We propose that the phenomenon of X-chromosome inactivation in females may constitute a risk factor for loss of T-cell tolerance; specifically that skewed X-chromosome inactivation in the thymus may lead to inadequate thymic deletion. Using a DNA methylation assay, we have examined the X-chromosome inactivation patterns in peripheral blood from normal females (n = 30), female patients with a variety of autoimmune diseases (n = 167). No differences between patients and controls were observed. However, locally skewed X-chromsome inactivation may exist in the thymus, and therefore the underlying hypothesis remains to be disproved.
A reduction in the sex ratio (male : female) is characteristic of most autoimmune disorders. The increased prevalence in females ranges from a modest 2:1 for multiple sclerosis , to approximately 10:1 for systemic lupus erythematosus . This tendency toward autoimmunity in females is often ascribed to hormonal differences, because in a number of experimental disease models estrogens exacerbated disease, and androgens can inhibit disease activity [3,4]. However, human studies have failed to demonstrate a clear-cut influence of hormonal environment on disease susceptibility to lupus or other autoimmune disorders. In addition, many childhood forms of autoimmunity, such as juvenile rheumatoid arthritis, exhibit female predominance . Interestingly, juvenile (type 1) diabetes is an exception to this general trend, with a sex ratio close to 1 in most studies . Therefore, it is reasonable to consider alternative explanations for the increased prevalence of autoimmune diseases in human females.
A unifying feature of autoimmune disorders appears to be the loss of immunologic tolerance to self-antigens, and in many of these diseases there is evidence that T-cell tolerance has been broken. The most profound form of T-cell tolerance involves deletion of potentially self-reactive T cells during thymic selection. Thus, lack of exposure to a self-antigen in the thymus may lead to the presence of autoreactive T cells and may increase the risk of autoimmunity. An elegant example of this has recently been reported .
The existence of X-chromosome inactivation in females offers a potential mechanism whereby X-linked self-antigens may escape presentation in the thymus or in other peripheral sites that are involved in tolerance induction. Early in female development, one of the two X chromosomes in each cell undergoes an ordered process of inactivation, with subsequent silencing of most genes on the inactive X chromosome . This phenomenon occurs at a very early embryonic stage , and thus all females are mosaic and may occasionally exhibit extreme skewing towards one or the other parental X chromosome. In theory, this may result in a situation in which polymorphic self-antigens on one X chromosome may fail to be expressed at sufficiently high levels in a tolerizing compartment, such as the thymus, and yet may be expressed at a considerable frequency in the peripheral soma. Thus, females may be predisposed to a situation in which they can occasionally express X-linked autoantigens in the periphery to which they have been inefficiently tolerized. Stewart  has recently speculated that such a mechanism may play a role in the predisposition to systemic lupus.
This hypothesis predicts that females with autoimmunity may be particularly prone to this mechanism of `inadequate tolerization' by virtue of extremely skewed X-chromosome inactivation. We therefore performed a comprehensive analysis of X-chromosome inactivation patterns in populations of females with multiple sclerosis, systemic lupus erythematosus, juvenile rheumatoid arthritis, and type 1 (insulin-dependent) diabetes mellitus, and in female control individuals. The results do not provide support for a major role for skewed X-chromosome inactivation in female predisposition to autoimmunity; however, neither is the underlying hypothesis disproved by the present data.
Materials and method:
DNA was obtained from female patients from the following sources: 45 persons with juvenile diabetes seen at the Virginia Mason Research Center in Seattle, Washington; 58 multiple sclerosis patients seen at the New York Hospital Multiple Sclerosis Center; 46 patients with systemic lupus erythematosus seen at the Hospital for Special Surgery (New York); 18 patients with juvenile rheumatoid arthritis seen at the Children's Hospital Medical Center in Cleveland. In addition, 30 healthy age-matched females were studied as normal controls.
Employing a modification of previously described methods , we utilized a fluorescent Hpa II/PCR assay of the androgen receptor (AR) locus to assess X-chromosome inactivation patterns. The AR gene contains a polymorphic CAG repeat, which is flanked by Hpa II sites. These Hpa II sites are methylated on the inactive X chromosome, and are unmethylated on the active X chromosome. By performing PCR amplification across this region after cutting with the methylation-sensitive enzyme Hpa II, the relative amounts of the methylated AR alleles can be quantitatively determined with a high degree of accuracy; variance on repeated assays is approximately 4% .
Skewing of X-chromosome inactivation is expressed as percentage deviation from equal (50:50) inactivation of the upper and lower AR alleles. Therefore, the maximal possible deviation is 50%, in which case all of the X chromosomes bearing one of the AR alleles are inactivated.
We examined X-chromosome inactivation patterns in several different populations. The results are summarized in Fig. 1. A wide range of X-inactivation skewing was observed in all five groups. Approximately 5% (nine out of 197) of individuals exhibited extreme skewing (greater than 40% deviation from a 50:50 distribution). However, there was no difference between the groups, either in the overall mean skewing, or in the fraction of individuals with extreme skewing (>40%).
Although the present study was not initiated in order to examine allelic variation in the AR gene per se, the data provide an opportunity to address this question. Excessively long CAG repeats in the AR are a rare cause of spinal-bulbar muscular atrophy , and AR repeat length appears to have an influence on the biology of certain tumors [14,15]. In this context, it has been shown that transcription of AR correlates inversely with repeat length . We therefore compared AR repeat length in control individuals and patients with autoimmunity. No differences were observed for mean repeat length, or for maximum and minimum repeat length, among the five groups.
The reason for the female predominance in most autoimmune diseases remains obscure. The present study was initiated in order to address the hypothesis that a nonhormonal mechanism related to X inactivation might be involved. The hypothesis rests on the idea that skewing of X inactivation might lead to a deficiency of tolerance induction in the thymus, particularly with respect to polymorphic X-linked autoantigens. The hypothesis predicts that skewed X inactivation would be more prevalent in females with autoimmune diseases than in female control individuals. This was not observed.
Nevertheless, these negative data do not rule out a role for X inactivation in female predisposition to loss of tolerance. A general model for how this mechanism might operate is shown in Fig. 2. Thymocytes undergo selection in the thymic parenchyma and, in the case of negative selection, the selecting elements appear to be derived from the bone marrow and consist mainly of thymic dendritic cells. If the thymic dendritic cell population exhibits random X inactivation, it is highly likely that differentiating thymocytes will contact dendritic cells that express self-antigens on both X chromosomes. This situation is outlined schematically on the left side of Fig. 2. However, if there is extremely skewed X inactivation in the thymic dendritic cell population, a particular thymocyte might not come into contact with dendritic cells that express one of the two X chormosomes. This would lead to a situation where T cells may undergo thymic maturation without having been negatively selected for antigens that are expressed on the predominantly inactive X chromosome. This situation is shown on the right side of Fig. 2.
In order for this mechanism to be physiologically relevant, some assumptions must be made. First, defective tolerance from skewed X inactivation should only be directed at X-linked antigens that are polymorphic, and for which the individual is heterozygous. Thus, this mechanism would not be expected to lead to lack of tolerance commonly, unless there are at least several highly polymorphic X-linked autoantigens in the population that are involved in thymic deletion events. Second, if this actually leads to autoimmunity, it also predicts that the initial break in tolerance that leads to disease should involve an X-linked autoantigen that is expressed in a peripheral nontolerizing site or circumstance.
A recent report  has elegantly demonstrated the importance of thymic deletion events in predisposition to autoimmune disease. The proteolipid protein (PLP) autoantigen is expressed in alternatively spliced forms, which exhibit tissue specific expression. A nonspliced variant is expressed in peripheral neural tissue. However, in the thymus a splice variant results in the lack of thymic expression of an immunodominant peptide. This results in loss of tolerace of T cells to this peptide, presumably on the basis of lack of thymic deletion of thymocytes that are reactive with this antigen. Interestingly, PLP is encoded on the X chromsome. However, there is no evidence that genetic polymorphisms control the level splicing of PLP within the thymus. Nevertheless, these data illustrate the potential importance of deficiencies in thymic deletion for autoimmune T-cell reactivity.
The present results suggest that if skewed X inactivation is relevant to thymic tolerance induction, then the effect does not depend on global skewing of X-chromosome inactivation, at least in the hematopoietic compartment. In this study we examined X-inactivation patterns in peripheral blood mononuclear cells, and the results should reflect the state of X inactivation in all mesenchymal tissues, including dendritic cells. X inactivation occurs at a very early time point in development, and thus the results in one tissue should reflect the general situation in the rest of the body. However, there may be exceptions to this. We have occasionally observed differences in X-inactivation patterns between buccal mucosa (an ectodermally derived tissue) and peripheral blood in the same individiual (unpublished observations). This could be a chance event, or it may result from selection for certain X-linked alleles during embryonic development, as has been described in carriers of X-linked immunodeficiencies .
Another consideration is that certain tissue microenvironments may be derived from very small numbers of founder cells, and thus may exhibit skewed utilization of one or the other X chromosome, even if the tissue as a whole is not skewed. This situation could vary over time. Thus, there may be time points at which certain thymic microenvironments are populated by dendritic cells that, for stochastic reasons, all utilize the same X chromosome. This would create a `window of opportunity' in which a given thymocyte, in a given selecting location, could escape negative selection by antigens on the inactive X chromosome. The likelihood of this happening would obviously depend on the number of dendritic cells that are usually contacted by a thymocyte during thymic selection. There is limited information on this point, although Stewart  has theorized that this number may be as low as 15. If this is the case, then escape from thymic deletion may still occur in females who are heterozygous for a relevant X-linked antigen, even if the hematopoietic cells in general do not exhibit extreme skewing.
In conclusion, we suggest that X-chromosome inactivation needs to be considered as a potential factor in the predominance of females in most autoimmune diseases. Our inability to show an increase in X-chromosome skewing in females with autoimmunity does not eliminate this as an etiologic contributor to loss of immunologic tolerance. Future experiments must be directed at a detailed analysis of tissue patterns of X inactivation, as well as at a search for potential X-linked autoantigens.