Primary Sex Determination in Drosophila melanogaster Does Not Rely on the Male-Specific Lethal Complex

Abstract

It has been proposed that the Male Specific Lethal (MSL) complex is active in Drosophila melanogaster embryos of both sexes prior to the maternal-to-zygotic transition. Elevated gene expression from the two X chromosomes of female embryos is proposed to facilitate the stable establishment of Sex-lethal (Sxl) expression, which determines sex and represses further activity of the MSL complex, leaving it active only in males. Important supporting data included female-lethal genetic interactions between the seven msl genes and either Sxl or scute and sisterlessA, two of the X-signal elements (XSE) that regulate early Sxl expression. Here I report contrary findings that there are no female-lethal genetic interactions between the msl genes and Sxl or its XSE regulators. Fly stocks containing the msl3¹ allele were found to exhibit a maternal-effect interaction with Sxl, scute, and sisterlessA mutations, but genetic complementation experiments showed that msl3 is neither necessary nor sufficient for the female-lethal interactions, which appear to be due to an unidentified maternal regulator of Sxl. Published data cited as evidence for an early function of the MSL complex in females, including a maternal effect of msl2, have been reevaluated and found not to support a maternal, or other effect, of the MSL complex in sex determination. These findings suggest that the MSL complex is not involved in primary sex determination or in X chromosome dosage compensation prior to the maternal-to-zygotic transition.

ONE of the first developmental decisions confronting a Drosophila embryo is to determine its chromosomal sex. The urgency of the decision likely reflects the need to ensure that the genic imbalances that result from having large heteromorphic X and Y sex chromosomes are compensated before they can exact deleterious effects (Vicoso and Bachtrog 2009; Conrad and Akhtar 2011; Stenberg and Larsson 2011). Genic imbalances in several X-encoded genes, known as the X-signal elements (XSEs), actually serve as the signals of chromosomal sex (Cline 1988; Erickson and Quintero 2007). The XSE products function prior to the large-scale onset of zygotic gene expression to initiate the sex determination program. Quickly engaged, the sex determination program then locks in place a mechanism that helps ensure balanced gene expression before development can be adversely affected (reviewed in Cline and Meyer 1996; Salz and Erickson 2010; Salz 2011).

The connection between sex and balanced genomic expression is the master regulatory gene Sex-lethal (Sxl) (Cline 1984). Male embryos, with one X chromosome and one set of XSEs per cell, never express Sxl protein and produce all the protein and RNA components of the Male Specific Lethal (MSL) complex (Cline and Meyer 1996; Salz and Erickson 2010; Salz 2011). The MSL complex then assembles and binds to the male X chromosome. There, acting in conjunction with more general chromosome buffering effects, the MSL complex helps adjust the relative expression of the male X so that it is in balance with expression from the autosomes (Conrad and Akhtar 2011; Stenberg and Larsson 2011; Sun et al. 2013a; Sun et al. 2013b; Figueiredo et al. 2014; Lucchesi and Kuroda 2015). In contrast, female embryos with two Xs and two sets of XSEs per cell transiently activate the Sxl establishment promoter, SxlPe, producing a brief pulse of Sxl protein (Keyes et al. 1992; Erickson and Quintero 2007). The Pe-derived Sxl protein directs a rapid transition to a stable maintenance mode of Sxl expression that depends on positive autoregulation of the splicing of the premessenger RNA (pre-mRNA) products of the sex nonspecific promoter SxlPm (Bell et al. 1991; Gonzalez et al. 2008). Once fully engaged, Sxl blocks the translation of key dosage compensation regulator msl2 (Bashaw and Baker 1995; Kelley et al. 1995; Lim and Kelley 2012). The absence of msl2 protein prevents the formation of the MSL complex and ensures that females also express their X chromosomes in concordance with the autosomes (Bashaw and Baker 1995; Kelley et al. 1995; Lyman et al. 1997).

The Sxl regulatory mechanism requires that the XSE genes be expressed differentially in early XX and XY embryos and indeed the strong XSE genes, scute (sc) and sisterlessA (sisA), encoding transcriptional activators of SxlPe, appear to be expressed in proportion to their gene dose (Cline 1988; Deshpande et al. 1995; Erickson and Quintero 2007; Lott et al. 2011). Curiously, this is not the case for every early expressed X-linked gene. RNA-sequencing (RNA-seq) analysis of staged single embryos has revealed that most X-linked genes undergo a partial dosage compensation prior to the time that the canonical MSL complex is thought to be active (Lott et al. 2011). The nature of this early zygotic dosage compensation (EZDC) is unknown, but could reflect gene-specific or more general MSL-independent “buffering” mechanisms to adjust genic balance (Stenberg et al. 2009; Lott et al. 2011; Stenberg and Larsson 2011; Philip and Stenberg 2013; Chen and Oliver 2015). Curiously, two Drosophila species with larger X chromosomes appear to lack this early dosage compensation (Lott et al. 2014).

The conventional view of sex and dosage compensation posits a temporally linear regulatory pathway in females: the XSE proteins initiate Sxl expression, autoregulatory splicing maintains Sxl activity, and Sxl protein thereafter prevents formation of the MSL complex ensuring an appropriate balance of X and autosomal expression. There is, however, an alternative model that upends this regulatory hierarchy. Based on reported genetic interactions between mutations affecting the msl complex and mutations in Sxl, and the XSE genes sisA and scute, as well as on correlated changes in gene expression, Gladstein et al. (2010) and Horabin (2012) have proposed that the MSL complex acts in early embryos of both sexes to generally elevate X-linked gene expression. With respect to sex determination, the alternative model posits that the MSL complex is part of an X-signal amplification mechanism that increases the fidelity of Sxl activation by elevating the concentrations of XSE proteins, both in absolute terms, and relative to maternal-supplied and autosomally expressed inhibitors. Likewise, the MSL complex facilitates the transition to stable splicing of SxlPm transcripts by doubling the amount of early Sxl protein produced in females and by doubling the levels of its SxlPm-derived pre-mRNA substrates (Gladstein et al. 2010; Horabin 2012).

The early embryonic MSL complex was also proposed to explain much of the EZDC activity observed by Lott et al. (2011). Initially, X-linked gene expression is elevated in both sexes but as Sxl protein is produced, it downregulates the production of msl2 protein reducing the relative activity of the MSL complex in females. The net result is partial dosage compensation of X-linked genes in the period before msl2 is fully repressed by Sxl protein (Horabin 2012).

Despite the appeal of a model that interconnects the beginning and ends of a regulatory pathway, and that offers a unified explanation for sex signal amplification and early zygotic dosage compensation, there are reasons to be skeptical of an early embryonic MSL function. One reason for concern is that the early embryonic MSL model has paradoxical and counterintuitive implications for the regulation of sex determination. The model postulates that the MSL complex normally elevates expression of the XSEs twofold, effectively providing females with a 4X and males a 2X concentration of XSE proteins (Gladstein et al. 2010). If this is so, how can it be that msl mutant females activate SxlPe, and wild-type males do not, when both should have equivalent 2X levels of XSE proteins? The proposal that Sxl downregulates msl2 and the MSL complex in females before autoregulatory Sxl splicing is fully established (Horabin 2012) implies that the Sxl regulatory scheme is programmed to decrease its ability to discriminate between the male and female signals before that task is actually accomplished. This seems odd for a high-fidelity regulatory system that otherwise seems constructed to continually amplify the differences between males and females (Gonzalez et al. 2008). A second reason for caution is that the proposed early embryonic function of the MSL complex depends on a reported maternal effect of msl2 (Gladstein et al. 2010; Horabin 2012). There is, however, no evidence suggesting that MSL2 could be deposited in eggs, nor do precisely staged single embryos contain maternally deposited msl2 mRNA (Bashaw and Baker 1995; Kelley et al. 1995; Rastelli et al. 1995; Franke et al. 1996; McDowell et al. 1996; Lott et al. 2011, 2014). A third reason for doubt is that the genetic evidence used to support an early embryonic role of the MSL complex is unpersuasive. Most importantly, the critical genetic experiments presented by Gladstein et al. (2010) relied on nonstandard controls and on comparisons with a reference strain that may have been particularly insensitive to the biological effects measured (Cline 1988).

For these reasons, I reexamined the proposed functions of the msl genes in Sxl regulation. My data and conclusions contradict those of Gladstein et al. (2010) and Horabin (2012). I found that none of the msl genes tested exhibited the genetic interactions previously reported to occur with components of the sex determination system. I discovered that the genetic interactions previously ascribed to msl3¹ were due to another mutation in the strain, and that msl3 mutations were neither required nor sufficient, for the strong female-lethal genetic interactions seen. My findings imply that the MSL complex does not participate in Drosophila primary sex determination. They also suggest that the MSL complex is unlikely to underlie the little understood early zygotic X dosage compensation process (Lott et al. 2011).

Materials and Methods

Experiments were done at 25°, except for those with sc^3-1, which were done at 29°. Flies were raised in uncrowded conditions on a standard yeast, sucrose, cornmeal, and molasses medium. Eclosion was the criterion for viability. Mutations and chromosomes not referenced in the text are described at FlyBase. MSL complex alleles were obtained from the Bloomington Drosophila Stock Center: mle⁹, msl1^KmB, msl1^γ216, msl2²²⁷, msl2^KmA, msl3¹, Df(3L)Exel6110, and Df(3L)BSC224; from James Birchler (University of Missouri): msl2^P17, msl2^P22, msl3¹, and msl3^MAK-1; from Mitzi Kuroda (Harvard University): msl1^L60, msl3¹, and P(msl3⁺-tap); and from Victoria Meller (Wayne State University): mle¹, msl1¹, msl1^γ269, msl1^L60, msl^undefined, msl2¹, msl3¹, and rox-1^Ex6. All msl alleles employed are thought to be null alleles or strong hypomorphs: mle⁹ (internal deletion); mle¹ (nonsense codon after 125 aa); msl1^γ269 (large internal deletion); msl1^L60 (deletion of most coding sequences); msl1^KmB (amorph or strong hypomorph); msl1^γ216 (deletion of 3′ end of gene); msl2¹ (amino acid substitution, amorph, or hypomorph); msl2²²⁷ (early frameshift); msl2^KmA (amorph or hypomorph); msl2^P17 (internal deletion/inversion); msl2^P22 (internal deletion/inversion); msl2^undefined (unknown, fully male lethal); msl3¹ (amorph or strong hypomorph); msl3^MAK-1 (amorph or strong hypomorph); see FlyBase for details of sequence changes or genetic evidence of function. The sc⁷¹ and sc⁹⁶ mutations are loss-of-function alleles with bristle phenotypes and reduced male viability that were recovered in a screen used to identify sisA null mutations (Walker et al. 2000). I used the null allele Sxl^f1 in the experiments detailed here instead of the Sxl^fP7BO deletion allele employed by Gladstein et al. (2010). Control experiments showed that both null Sxl alleles produced equivalent results in experiments with msl1, msl2, and msl3 mutations (data not shown). Strains not carried by the Bloomington Drosophila Stock Center are available upon request.

Results and Discussion

The experiments described here rely on well-defined genetic interactions among components of the primary sex determination signal and their molecular target, Sxl (Cline 1984, 1988, 1993). The key principle is that the dose-sensitive nature of the X-chromosome counting process means that there is an expectation that any two mutations affecting X counting will exhibit sex-specific lethal interactions between them. Females individually heterozygous for the X-signal elements sisterlessA (sisA), or scute (sc), or for Sxl, are fully viable, but are sensitized to the loss of another X-signal component. Females carrying a second mutation, sc sisA /++ or sisA Sxl/++, for example, suffer reduced viability because of reduced expression of SxlPe and a decreased ability to transition to stable autoregulatory Sxl pre-mRNA splicing. Sxl, scute, and sisA mutations also show female-lethal interactions with maternal effect loci, such as daughterless, that are defective for maternally supplied factors needed to control Sxl expression. Most experiments in this paper address female-lethal interactions that have been reported to occur between maternal components of the MSL complex and scute, sisA, or Sxl (Gladstein et al. 2010). My general strategy was to cross msl mutant or control mothers with males carrying an X chromosome with Sxl or XSE mutations and to compare the viability of the experimental daughters with their appropriate male or female siblings. While the genetic interactions between Sxl, scute, and sisA are well defined, the magnitudes of the lethal effects observed can vary dramatically depending on genetic background (Cline 1988). For this reason, whenever possible, viability comparisons were made between siblings or between the offspring of sibling mothers.

Neither mle, nor msl1, nor msl2 interact with Sxl or XSE mutations

The data most supportive of a role of the MSL complex in primary sex determination are the apparent female-lethal synergisms between the null Sxl^fP7BO allele and maternal msl mutations reported in figure 1A of Gladstein et al. (2010). The strong female-lethal effects reported for Sxl^fP7BO/+ progeny of msl3¹ (1.4% survival), msl2¹ (31% survival), msl1^L60 (23% survival), and even the weak effect of maleless¹ (mle¹) mothers (76% survival), under conditions where all female progeny would be expected to survive, hint that components of the MSL complex may be needed to ensure that Sxl is activated in females (Gladstein et al. 2010).

Because of the central importance of these genetic interactions, I reexamined the female-lethal synergisms reported between maternal msl mutations and Sxl null alleles. I also added tests for interactions with the X-signal elements, sisA and scute, which should exhibit female-lethal interactions if the MSL complex is involved in the establishment phase of sex determination (Cline 1988; Gladstein et al. 2010). I observed a strong female-lethal interaction between maternal msl3¹ and all three zygotic sex determination mutations (Table 1, line 4), as reported previously (figures 1A and 2A of Gladstein et al. 2010). In contrast, I found that neither mle⁹, nor msl1^KmB, nor msl2²²⁷ mutants showed any indication of lethal genetic interactions with Sxl or the sex signal components (Table 1, lines 1–3). Because the mle, msl1, and msl2 alleles in Table 1 are different from those used in the earlier study, I tested for lethal interactions using additional mle, msl1, and msl2 alleles, including the three used by Gladstein et al. (2010). In every case, I found that female viability was unaffected by the maternal msl genotype (Supporting Information, Table S1).

Table 1. Test for lethal interactions between maternal msl mutations and individual sex determination mutations.

	Viability (%) of females of indicated X chromosome genotype compared to sibling males (no. reference)
Maternal 2nd or 3rd chromosome genotype	Sxlf1/ +		sisA1/ +		scM6/ +
mle9	99	(99)	97	(87)	123	(100)
msl1KmB	107	(122)	99	(80)	91	(67)
msl2227	101	(147)	95	(94)	115	(80)
msl31	15	(88)	26	(95)	12.5	(183)

Homozygous msl females generated from stocks of genotypes: mle⁹ cn¹ bw¹/CyO; msl1^KmB/CyO; msl2²²⁷ bw¹/CyO; or msl3¹ red/TM3, Sb¹ Ser¹ were crossed to y w cm¹ Sxl^f1 ct⁶/Y; y cm¹ ct⁶ sisA¹/Y; or sc^M6 w/Dp(1;Y)y²61l (sc⁺) males at 25°.

mle, msl1, and msl2 do not enhance the female lethality of sc sisA heterozygotes

A possible explanation for the discrepancies between my findings and those of Gladstein et al. (2010) would be if some undefined aspect of the genetic background or experimental conditions rendered the fly lines I used less sensitive to the lethal synergism between maternal msl mutations and females heterozygous for a single sex determination mutation. In other words, because females heterozygous for Sxl^f1, sisA¹, or sc^M6 mutations are fully viable, perhaps the demand for a lethal synergism with maternal msl alleles was too stringent a criterion. To test this possibility, I carried out sensitized crosses in which female viability was partially compromised by virtue of being heterozygous for both the sc^3-1 and sisA¹ alleles. Two control crosses (A and B of Table 2) illustrate an important point about the lethal interactions between the two XSE mutations—viability is sensitive to undefined aspects of genetic background (see Cline 1988). Here that difference is manifest as a threefold difference in viability between the sc^3-1 sisA¹/++ progeny of two lab stocks, yellow white (9.5% survival) and yellow white singed (33% survival), that would naively be expected to produce equivalent outcomes. In table 2 of his seminal paper defining the sex determination signal, Cline (1988) documented the magnitude of this variation on the related female-lethal synergism between Sxl^f1 and sisA¹. Depending on the wild-type stock used, viability of Sxl^f1 sisA¹/++ daughters ranged from <1% up to 79%. Because of the potential for undefined aspects of genetic background to influence results, each sensitized test performed here compared the viability of the progeny of sibling mothers either homozygous or heterozygous for the msl alleles.

Table 2. mle, msl1, and msl2 show no female-lethal interactions with XSE mutations in sensitized crosses.

		Progeny recovered from crosses with sc3-1 sisA1/Y males
		Experimental daughters sc3-1 sisA1/+ +			Control sons sc+ sisA+/Y
Cross	Maternal MSL genotype	Zygotic MSL genotype	Viability (%) relative to sibling males	No.	No.
A	+ (y w control)	+/+	9.5	12	126
B	+ (y w sn control)	+/+	33	28	84
C	mle9	mle9/+	49	38	77
D	mle9/CyO	mle9/+	17	16	93
		CyO/+	11	10	89
E	mle1	mle1/+	22*	32	147
F	mle1/In(2L)Gla	mle1/+	25	33	132
		In(2LR)Gla/+	34	41	122
G	msl1L60	msl1L60/+	12	21	171
H	msl1L60/ In(2L)Gla	msl1L60/+	12	7	57
		In(2LR)Gla/+	12	6	51
I	msl1KmB	msl1KmB/+	11	7	63
J	msl1KmB/ In(2L)Gla	msl1KmB/+	10	13	125
		In(2LR)Gla/+	2.4	3	122
K	msl2227	msl2227/+	50	92	183
L	msl2227/CyO	msl2227/+	27	23	85
		CyO/+	25	21	83
M	msl31	msl31/+	<0.25	0	401
N	msl31/TM3, Sb1	msl31/+	18	12	68
		TM3/+	12	8	65

Mothers homozygous and heterozygous for MSL mutations were siblings from crosses of the form: msl/msl ♀♀ × ♂♂ msl/Balancer. Full genotypes of stocks used as in Table 1 and Table S1, except y w; y w sn, and msl1^KmB/In(2L2R)BcGla¹. Males were: sc^3-1 w cm¹ ct⁶ sisA¹/Y. Crosses at 29°. *Female viability in cross E was not significantly different from either the total, or the Gla/+, female progeny of cross F (P = 0.13 for both, Fisher’s exact test).

I used msl3¹, the only msl mutation that appeared to interact with individual XSE components in the experiments of Table 1 and Table S1, as a proof of principle for the maternal-effect enhancement of sc sisA/++ lethality (Table 2, crosses M and N). The sc sisA combination exerted a considerable lethal effect on its own, as only 15% (20/133) of sc sisA/++ daughters of msl3¹/+ heterozygotes survived (cross N). As expected, the lethal impact of the XSE mutations was strongly enhanced (>50-fold) when mothers were homozygous for msl3¹ (0 daughters/401 sons) (Table 2, cross M). In contrast, and in agreement with the results in Table 1 and Table S1, I found that none of the other msl mutations enhanced the lethality of sc^3-1 sisA¹ /++ heterozygotes (Table 2, crosses C–L).

There are no zygotic lethal interactions between msl and XSE mutations

The results of the crosses shown in Table 1, Table 2, and Table S1 provide no support for the claim (Gladstein et al. 2010; Horabin 2012) that maternal mle, msl1, or msl2 mutations affect the assessment of chromosomal sex or influence the activity of Sxl. Might there still be a role for zygotically expressed products of these, or other, MSL complex genes in Sxl activation? The question is important because data presented in figure 2B of Gladstein et al. (2010) were interpreted as showing that zygotically contributed mle¹, msl1^L60, msl2¹, msl3¹, males absent on the first² (mof²), and RNA on the X^ex6 (rox^ex6) alleles all enhanced the lethality of sc^3-1 sisA¹/++ females. Any such zygotic lethal interactions should have been apparent in the results shown here in Table 2 as a reduction in the number of msl/+ female offspring compared to their balancer-bearing sisters. In no case was such a deficit observed (Table 2, crosses D, F, H, J, L, and N). Reasoning that reducing the dose of two zygotic components of the MSL complex should have a stronger impact on viability than the loss of one, I asked whether a msl1^L60 msl2 double mutant chromosome enhanced the lethality of females carrying the sc and sisA mutations. Once again, I found no evidence of a lethal interaction as both the double msl1 msl2 mutant and balancer chromosome-bearing sc⁷¹ sisA¹ /++ females were equally viable (Table 3A).

Table 3. There are no zygotic female-lethal interactions between msl1 and msl2, msl3 or rox1 mutations and XSE mutations.

X Chr. genotype	2nd Chr. genotype	3rd Chr. genotype	Viability (%) compared to reference ♀	No.
Cross A: msl1L60 msl2/In(2LR)Gla ♀♀ X ♂♂ y sc71 cm1 ct6 sisA1/Y; Dp(1;2)Hwbap (sc+)/+
++/sc71 sisA1	+/Dp(1;2)sc+	msl1L60 msl-2/+ +	90	81
++/sc71 sisA1	+/+	msl1L60 msl-2/+ +	14	13
++/sc71 sisA1	+/Dp(1;2)sc+	Gla/ +	Reference	90
++/sc71 sisA1	+/+	Gla/ +	13	12
Cross B: y sc96 cm1 ct6 sisA1 /FM7c ♀♀ X ♂♂ +/Y; msl31 / Rap1
X Chr. genotype		3rd Chr. genotype	Viability (%) compared to reference ♀	No.
sc96 sisA1/++		msl31/+	32	24
sc96 sisA1/++		Rap1/ +	26	17
FM7c/ + +		msl31/+	114	75
FM7c/ + +		Rap1/ +	Reference	66
Cross C: y w rox1ex6 + +/ + + + cm1 ct6 ♀♀ X ♂♂ sc3-1 w cm1 ct6 sisA1/Y
X Chr. genotypea			Viability (%) compared to reference ♂	No.
rox1ex6 sc3-1 sisA1			80	103
rox1+/sc3-1 sisA1			57	75
rox1ex6/Y			Reference	129
rox1+/Y			102	131

Full genotypes of crosses as shown. Dp(1:2)Hw^bap is a duplication providing sc⁺ function (Cline 1988). It dominantly expresses ectopic abdominal bristles. Cross C at 29°.

^aOnly flies nonrecombinant for the w-ct interval were scored.

The rox-1 RNA component of the MSL complex is not provided maternally and the transcript is expressed zygotically in early embryos (Lim and Kelley 2012). Gladstein et al. (2010) reported that the rox-1^ex6 allele exhibited a strong zygotic female-lethal synergism with sc and sisA mutations. Cross C of Table 3 shows the results of a cross in which females carrying marked rox-1^ex6 and rox-1⁺ X chromosomes were crossed to sc^3-1 sisA¹/Y males. I scored male and female progeny nonrecombinant for the w ct interval and found no evidence for a lethal interaction between the rox-1^ex6 mutation and the XSEs. Indeed, the viability of sc^3-1 sisA¹/++ females inheriting the wild-type rox-1⁺ allele (57%) was more compromised than those bearing the rox-1^ex6 mutation (80% survival).

msl3¹, alone among the msl lines I tested, exhibited lethal interactions with XSE mutants (Table 1, Table 2, and Table S1). Gladstein et al. (2010) reported that this lethal interaction also had a zygotic component. The results of cross N of Table 2 indicate, however, that the lethal interaction with msl3¹ flies is exclusively maternal as both msl3¹/+ and TM3/+ female progeny were equally viable. Because balancer chromosomes are not necessarily neutral in their effects on primary sex determination (Cline 1988), I wished to determine if there was any evidence of a zygotic interaction between msl3¹ and sc and sisA under conditions where the TM3 balancer normally carried in the msl3¹ stock was absent. Accordingly, I crossed males transheterozygous for msl3¹ and the dominant eye marker Rap¹ with females carrying the sc⁹⁶ and sisA¹ alleles balanced with FM7. Heterozygous sc⁹⁶ sisA¹/++ progeny were equivalently viable whether they carried msl3¹ or Rap¹, confirming there is no zygotic component of the lethal interaction between the msl3¹ chromosome and the XSEs (Table 3, cross B).

The data presented here do not support the hypothesis that the MSL complex functions in the early regulation of Sxl. Despite extensive tests, I was unable to identify any of the genetic interactions that would be predicted to occur between mle, msl1, or msl2 and either Sxl or the two key zygotic regulators of Sxl, scute, and sisA. Nor was I able to find support for the claimed zygotic role of rox-1 in early Sxl activation. In this context, the strong female-lethal maternal effect observed with msl3¹ females suggests one of two explanations: Either msl3 regulates Sxl independent of the MSL complex or the msl3¹ stock contains an unrecognized mutation(s) responsible for the maternal effect interaction.

An unidentified mutation is responsible for the interactions seen with an msl3 mutant

To determine which of these alternatives is true, I examined other msl3¹ isolates and additional msl3 alleles to see if they retained the interaction with the X-signal components. I obtained four additional msl3¹ isolates from J. Birchler (University of Missouri), V. Meller (Wayne State University), and M. Kuroda (Harvard University), some labeled by their historical synonyms, mle(3)132 and msl3^P. Each of the msl3¹ stocks, regardless of origin, exhibited maternal-effect lethal interactions with Sxl, scute, and sisA (data not shown). I also received the msl3^MAK-1 allele from J. Birchler but found that the stock had acquired a mutation, rendering the second chromosome homozygous lethal, so I was unable to test directly if msl3^MAK-1 mutants exhibited the maternal-effect lethal interaction. Instead, I carried out complementation tests with msl3¹, msl3^MAK-1, and two chromosomal deficiencies, Df(3L)Exel6110 and Df(3L)BSC224, which are deleted for all msl3 coding sequences (Table 4). I found that msl3^MAK-1 and the two deficiencies fully complemented the female-lethal maternal effect of the msl3¹ chromosome as evidenced by the absence of any lethal interaction with sisA¹ or sc^M6 (Table 4). The complementation tests thus demonstrate the female-specific maternal effects previously attributed to msl3 are likely due, at least in part, to one or more additional mutations in msl3¹ fly stocks.

Table 4. msl3 is not sufficient for the maternal-effect female-lethal interaction with XSE mutations.

	Viability (%) of females of indicated X Chr. genotype compared to sibling males (no. reference)
Maternal 3rd Chr. genotype	sisA1/ +		scM6/ +
msl31	42	(117)	9.2	(174)
msl3MAK-1/msl31	110	(152)	103	(241)
Df(3L)Exel6110/msl31	102	(101)	102	(477)
Df(3L)BSC224/msl31	92	(104)	99	(135)

Experimental mothers derived from crosses: msl3¹ red ♀♀ X ♂♂ msl3¹ red/TM3, Sb¹ Ser¹ or ♂♂ msl3^MAK-1/TM3, Sb¹ Ser¹ or ♂♂ w¹¹¹⁸; Df(3L)Exel6110/TM6, Tb¹, or ♂♂ w¹¹¹⁸; Df(3L)BSC224/TM6c, Sb¹ cu^1. Paternal genotypes: y cm¹ ct⁶ sisA¹/Y or sc^M6 w/Y.

An important question is whether the maternal effect lethal interaction requires the loss of msl3 function, as might be the case if both a deficit in msl3 and another gene were required, or if the lethal effect is entirely independent of msl3. To address this point, I asked if msl3¹ females that carried an msl3⁺ transgene that fully complements msl3 male lethality also exhibited the maternal effect. Initially I tested a y w; P[w⁺; msl3⁺-TAP]; msl3¹ stock (Alekseyenko et al. 2006) and found that the stock did not exhibit the female-lethal maternal effect (data not shown). This result could have occurred either if an msl3 mutation is necessary, but not sufficient for the maternal defect, or if the responsible mutation(s) had been lost from the strain. I therefore generated two independent w¹¹¹⁸; msl3¹/TM3 stocks that retained the maternal-effect female-lethal interaction and introduced the P[w⁺; msl3⁺-TAP] transgene into each, taking care to ensure that the suspect msl3¹ chromosome from the y w transgenic stock was excluded. Sibling w¹¹¹⁸; msl3¹ mothers, differing only in whether or not they carried a copy of P[w⁺; msl3⁺-TAP], were tested to determine if there was a female-lethal interaction with the sc^M6 allele (Table 5). I found that the presence of the fully functional msl3⁺ transgene had no effect on the female-lethal interaction as sc^M6/+ female viability was equally compromised regardless of whether the maternal genotype was functionally msl3⁺ or msl3⁻ (Table 5). Note that crosses A and B in Table 5 incorporated a second chromosome duplication of sc⁺ as a control to ensure that the female-lethal interactions observed remained dependent on the sc^M6 allele present in the progeny females. Taken together, the complementation experiments with msl3 mutants (Table 4) and msl3⁺ transgenes (Table 5) establish that msl3 mutations are neither necessary nor sufficient for the observed maternal-effect interactions with the XSE genes.

Table 5. The msl3¹ mutation is not necessary for the female-lethal maternal-effect interactions with the XSE scute.

	♀♀ X ♂♂ scM6 w/Y; Dp(1;2)Hwbap (sc+)/+
Crosses A and B	Maternal genotype	2nd Chr. genotype	Female viability (%) compared to brothers	No. female progeny	No. male progeny (reference)
Cross A	w; msl31	+/Dp(sc+)	98	65	66
	“	+/+	4	3	79
Cross B	w: P{msl3+-TAP}/+; msl31	P{msl3+-TAP}/Dp(sc+)	98	63	64
	“	P{msl3+-TAP}/+	7	4	60
	“	+/Dp(sc+)	127	71	56
	“	+/+	8	6	72
Crosses C and D	♀♀ X ♂♂ scM6 w/ Y
	Maternal genotype	2nd Chr. genotype	Female viability (%) compared to brothers	No. female progeny	No. male progeny (reference)
Cross C	w; msl31	+/+	2	6	284
Cross D	w: P{msl3+-TAP}/+; msl31	+/+	0.9	1	113
	“	P{msl3+-TAP}/+	<1.1	0	94

Females in crosses A–D were derived from independently isolated w¹¹¹⁸; msl3¹ lines. It is not known if the msl3¹ chromosomes retain the red allele. Experimental mothers derived from crosses between w¹¹¹⁸; msl3¹ females and w¹¹¹⁸; P{w⁺,msl3⁺-TAP]/+; msl3¹ males and differ only in the presence or absence of the msl3⁺ transgene. Male genotypes are as shown.

Collectively, the data presented here argue that neither the MSL complex nor its component parts participate in the early embryonic activation of Sxl. This undercuts the notion that females use the MSL complex to amplify the X chromosome signal (Gladstein et al. 2010) and eliminates an important challenge to standard models for sex determination and dosage compensation. My experiments also suggest the existence of at least one novel maternal regulator of primary sex determination. The molecular identification of this locus, which has escaped detection in a variety of screens, is a high priority, as it is likely to offer new insights into the mechanism of X-chromosome counting.

Comparisons of differing results

The conclusions I reach in this paper are in direct opposition to those made earlier by Gladstein et al. (2010) and Horabin (2012), and the data also appear to be in conflict. How then can one explain the differences between the results in this report and those published previously? With respect to msl3 the answer is clear. Because Gladstein et al. (2010) neither mapped the responsible lesion nor performed a complementation test, they reached the incorrect conclusion about the involvement of msl3 in sex determination. Remarkably, Uenoyama et al. (1982) offers evidence that the original isolate of msl3¹ contained a maternal-effect mutation exhibiting a female-lethal interaction with Sxl^f1. For convenience, data from tables 1 and 2 of Uenoyma et al. (1982) are reproduced here with explanatory notes and current nomenclature as Table S2. Crosses 1a and 1b of Uenoyma et al. (1982), comparing sibling msl3¹ red and msl3¹ red/TM3 mothers, showed a modest female-lethal maternal-effect interaction with Sxl^f1. Crosses 2a and 2b, however, reveal that when the experimental mothers carried a recombinant msl3¹ red chromosome that also included the distal marker ebony (e), the maternal-effect lethal interaction was lost (see Table S2). In retrospect, it is not surprising that the msl3¹ stock would carry additional mutations. The msl3¹ chromosome was recovered from a wild population in Japan in the 1970s (Uchida et al. 1981) and may well have experienced mobilization of P-elements while introducing markers from laboratory strains. In fact, Uchida et al. (1981) reported the original chromosome also bore a nonsex-specific maternal-effect lethal that mapped between red and ebony. The relationship between that mutation and the one responsible for the interactions with Sxl and the XSEs is not known.

It is less obvious what might explain the differences between what is reported here and what was published earlier for mle, msl1, and msl2. I suggest the resolution is that the differences are more illusory than real because Gladstein et al. (2010) drew their conclusions from equivocal genetic experiments. Consider figures 1B and 2B of Gladstein et al. (2010), which addressed whether maternal or zygotic msl mutations enhanced the female-lethal effects of the sc^3-1 sisA¹ X chromosome. Instead of comparing the viability of the offspring of heterozygous and homozygous msl mothers, or between sibling sc sis/++; msl/+ and sc sisA/++; +/+ females, Gladstein et al. (2010) measured viability with reference to the unrelated wild-type Oregon R strain. They concluded that each msl had maternal and zygotic effects on female viability because the lethal effects observed with the msl mutants were greater than seen with the Oregon R controls. The problem with this approach is that it relies on quantitative comparisons between flies of unrelated genetic backgrounds. This was compounded by the choice of Oregon R, a strain particularly ill-suited to serve as a reference because it is among those least sensitive to the female-lethal effects of reduced XSE gene dose (Cline 1988). In effect, the choice of control strain may have predetermined that stronger female-lethal effects would be observed with the msl mutant strains than with the control, which would have given the appearance of female-lethal interactions where none existed.

The only cases where there appear to be actual conflicts between my findings and those published earlier are with respect to the abilities of maternal msl1^L60, msl2¹, and perhaps mle¹, to create synthetic female-lethal interactions with a Sxl null allele (figure 1B in Gladstein et al. 2010 vs. Table 1 and Table S1). I cannot explain the discordant data because Sxl/+ heterozygotes should be fully viable in the absence of interacting mutations, but note that Gladstein et al. (2010) examined only a single mutant line for each locus and did not determine if the effects were dominant or recessive or if they were maternal. The latter points are important because if the maternal effects were equal between mothers homozygous and heterozygous for these null, or near null, msl alleles, that result would suggest the lethally interacting loci were unlikely to be the msl genes.

Maternal effects of the msl2 gene?

The question of whether there are maternally contributed msl gene products is of crucial importance for assessing a possible role of the MSL complex in the early embryo, whether related to sex determination or early dosage compensation (Gladstein et al. 2010; Lott et al. 2011; Horabin 2012). There is strong evidence that maternally provided mle and msl1 products function in the embryo as assembly of the MSL complex is delayed, and the lethal period shortened, in the male progeny of mothers mutant for the two loci (Belote and Lucchesi 1980; Franke et al. 1996). The most crucial questions concern msl2 as the protein is essential for the formation and function of the MSL complex. Numerous experimental measures suggest that there is no maternal contribution (Bashaw and Baker 1995; Kelley et al. 1995; Rastelli et al. 1995; Franke et al. 1996; McDowell et al. 1996; Lott et al. 2011, 2014). RNA-seq data are potentially helpful, but findings from precisely staged embryos that lack msl2 mRNA (Lott et al. 2011) have been criticized for lack of statistical power (Horabin 2012), whereas bulk samples from 0- to 2-hr-old embryos that contain msl2 mRNA (FlyBase, cited by Gladstein et al. 2010 and Horabin 2012) could be explained by even moderate contamination with older embryos. It is thus worth reexamining the genetic evidence cited (Gladstein et al. 2010; Horabin 2012) in support of a maternal effect of msl2. Apart from the incomplete experiment in figure 1A of Gladstein et al. (2010), which does not formally address a maternal effect, that evidence comes from crosses performed by Uenoyama et al. (1982). As reprinted in Table S2, two effects were observed in crosses between msl2²⁷ mothers and Sxl^f1/Y; msl2²⁷/SM1 fathers: a small female-lethal effect and a low frequency of apparent sex transformations in homozygous msl2²⁷ female progeny (Uenoyama et al. 1982). With respect to viability, there was no evidence for a maternal effect as there was no difference between the female progeny of homozygous and heterozygous msl2²⁷ mothers (88 vs. 86% viability). With respect to sex transformation, 8% of female msl2²⁷ progeny of homozygous mothers exhibited at least one sex-transformed structure compared to 4% of the msl2²⁷ progeny of heterozygous mothers (Table S2, crosses IIIa, IIIb; Uenoyama et al. 1982). Given that the basis of the sex transformations remains mysterious, that they occur only in 4–8% of progeny, and are affected by the zygotic genotype, the 1982 sex transformation data do not make a convincing case for an msl2 maternal effect. In light of the strong evidence that there is no detectable maternal msl2 protein, or mRNA, and the absence of genetic evidence for an msl2 function in embryos, the simplest interpretation is that msl2 mRNA and protein first appear, and the MSL complex first assembles, after Sxl activity has been set in cycle 14 (Franke et al. 1996; Lott et al. 2011, 2014).

Conclusions

The data presented here indicate that the MSL complex does not participate in the early regulation of Sxl in female embryos. Most msl mutants did not exhibit the genetic interactions predicted to occur if the complex directly, or indirectly, affects the early steps in sex determination. In the sole case where an msl mutant did interact as predicted, the msl3 mutation was shown to be unnecessary for the effects, which appear instead, to be the result of at least one unidentified maternal-effect locus. My experimental findings nullify the rationale behind an important alternative model of primary sex determination. By extension, they also undercut the proposal that the MSL complex regulates the process by which some X-encoded genes are dosage compensation prior to the large-scale activation of the zygotic genome (Lott et al. 2011).