs. 58%, P < 0.001, and 103% vs. 46%, P < 0.001, respectively). These results further support the ability of class I alleles to contribute to viability through specific mutations in the UL domain.

Interestingly, the three Su(z)2¹ derivatives that delete much if not all of the UL domain, but retain full RF and CP domains, can result in increased viability when heterozygous with e26, and we have categorized them into a new group called class II. Specifically, class II/e26 is more viable than Df/e26 when e26 is maternally inherited (Figure 3A, 79–105% vs. 46%, respectively, 0.001 < P < 0.04, one-way ANOVA). Note, however, that the increased viability is not seen in the reverse direction, that is, when e26 is paternally inherited (Table S1, 46–68% vs. 58%, P > 0.37). We do not have an explanation for this observation, but maternal or paternal effects may play a role. Regardless, these data support the prediction of the cascade model that disruption of the UL domain, either by class I or class II mutations, can uncover a function that supports viability in the absence of the CTR (Figure 3B).

Finally, the remaining seven alleles, which are unable to increase viability when heterozygous with e26, were placed into a final group called class III. This class includes the two UL point mutations, S141F and C214Y, which, compared to the Df, are significantly less viable when heterozygous to e26 (2–7% vs. 46–58%, P ≤ 0.006). It also includes all six RF and CP mutations; five of the alleles are significantly less viable than Df/e26 in all crosses (0–5% vs. 46–58%, P < 0.001), and I96N/e26 is significantly less viable than Df/e26 only when e26 is paternally inherited (20% vs. 58%, P < 0.001) (Figure 3A). One possible interpretation of the data are that these class III alleles retain some residual dominant negativity of Su(z)2¹ that is responsible for the lack of viability.

Overall, the marked differences in viability between the different classes with e26 aligns with our proposed model of regulation between the Su(z)2 domains (Figure 3B). For example, the class III mutations that disrupt the RF or CP domains and preserve the UL domain fail to increase viability with e26, consistent with the prediction that the RF and CP domains house a core function of Su(z)2 that may be suppressed by the UL domain. Conversely, loss of the UL domain would eliminate the suppression and release this core function, as reflected in the results of class II truncation alleles. Our data also suggest that disruption of the UL domain by point mutation may have the same effect, as the switch from residual dominant negativity of the class III UL mutations to the increase in viability by the class I mutations suggest that only certain point mutations can disrupt the UL domain sufficiently to release the function of the RF and CP domains. Overall, the crosses with e26 have allowed us to dissect the functional relationship of the different domains of Su(z)2, and these results are consistent with a model of intramolecular regulation between the domains.

Potential mechanisms of Su(z)2 regulation

In summary, our studies suggest that the core function of Su(z)2 is encoded in the HR, and the role of the CTR may be to help release this function. This interpretation is based on our surprising discovery that, in some genetic backgrounds, the CTR, which had been thought to encode the core function for Polycomb-based silencing, is actually dispensable. Specifically, the lethality seen with loss of the CTR can be rescued by mutations within the HR of Su(z)2. Thus, this study presents a new perspective on the relationship between the CTR and HR. In particular, CTR may function primarily to regulate the activities of the HR and, thus, it may be the HR that facilitates stable silencing by PcG proteins. This model would explain how our mutations within the UL domain can phenocopy a potential role of the CTR and permit CTR-independent viability.

There are multiple, but not mutually exclusive, scenarios that reflect the regulatory cascade between the CTR and the HR (Figure 4). The CTR may be subject to post-translational modifications, as seen by the phosphorylation of the CTR of Su(z)2 (Bodenmiller et al. 2008), which may alter the overall function and/or stability of Su(z)2 (Figure 4A). The CTR may also interact with the HR directly (Figure 4B), or engage in its own interactions that may inhibit, modulate, or participate with those that occur with the HR (Figure 4C). Indeed, the CTR of Psc is necessary and sufficient to bind cyclin B as part of a complex that also requires the HR (Mohd-Sarip et al. 2012). Alternatively, the propensity of the CTR and other proteins, such as M33/Cbx2, to bring nucleosomes together in vitro (Francis et al. 2004; Lo and Francis 2010; Grau et al. 2011) may actually reflect a novel function in vivo, such as a sampling and “reading” of the chromatin landscape and/or the associated chromatin-bound factors, providing an opportunity for the local regulation and prioritization of the varied PcG interactions that occur with the HR (Figure 4D) (Gao et al. 2012; Junco et al. 2013; Kassis and Brown 2013; Morey et al. 2013). Overall, our genetic analysis reveals a novel interaction between the CTR and HR and further study should elucidate the impact of this intramolecular regulation of Su(z)2 on PcG function.

Figure 4.

Potential mechanisms for the observed regulation of the HR by the CTR. (A) Post-translational modifications (green diamonds) lead to suppression of the UL domain. (B) Intramolecular inhibition of the UL domain by the CTR. (C) Protein X binds the UL and inhibits the RF and CP, whereas protein Y binds the CTR and the UL, allowing the RF and CP to engage in other interactions. (D) Chromatin compaction causes inhibition of the UL domain, allowing the RF and CP to perform its activities.

Our findings may also explain the absence of a CTR among the mammalian homologs of Su(z)2, Bmi-1, and Mel-18 (Brunk et al. 1991; van Lohuizen et al. 1991; Grau et al. 2011; Beh et al. 2012), as it is possible that mammalian HRs have evolved in such a way as to compensate for the absence of a CTR. On the other hand, it may be that the CTR function of chromatin compaction, first discovered with Drosophila proteins in vitro (Francis et al. 2004), still exists in mammals but is found within a different PcG member, M33/Cbx2 (Grau et al. 2011). Interestingly, this protein interacts with Bmi-1 and Mel-18 (Gao et al. 2012), preserving an evolutionary link in trans between compaction and the HR. Thus, our observation of CTR-independent viability may reflect a core function of the HR of Su(z)2 that is evolutionarily conserved and intimately associated with protein interactions.

Supplementary Material

Supplemental material is available online at www.genetics.org/lookup/suppl/doi:10.1534/genetics.116.187849/-/DC1.