Background
Approximately one third of autism spectrum disorder (ASD) patients have an identifiable pathogenic genetic variant, and as sequencing becomes more widespread, this percentage increases [
1]. Although mouse models of ASD-associated mutations have successfully reconstituted some ASD phenotypes, they are limited by constraints in recapitulating complex human behaviors, by the tendency for replicable phenotypes to be most robust in homozygous knockouts, which exaggerate deficiency states, and by the inability to recapitulate background genetic factors that influence autistic syndromes in individual human patients [
2]. Induced pluripotent stem cells, which are derived from reprogrammed somatic cells, offer researchers studying autism a reliable method to obtain genetically identical neurons from noninvasively retrieved patient samples. A small number of recent papers have begun to explore cellular phenotypes of autism observed in human induced pluripotent stem cell (hiPSC)-derived neurons [
2‐
6]. HiPSC models not only offer researchers a method to elucidate the molecular biology of ASD, but they also provide a means of identifying and testing drug therapies.
An important aspect of any cellular model of disease involves the issue of whether the cells faithfully reflect the normative process of X chromosome inactivation. Although only a minority of ASD susceptibility loci reside on the X chromosome, X-linked mutations are well-known causes of related developmental disorders. Perhaps even more importantly, there is growing evidence that supports the hypothesis that there is a “female protective effect” in which the pronounced sex ratio universally observed in ASD is resolvable to the interaction between female sex and inherited autism liabilities, which are predominantly polygenic and autosomal in nature [
7‐
9]. A proposed mechanism for this phenomenon is that genetic loci expressed on the noninactivated X chromosome modulate the phenotypic expression of autosomal loci conferring risk for ASD, which would make faithful recapitulation of X-inactivation critical to the development of valid cellular models of the condition [
7]. Here, we review what is known about this important detail of hiPSC-derived cellular models, based upon the current literature.
Random X chromosome inactivation in female humans occurs early during the developmental process, resulting in the equalization of X chromosome-linked gene dosages across sexes. There are four steps associated with X chromosome inactivation (XCI), referred to as counting, choosing, initiation, and maintenance. All but maintenance are orchestrated by the X-inactivation center (Xic), which contains the noncoding RNA-encoding X-inactive specific transcript (Xist) locus. Xist RNA surrounds the X chromosome selected to be inactivated and is necessary for the initiation of its silencing. Although the exact mechanism involved in counting remains elusive, there is evidence to suggest that the cell utilizes the ratio between X chromosome number and haploid autosomal chromosome number in order to ensure that only one X chromosome is active [
10]. This process is so precise that only one X chromosome is active even in instances where individuals have more than two X chromosomes [
11]. Xist expression is crucial for initiation of silencing in developing embryos. However, once cells have differentiated, Xist is unable to induce inactivation [
12]. Xist forms a “cloud” around the inactivated X chromosome by binding areas along the length of the chromosome, which are located near regions where Xist is actively transcribed [
13]. This results in the loss of RNA polymerase II, initiation factors, and other components necessary for transcription in the areas covered by Xist [
12]. However, there are regions on the inactivated X chromosome that sustain active transcription and “escape” inactivation by looping out of the chromosomal territory covered by Xist [
13].
Early in development, when Xist plays the major role in inactivation, XCI is reversible. Once the initiation phase transitions to the maintenance phase, however, the inactivation status is sustained [
12]. This progression is characterized by alteration in DNA methylation and histone modification. The Polycomb group (PcG) complex, a protein complex that mediates repressive histone 3 lysine 27 trimethylation (H3K27me3 ) to silence transcription, drives some of these epigenetic changes [
12].
Generally, X-inactivation is a random process that results in equal expression of both X chromosomes in tissue [
10]. However, under certain circumstances, one X chromosome is preferentially inactivated, which may occur when one of the X chromosomes actively expresses a severely detrimental allele [
14]; such “skewing” of XCI, however, has not been consistently observed in samples of patients with ASD [
15]. There are two methods that result in nonrandom expression of an X chromosome. The first method arises when the cells preferentially inactivate one X chromosome over the other. The second method occurs when cells expressing a particular X chromosome are selected against [
14].
The extent to which these processes are recapitulated in the development of human-induced pluripotent stem cell-derived neuronal cell models is currently poorly understood and, while many papers discuss the modeling of neurodevelopmental disorders in hiPSCs, the literature is scarce regarding the status of XCI in hiPSCs and the potential impact on neurodevelopmental disease modeling [
16‐
18]. In this review, we attempt to provide a concise summary of the literature regarding XCI in hiPSCs while addressing how discrepancies in the literature impact ASD hiPSC modeling.
Conclusions
Human iPSCs represent an ethically acceptable, increasingly widely available resource for developing cellular models of disease. These models are particularly important in autism research, given the relative lack of availability of the affected primary tissue (brain) for biological studies, marked heterogeneity in the genetic influences on ASD across affected individuals (making each individual representative of a rare/unique combination of deleterious and background genetic factors), the prevalence of the condition, and the availability of effective technologies for reprogramming noninvasively derived cells from individual patients [
9]. By recapitulating the unique phenotypes of specific patients, hiPSCs have the capacity to reveal cell-autonomous mechanisms influencing ASD and related disorders and to provide a platform for assessing the effects of correcting particular genetic susceptibilities.
To date, there have been few published results of hiPSC models of nonsyndromic ASD [
4,
36]. Griesi-Oliveira et al. did study the effects of a mutation in MeCP2, a gene found on the X chromosome, on TRPC6, a gene believed to be associated with a nonsyndromic version of ASD. Since, in this case, the loss-of-function mutation was on the X chromosome and was severe enough to influence it to be inactivated, the status of XCR was not particularly relevant to the study [
4]. While there are additional published reports involving hiPSC models of Rett syndrome, there are, again, many inconsistencies among groups regarding the status of XCR [
22‐
24].
We conclude that variabilities in the status of X-inactivation and X-reactivation in hiPSCs are significant factors to be considered when implementing cellular models of autism and related neurodevelopmental disorders. The multiple variations in XCI status of clonal hiPSC populations include complete reactivation, complete inactivation, mixed inactivation and reactivation, and erosion. These complexities indicate that variation in reprogramming procedures and/or failure to precisely characterize XCI status can have significant consequences for the validity of a cellular model [
27]. This is true for conventional technologies for reprogramming and should be clarified whenever new or emerging methodologies for transformation of somatic cells (e.g., Richner et al.) are implemented in female cellular models [
37]. Careful evaluation of the extent to which X-inactivation has occurred, as well as attention to which X chromosome is inactivated in a given cell population, using available approaches involving immunocytochemistry, RNA FISH, PCR, and sequencing, are important aspects of the design and execution of hiPSC-based cellular models of disease [
22,
24,
34]. It is always important to recognize that there may be other factors influencing XCI in the development of actual tissue, such as parent of origin influences, that further complicate faithfully recapitulating a human phenotype in vitro [
38,
39].
The discrepancies in XCR status in human hiPSCs reviewed here underscore the need for a greater understanding of how the parameters involved in reprogramming and culturing of hiPSCs affect their pluripotent cell state, differentiation potential, and other phenotypic properties relevant to their use for modeling human disease. While it may yet take years to fully define the molecular basis of reprogramming and factors that influence acquisition and maintenance of particular pluripotent states, it is important to identify specific conditions under which hiPSCs with stable XCI status can be obtained and maintained. This will provide a reliable and standardized method for researchers to generate hiPSC models that can more faithfully recapitulate the state of the cells in vivo and that therefore have higher utility for cellular modeling of human disease states.
Abbreviations
AR, androgen receptor; ASD, autism spectrum disorder; ESCs, embryonic stem cells; FISH, fluorescence in situ hybridization; H3K27me3, histone H3 trimethyl lysine 27; hiPSCs, human induced pluripotent stem cells; iPSCs, induced pluripotent stem cells; LIF, leukemia inhibitory factor; MEF, mouse embryonic fibroblast; mESCs, mouse embryonic stem cells; miPSCs, mouse induced pluripotent stem cells; PcG, polycomb group complex; PCR, polymerase chain reaction; XCI, X chromosome inactivation; XCR, X chromosome reactivation; Xic, X-inactivation center.