In animals with an unmatched (heteromorphic) pair of sex chromosomes, all sex differences in the phenotype originate from the unequal effects of the sex chromosomes because they are the only factors that consistently differ between male and female zygotes. In species such as mammals in which the sex chromosomes are XX and XY, a fundamental goal is to identify the sex-biasing effects of the two sex chromosomes on phenotypes. This review discusses the strategies for identifying X and Y genes or mechanisms that cause sexual bias, using mouse models that differ in their sex chromosome complement, including differences in the parental source of the X chromosome.

Primary among the sexually unbalanced effects of X and Y genes are those that cause sexual differentiation of the gonads. In mammals, the most important sex differences in the gonads are thought to be initiated by the Y-linked gene Sry, which is expressed in the indifferent gonadal ridge of males and induces there a cascade of molecular and cellular events that commit that tissue to a testicular fate [1–4]. In the absence of Sry in the XX female, ovaries develop through an active developmental program that includes inhibition of testicular pathways [5]. Therefore, the developmentally limited male-specific expression of Sry in the gonads sets up a lifelong sex difference in secretion of testicular vs. ovarian hormones. Because sex steroid hormones, especially androgens and estrogens, are considered the major class of proximate signals that act throughout the body to cause sexual differentiation, the Sry effect has long been thought to be primarily responsible for determining the sexual phenotype of most mammalian species.

In addition to the effects of Sry on gonadal tissues, X and Y genes may be expressed at different levels in XX and XY non-gonadal cells as a consequence of their imbalance in the genomes of all XX and XY cells. By virtue of this imbalance, X and Y genes are known to cause numerous sex differences in phenotype [6–11]. These X and Y gene effects are here called “direct” effects of the sex chromosomes, because the sex bias in the X and Y gene expressions acts directly to cause sex differences in non-gonadal tissues instead of indirectly via an action on the gonads to induce sex differences caused by gonadal hormones [12]. Historically, it has been much easier to study the effects of gonadal hormones, which can be administered or blocked or withdrawn by simple experimental procedures. It has been much more difficult to discover the direct sex-biased effects of the X and Y genes on non-gonadal tissues, because altering the sex chromosome complement usually causes changes in the type or function of the gonads, and therefore introduces confounding differences in the level of gonadal secretions.

Here, we review mouse models (see also [13]) that enable the separation of sex chromosome complement effects from the hormonal confounds, thus facilitating the identification of direct sex chromosome effects (direct SCEs) that contribute to sex differences. We focus especially on direct SCEs that are expected to contribute to sex differences in the function of normal XX and XY cells. The goal of our studies, and of this review, is not only to discuss how to detect direct SCEs but also to map out a strategy for identifying the X or Y genes that cause sex differences in cells. The strategy represents a logic tree, which progressively narrows down the potential candidate genes responsible for direct SCEs.

The mammalian sex chromosomes are thought to have evolved from an ancestral pair of autosomes. One autosome, the proto-Y chromosome, acquired a dominant male-determining locus, which led to the loss of recombination with the proto-X, and ultimately, to the wholesale loss of gene-encoding DNA from the Y chromosome [14, 15]. However, there have been gene additions to the X and/or Y chromosome subsequent to their divergence [16–20]. The present-day Y chromosome is usually small and gene-poor relative to the X chromosome [21, 22]. However, the mouse Y has an unusually high gene count as a result of massive gene amplification that is thought to be driven by an ongoing post-meiotic X-Y genomic conflict [22, 23]. During meiosis, the X and Y chromosomes pair at the pseudoautosomal regions (PARs), thus enabling the X and Y PARs to recombine [24–27]. The X and Y PARs are therefore identical on average between males and females and are not thought to cause sex differences in phenotypes. The number and type of sex chromosomes present in females or males are referred to as the sex chromosome complement, and we use this term to encompass the parental imprinting and inactivation status of the X chromosomes (see below). For the purposes of this review, we break down the sex chromosome complement into a number of components that could contribute to sex differences in phenotypes. Aside from the PARs, which are present in two doses in female and male mice, the genomic dosage of these components is not the same in males and females (Table 1 and Fig. 1). Thus, females have two copies of the non-PAR region of the X chromosome (NPX), whereas males have one NPX and one non-PAR Y (NPY). Females also differ from males in that they receive an X chromosome from each parent, whereas males only receive an X from their mother. Because of parental imprinting, gene expression can differ between the maternal and paternal X chromosomes. Also, as is discussed more fully below, one of the X chromosomes in females is “inactivated.” Thus, about half of female cells inactivate the maternal X chromosome and experience the paternal X imprint (unlike any male cell), whereas the other half of the cells inactivate the paternal X chromosome and experience the maternal X imprint (similar to male cells). The inherent imbalance in the sex chromosome complement that results in a difference in gene expression between XX and XY cells can therefore be summarized as follows: (1) the absence vs. the presence of the NPY genes, (2) the difference in genomic dose of NPX genes (two vs. one copies), (3) any differences due to the parental imprint on the NPX genes, and (4) the presence vs. the absence of an inactive X chromosome (Table 1).

Genotypes	Gonads	NPX	NPY [−Sry]	Sry	PAR [−Sts]	Sts	Xm	Xp	Xi
XX (40,XX)	F	2	0	0	2	2	1	1	1
XY (40,XY)	M	1	1	1	2	2	1	0	0

Sry and Sts are separated from the chromosomal regions that encompass them because mouse variants with sex chromosomes deleted for one or both of these genes are utilized in crosses

NPX non-PAR X genes, NPY non-PAR Y genes (excluding Sry), PAR pseudoautosomal region (excluding Sts), X ^p X chromosome of paternal origin, X ^m X chromosome of maternal origin, X ⁱ the inactive X which may be of maternal or paternal origin

Without any compensatory mechanism, the double genomic representation of X genes in XX cells, compared to the single representation of those genes in XY cells, would create an imbalance in X gene expression between the sexes (F>M); such a chromosome-wide sexual imbalance in expression is thought to be disadvantageous because X genes would not be in balance in both sexes with autosomal genes that they drive or that drive them [28]. X-inactivation, the process by which one X chromosome is largely transcriptionally silenced in non-germline XX cells, is an effective mechanism that reduces the imbalance in X gene expression. However, X-inactivation does not eliminate sex differences in the expression of X genes but reduces it to match approximately the level of sexual dimorphism in the expression of genes encoded on autosomes, which are present in two copies in each sex [29]. It also introduces the possibility of epigenetic effects of the inactive X, which are discussed more fully below.

X inactivation leads to mosaicism of X gene expression in XX mice. In eutherian mammals such as the mouse, the choice to inactivate the paternal or maternal X chromosome is random and patchy in tissues stemming from the inner cell mass of the blastocyst (i.e., the tissues giving rise to the body of the embryo), so that some somatic XX cells predominantly express X genes inherited from the mother, and others express X genes inherited from the father [30]. In outbred or natural populations that carry genetic polymorphisms, there are two types of X mosaicism—of the alleles and imprints. Because the maternally derived X chromosome will have many alleles that are different from those on the paternally derived X chromosome, roughly half of the cells in the XX mice will express the paternal X alleles and show the effects of any paternal imprint on X genes, and the other half will express the maternal X alleles and evince the effects of any maternal imprint on X genes. Thus, females are expected to have fewer extreme tissue phenotypes, because the extreme effects of specific X alleles (e.g., causing abnormal function) is often buffered by the expression of another allele in other cells of the same individual [31–33]. These effects are thought to lead to sex differences in the variability of tissue phenotypes (XX females less variable than XY males), and broader adaptiveness of female tissues to diverse environments because of the inherently greater diversity of X alleles within an individual female vs. an individual male. However, an overall reduced variability of X gene expression in human females, relative to males, has not been supported [34] but might be important under extreme environmental conditions or disease.

Thus, we can expand the list of possible sources of sex differences in phenotype arising from differences in sex chromosome complement, to include the following: (5) The inherent mosaicism of XX tissues arising from random X inactivation, which can lead to less phenotypic variability among females than males and to average differences in phenotype between the sexes; (6) sex differences in the prevalence of specific X alleles that may also lead to average phenotypic differences between the sexes. However, when studying inbred mice, all X alleles are the same in the two sexes, so these additional sources of sex differences are not operative.

Finally, it is possible that the difference in sex chromosome complement between normal males and females can lead to direct SCEs that are linked to the epigenetic status of the sex chromosomes rather than differences in X and Y gene expression [35–38]. One example of this arises from the fact that in somatic cells, the inactive X (Xⁱ) in females is rich in heterochromatin, reflecting the limited transcriptional activity. Heterochromatinization depends on enzymes that introduce repressive marks on histones, and it is hypothesized that heterochromatinization of the X competes with the introduction of repressive marks at some autosomal loci, thus affecting autosomal gene expression in females. This could be linked to a direct SCE even though it is not linked to an XX vs. XY difference in X gene expression. We do not consider this type of epigenetic contribution to direct SCEs further.

An ideal mouse model for identifying direct SCEs is one that provides mice that differ in their sex chromosome complement (XX vs. XY) but not in the level of gonadal secretions. In practice, this is an elusive goal, but some experimental paradigms are informative even if strict control of gonadal hormone levels may not always be possible.

The strategies outlined below are based on the assumption that the direct SCEs are a consequence of differences in gene expression resulting from the difference in sex chromosome complement. If the tissue underlying the direct SCE can be identified with a high degree of confidence, then a case may be made for using RNA sequencing (RNAseq) on this tissue from the FCG in order to provide a list of candidate genes at this stage (see sections below beginning with “Candidate genes for direct SCEs). We will refer to this tissue as the “target tissue.” The candidate genes will be either on the NPX or NPY. The following crosses provide a logical approach to determining the extent to which NPX and/or NPY genes are responsible for direct SCEs.

Important: see Additional file 1, which provides the details of the three-generation breeding strategy needed to produce the fathers for Cross B. This involves progeny from Cross C.

Identifying NPX and NPY effects using Cross A: XX♀ x XY*♂

This cross is the best starting point for identifying potential NPX and NPY effects. The fathers have a variant Y chromosome (denoted Y*) that leads to the generation of a minute X chromosome derivative (termed Y*^X for historical reasons) composed of a complete PAR, no NPY, and a small subsegment of NPX (NPX+) currently estimated to be less than 1% of the total NPX (Table 3, Fig. 3) [70, 71]. As we shall see, the resulting X^mY*^X female progeny are used for Cross C that enables NPX dosage effects to be distinguished from NPX parental imprinting effects.

Genotypesa	Gonads	NPXb	NPY	Sry	PAR [−Sts]b	Sts	Xm	Xp	Xi
XmO (rare)c	F	1	0	0	1	1	1	0	0
XmY*X	F	1+	0	0	2	2	1	0	0
XmXp	F	2	0	0	2	2	1	1	1
XmY*	M	1+	1	1	3	1	1	0	0
XmXpY*	M	2	1	1	3	1	1	1	1

^aParental sources of the X chromosomes (X^m: maternal; X^p: paternal) are included in Tables 4, 5, and 6, since these crosses are also used to identify effects of parental X imprinting

^bFigure 3 illustrates the structures of the Y*, X^Y*, and Y*^X chromosomes and the gene content of Y*^X that includes the minute + subsegment of NPX that is also present in Y*. Note that the Y* and X^Y* PARs have a duplication of PAR regions A and B but lack PAR region C that encompasses the Sts locus

^cX^mO are rare (~1% [75])

1. 1.

   Identifying an effect of NPX (two vs. one copies).

   A comparison of X^mX^p vs. X^mY*^X is a test of the effect of the number of NPX segments (2 vs. 1), keeping the number of PARs = 2.

   Cross A also tests for an NPX effect in the comparison of the two gonadal male groups, X^mX^pY* with X^mY*, although in this case, the comparison deviates from the precise genetic conditions found in the XX-XY comparison in three ways: the presence of NPX+ in X^mY*; single copy of Sts in X^mY* instead of two copies found in XX or XY; and three copies of the region of the PAR that does not contain Sts, instead of two in XX or XY (Table 3). If the phenotypic difference between females containing one vs. two copies of NPX is similar to the difference between males with one vs. two copies of NPX, these deviations from normal XX and XY are generally considered minor [47].

    
2. 2.

   Identifying an NPY effect.

   The effect of NPY can be detected in a comparison of X^mY*^X vs. X^mY* and X^mX^p vs. X^mX^pY*. These comparisons are confounded with differences in the number of PAR segments as shown in Table 3. The Y* also has a different strain origin (LT/Sv) from that of the Yˉ from the FCG (strain 129/SvEv-Gpi1c). Thus, if the investigator finds no effect of the Y chromosome in this cross, it could be a Y chromosome strain difference. On the other hand, if a Y effect is found, then one can conclude that at least some Y chromosomes create a difference relative to XY*^X mice, which therefore may explain the original sex chromosome effect in FCG. Because the Y* chromosome encodes Sry, the NPY effect could be the result of testicular secretions; that possibility will already have been evaluated by a study of FCG mice, which will indicate whether the phenotype is influenced by gonadal hormones or not. If Cross A is used without knowledge of the outcomes from studies of FCG mice, it will not be possible to eliminate the hormonal confound of the Y* chromosome.

    
3. 3.
   Interpretation of outcomes

   1. (a)

      NPX copy number effect only. The different effects of one vs. two copies of NPX can be caused by dosage differences (inherently higher expression of genes that escape X inactivation in XX than XY) or by X imprinting effects (different expression levels of paternally vs. maternally imprinted X chromosomes). Importantly, this cross tests for an effect of NPX copy number both in gonadal males (with Sry) and gonadal females (without Sry). If an NPX copy number effect is found in Cross A and in FCG mice, the results indicate that the effect does not depend on the Sry transgene in the FCG model.

       
   2. (b)

      NPY effect only. The SCE is attributed to NPY genes. Numerous NPY genes have an X “partner” gene that theoretically has a similar function (see section on “Candidate genes for direct SCEs” below). If the NPY effect is caused by one of these genes, the implication is that the NPY gene has an effect distinct from its X partner.

       
   3. (c)

      NPX and NPY effects. In this case, it is possible that X and Y partner genes are involved, with the X gene escaping X inactivation [50]. Alternatively or additionally, the NPX effect could be due to imprinting.

       
    
4. 4.
   Advantages and drawbacks

   1. (a)

      This a relatively simple cross that enables comparisons to be made quickly and economically.

       
   2. (b)

      Some of the X^mY*^X female offspring can be utilized as mothers for Cross C that enables parental X imprinting effects to be identified.

       
   3. (c)

      It has the minor disadvantage that the Y* chromosome has a different strain origin to that in the FCG cross.

       
   4. (c)

      It has the disadvantage that Y* encodes Sry, and thus all NPY-bearing mice are gonadal males, and differ hormonally from gonadal females that lack a Y chromosome. Thus, it does not test for an effect of NPY independent of its effect on gonadal secretions, unless prior studies of FCG mice indicate that the phenotype is not sensitive to gonadal hormones. This disadvantage is remedied in Cross B.

       
   5. (d)

      NPX is subject to X inactivation in X^mX^p. A minor concern is that in X^mY*^X the small NPX+ segment present on Y*^X is exempt from X inactivation because it does not include Xist. Genes in this segment are therefore expected to be expressed from both the X and Y*^X chromosomes [72]. In contrast, the same segment in X^mX^p is expected to undergo random X inactivation and thus be expressed only from the active X chromosome. If these genes are expressed in the target tissue of the direct SCE, it is therefore likely that they will be expressed higher in XY*^X than XX. However, only a few known genes are encoded by NPX+ (Fig. 3) and they can be discounted if these genes show no difference in expression in the relevant target tissue.

       
    

Here, we provide an overview of the sex chromosome genes whose differential expression between XX females and XY males may lead to direct SCEs in the target tissue in which the direct SCE originates. These comprise the following: (i) X genes that have parental imprints that lead to differential expression between males and females in the target tissue; (ii) X genes that escape X inactivation in the target tissue, an important group of which are X genes that have functionally similar partner genes on the Y chromosome (X-Y gene pairs that are termed “gametologs”); (iii) Y genes that are expressed in the target tissue of XY males.

Here, we consider strategies for identifying the genes underlying a direct SCE when NPX (but not NPY) has been implicated in Cross A or B. The first step is to use the FCG RNAseq data to generate a list of X genes that are differentially expressed in the target tissues of XXF vs. XYF and/or XXM vs. XYM FCG mice, under conditions in which the differences are unlikely to be caused by gonadal hormones.

The X genes that exhibit expression differences compatible with the direct SCE identified in the FCG mice would be prioritized for further study. The X genes might first be ranked with respect to the fold difference in the level of expression between the relevant FCG genotypes (e.g., XXF vs. XYF), and/or by the p value of a statistical test that estimates the reliability of the difference in the expression between groups. These candidate genes could be further prioritized based on their known characteristics and potential relevance to the phenotypes under investigation. If necessary, candidate genes would be checked using quantitative RT-PCR to confirm the expected group differences. If the candidate NPX gene is a reported X escapee, and is expressed higher in XX than in XY, as expected from its status as an escapee, its importance as a gene causal to the SCE would be established by comparing mice with one vs. two copies of the gene. For example, if Ddx3x (Table 6) is the candidate, one might compare WT XX mice (with two copies of Ddx3x) with XX mice with a heterozygous knockout of Ddx3x (one copy). If these groups of mice differ in the phenotype showing the SCE, such that one vs. two copies of Ddx3x mimics the phenotypic effects of one vs. two copies of NPX, then Ddx3x is a likely contributor to the SCE. This experiment determines if dosage differences in Ddx3x alone are necessary to show the SCE. An additional strategy is to add a transgene encoding Ddx3x to an XO or XY*^X mouse, to determine if the phenotype is affected appropriately in mice with one copy vs. more than one copy of the candidate X escapee. The transgenic add-in experiment aims to test if the dosage differences in the X escapee are sufficient by themselves to mimic the SCE. The main caveat is that expression levels of the transgene may be difficult to control depending on its promoter, position effects, etc.

The experiments comparing Cross A vs. C may suggest that the NPX effect is the result of imprinting. In this case, transcriptome profiling may reveal either XX > XY or XY > XX expression patterns (Table 8) and differences in expression of transcripts in X^mY*^X vs. X^pY*^X. To provide further strong support for an NPX imprinting explanation of an SCE, one might measure the effects of parent of origin on the epigenetic status of the candidate gene and manipulate its expression to mimic the differences in expression found in X^mY*^X vs. X^pY*^X, to demonstrate that these differences in expression alone are necessary and sufficient to mimic the SCE, at least in part.

The manipulation of NPX expression is best performed in a tissue-specific manner, assuming that the tissue of interest is known. For example, a floxed allele of the candidate gene can be knocked down with a tissue-specific Cre. Indeed, the use of tissue-specific knockdown or overexpression may help determine the tissue in which the candidate gene causes the SCE.

Here, we consider strategies for identifying the genes underlying a direct SCE when NPY (but not NPX) has been implicated in Cross A or C. The first step is to check the RNAseq data to provide a list of the Y genes expressed in the SCE target tissue. Prioritizing the candidate genes may be possible based on the current information about the function of the Y genes or their similar X gametologs.

Given that only seven Y genes (Uba1y, Kdm5d, Uty, Usp9y, Ddx3y, Eif2s3y, and Ssty2) have been shown to be widely expressed in non-gonadal tissues as assayed by RNAseq [22], the “prioritized list” of genes expressed in a target tissue with a putative NPY-linked SCE will likely be limited to one or a very few genes. Thus, it makes sense to move on to the stage of manipulating candidate Y gene expression in order to demonstrate that the SCE of interest is a consequence of the presence vs. the absence of a particular NPY gene or genes. The basic approach is the same as for candidate NPX genes discussed in the last section “Identifying NPX genes that cause direct SCEs”. Keeping other factors constant (e.g., sex chromosomes and hormones), the investigator varies the presence/absence of one candidate Y gene to determine if that manipulation causes a phenotypic difference similar to the SCE caused by the comparison of XX vs. XY. Two basic approaches are available, either knocking out the NPY gene in XY mice [98–101] and comparing the knockout (KO) to WT XY or adding a transgenic copy of the candidate NPY gene in mice with one X chromosome (e.g., XO or XY*^X) and comparing to controls without the transgene [73, 102]. It would be best to start with global KO or transgenic, where the genetic manipulation affects all tissues, but tissue-specific gene targeting or transgenesis would ultimately be a powerful demonstration of the effect of the NPY in the target tissue. Transgenic lines encoding autosomal copes of NPY genes (Uty, Kdm5d, Ddx3y, Usp9y, Ubely, and Eif2s3y) were developed in the Burgoyne Lab and are being backcrossed to B6 in the Arnold Lab.

It is possible but less likely that multi-copy NPY genes account for the SCE. Manipulating these genes is more problematic. One option is to knock down the expression using a transgene that delivers small interfering RNAs such as has been successful for the multi-copy NPY gene Sly [103].

Investigators who have obtained evidence for NPY involvement might consider using crosses in Additional file 4 to map the NPY genes that underlie the direct SCE to specific regions of the Y. The first cross involves a commercially available Sxr ^a mutation that involves translocation of Y short arm (Yp) genes to the tip of the X or Y PAR and will establish if the SCE involves Yp genes that map to Sxr ^a [104]. If this is not the case, the second cross utilizing a Y deletion (Y^d1) should provide confirmation that multi-copy NPY genes are involved. A contact e-mail is provided in Additional file 4 for those wishing to use this Y^d1 deletion.

Here, we consider the strategies for identifying the genes underlying a direct SCE when NPX and NPY have similar effects in Cross A and/or B. The prime candidate genes are the ancestral X-Y gene pairs Uba1y/Uba1, Kdm5d/c, Zfy/x, and Uty/x (Utx A.K.A. Kdm6a), Usp9y/x, Ddx3y/x, and Eif2s3y/x; Table 6). As discussed in the section above, “Candidate genes for direct SCEs”, the X partners of these gene pairs are widely expressed dosage-sensitive regulatory genes [19, 20]. The conservation on the Y chromosome of the Y partner gene during evolution is thought to have been driven by dosage balance of expression and function of the X and Y gametologs, which implies equivalent function of the two genes. However, over time, there were varying degrees of sequence divergence between the X and Y copies, with the Y copies often acquiring some specific functions. Thus, a central unresolved question is whether the X and Y partner genes have similar or divergent function in different tissues and life stages. Limited evidence to date indicates that the function of the X and Y genes outside of the testes is different in some cases such as Uty/Utx [105].

The RNAseq data provides a list of the X and Y genes that are expressed in the target tissue, so it is reasonable to prioritize any X-Y gene pairs. If any X genes are known to show linkage to the trait, they would have a high priority whether or not they are a member of an X-Y gene pair. For X-Y gene pairs, the relative abundance of X + X transcripts in XX females compared to X + Y transcript abundance in males are not very informative when the X and Y transcripts have diverged such that the encoded proteins are not equivalent in activity and/or function. Once again, the most direct approaches to identifying the X and Y genes involved are gene targeting and transgenesis.

Nevertheless, the possible balance of effects of X-Y pairs could lead to an unusual scenario, in which Crosses A and B show effects of NPX dose, or NPY dose, when in fact, the study of FCG mice shows no direct SCE. In this case, a likely scenario is that both members of an X-Y gene pair have effects on the phenotype but that the effects are balanced, and therefore produce no phenotypic difference in XX vs. XY. Similarly, when gene dose is manipulated through knockout of single copies of X-Y gene pairs, reducing the dose of both X and Y partners could show a phenotypic effect that is balanced and does not contribute to sex differences in phenotype. The balance of X and Y partner genes could be dynamic, however, changing under different life conditions (stress, disease, developmental stage, tissue type, and hormone levels) [50].

Most investigators will be drawn to the study of FCG mice if they hypothesize a direct SCE because classical experiments of hormone removal and replacement fail to explain a known sex difference in phenotype [69]. However, FCG mice can be informative even if a sex difference in the phenotype is not established. In some cases, two sex-biasing factors typical of one sex (e.g., female levels of estradiol and a second X chromosome [47]) can reduce or eliminate the effects of the other [46, 106]. Reducing the effects of one sex-biasing factor can therefore “unmask” the effects of another. The “compensatory” actions of hormones and direct SCEs provides a rationale for comparing groups of FCG mice even if there is no overt sex difference in a phenotype [106]. In published studies, FCG mice have often been gonadectomized to allow comparison of groups under conditions in which the levels of gonadal hormones are equal, to reduce the possibility of gonadal hormonal confounds for direct SCEs. Direct SCEs discovered in gonadectomized (GDX) mice are sometimes, however, reduced if the mice have gonads or are treated with hormones after GDX [107]. Direct SCEs that are smaller in the presence of gonads or gonadal hormones are potentially important because various life events (aging, stress, disease) can result in reduced levels of gonadal hormones, which can result in the emergence of direct SCEs that contribute to sex differences in physiology. The interaction of direct SCEs and hormones, when the effect of one sex-biased variable is conditioned by the level of the other factor, indicates that the two types of variables (sex chromosome genes, gonadal hormones) have convergent actions on gene pathways influencing emergent phenotypes including the disease [45]. The molecular nature of this interaction is unstudied to date.

Our attempt here has been to provide the basis for a logical dissection of sex chromosome effects on any phenotype in mouse. The decision tree that summarizes this approach is shown in Fig. 4. Because the difference in the XX and XY genomes is fundamentally a difference in copy number of large segments of X and Y DNA, it makes sense to start by varying the number of these segments in combination or in isolation, to narrow down the segments that make XX and XY mice different. A progressive narrowing of choices can ultimately lead to the discovery of new genetic factors that make males and females different. The approaches suggested here have the advantage that they mimic the natural differences between XX and XY tissues. Other methods to search for X chromosome effects, for example, linkage mapping for specific traits, are not equivalent and may not detect differences caused by one vs. two X chromosomes. For example, study populations may not have any variation in the X genes that contribute to direct SCEs or variation in gene sequence may not mimic effects of X chromosome number. Similarly, studying the effects of different strain origin of the Y chromosome (e.g., by comparing consomic Y strains) may uncover effects of Y chromosome variation but may not mimic the effects of the presence vs. absence of the Y chromosome as it relates to sex differences in phenotype.

The approaches suggested here have already demonstrated that both X and Y segments cause differences in phenotype [11, 50]. Specific Y genes have been identified that control specific aspects of spermatogenesis [73, 102, 104, 108, 109], but to date, no specific X gene has been identified that causes a sex chromosome effect. Because of the advent of more efficient gene targeting methods, we expect this situation to change soon. It will be interesting to discover which X imprinted or escapee genes cause differences in physiology and disease, and how these and Y genes interact with gonadal hormones to cause emergent sex differences in phenotype.