Abstract

   Monosomy X (45,X) is associated with Turner syndrome and pregnancy loss
   in humans, but the underlying mechanisms remain unclear. We therefore
   undertook an exploratory study of the transcriptomic landscape of
   clinically relevant human fetal 45,X tissues (including pancreas,
   liver, kidney, skin, placenta) with matched 46,XX and 46,XY control
   samples between 11 and 15 weeks post conception (n = 78). Although most
   pseudoautosomal region 1 (PAR1) genes are lower in monosomy X tissues,
   we also found reduced expression of several key genes escaping X
   inactivation (e.g., KDM5C and KDM6A), several ancestral X-Y gene pairs,
   and potentially clinically important transcripts such as genes
   implicated in ascending aortic aneurysm. In contrast, higher expression
   of an autosomal, long non-coding RNA (OVCH1-AS1) is seen in all 45,X
   tissues. In the placenta, lower expression of CSF2RA is demonstrated,
   likely contributing to immune dysregulation. Taken together, these
   findings provide insights into the biological consequences of a single
   X chromosome during early human development and potential insights in
   genetic mechanisms in Turner syndrome.

   graphic file with name 42003_2025_7699_Figa_HTML.jpg

   Subject terms: Embryology, Development, Gene expression,
   Transcriptomics, Molecular medicine
     __________________________________________________________________

   Transcriptomic analysis of monosomy X (45,X) tissues during early human
   fetal development provides insight into sex chromosome gene dosage and
   potential mechanisms and management in Turner syndrome.

Introduction

   Complete or partial loss of the second X (sex) chromosome in humans
   occurs in approximately 1:2500 girls and women and is associated with
   Turner syndrome (TS)^[60]1,[61]2. Around 50% of individuals with TS
   have a monosomy X (45,X) karyotype but other variations in karyotype
   are often seen (e.g., isochromosome Xq (46,X,i(Xq), ring X (mosaic
   45,X/46,X,r(X)), Xp or Xq deletions, and 45,X/46,XX or 45,X/46,XY
   mosaicism)^[62]1–[63]4. Although monosomy X is the only chromosome
   monosomy compatible with survival in humans, it is estimated that only
   a proportion of monosomy X fetuses survive to term and many pregnancies
   are spontaneously lost in the first or early second trimester, often
   before a pregnancy is recognized^[64]5–[65]7.

   Girls and young women with TS/monosomy X can present with many
   different features and at different ages. In the newborn period, the
   diagnosis may be suspected due to lymphedema, congenital cardiovascular
   anomalies (e.g., coarctation of the aorta), renal features (e.g.,
   horseshoe kidney) or distinct physical signs (e.g., wide chest, widened
   neck skin)^[66]8. Transient hyperinsulinism and hypoglycemia have also
   been reported^[67]9,[68]10. In childhood, early features include
   impaired growth and recurrent otitis media, whereas absent puberty and
   primary ovarian insufficiency (POI) present later in teenage
   years^[69]8,[70]11,[71]12. Higher risks of long-term co-morbidities in
   adulthood are described, such as diabetes mellitus, weight gain,
   hypertension, raised liver enzymes, hearing impairment, hypothyroidism,
   autoimmunity, acquired cardiovascular disease, and skin
   nevi^[72]12–[73]16. Women with TS have an overall increased mortality
   partly accounted for by aortic root dilatation and
   dissection^[74]13,[75]16–[76]18. Thus, the clinical features associated
   with TS can affect many different systems and may have some origins in
   early fetal development^[77]2. Identifying underlying mechanisms that
   drive the clinical features of TS is important as they may help to
   develop personalized medicine strategies and improve care for girls and
   women in the long term^[78]1,[79]2,[80]19.

   Key genetic mechanisms hypothesized to drive TS phenotypes are usually
   related to the complete or partial loss of the second sex
   chromosome^[81]1,[82]2,[83]20,[84]21. Haploinsufficiency of genes in
   the pseudoautosomal (PAR) regions (PAR1, PAR2), which have Y chromosome
   homologs, have been associated with TS phenotypes such as short stature
   and cubitus valgus (e.g., SHOX) or cardiac QTc interval
   duration^[85]2,[86]22–[87]24. Monosomy X may also be associated with
   reduced dosage of genes that normally escape
   X-inactivation^[88]2,[89]20,[90]21,[91]25,[92]26. X inactivation is a
   process whereby one X chromosome is transcriptionally silenced in
   female mammalian cells to equalize dosage of gene products from the X
   chromosome between 46,XX and 46,XY individuals^[93]7,[94]25,[95]27.
   This process is mediated largely by the non-coding RNA transcripts,
   XIST/TSIX^[96]25,[97]27. It is well established that some genes on the
   X chromosome escape X inactivation; these genes are normally
   biallelically expressed from both X chromosomes in 46,XX females, but
   this may not occur in those with a 45,X
   karyotype^[98]20,[99]21,[100]25,[101]28,[102]29. Moreover, a core
   subset of X chromosome genes that might drive somatic sex differences
   has recently been proposed^[103]27.

   In addition to direct effects of reduced PAR gene dosage, other
   mechanisms linked to the pathogenesis of TS include disruption of X
   chromosome genes that have a knock-on (or “ripple”) effect on other
   parts of the X chromosome itself^[104]2,[105]20,[106]21, including
   important non-coding RNAs such as JPX (a key regulator of the X
   inactivation gene, XIST)^[107]20,[108]26,[109]30, or X chromosome genes
   that influence autosomal genes with regulatory functions such as
   ubiquitination, chromatin modification, translation, splicing, DNA
   methylation and circular RNA
   generation^[110]2,[111]20,[112]21,[113]26,[114]28–[115]32. Unravelling
   these complex interactions requires whole genome transcriptomic
   analysis at a suitable scale and in relevant tissues.

   To date, most transcriptomic studies investigating the pathogenic basis
   of TS have analyzed blood leukocytes/peripheral blood mononuclear cells
   (PBMCs)^[116]21,[117]26,[118]29. Direct sampling of tissues strongly
   associated with phenotypic features such as diabetes, hypertension and
   obesity is much more challenging. Recently, transcriptomic analysis of
   adult fat and muscle biopsies has been reported from individuals with
   TS and related sex chromosome aneuploidies such as 47,XXY (Klinefelter
   syndrome)^[119]20,[120]32. This approach is starting to provide insight
   into the effects of sex chromosomes in different tissues, and to
   identify core haploinsufficient X chromosome genes associated with a
   45,X karyotype^[121]20,[122]33,[123]34. However, the biochemical and
   transcriptomic profile of adult tissues may also be influenced by
   confounding factors such as medication, inflammation, diet or the
   complex interplay between different systems (e.g., fat, muscle and
   pancreas in insulin sensitivity), so assessing the “pure” monosomy X
   transcriptome is difficult.

   Given the fact that many clinical features associated with monosomy X
   are present in early postnatal life, and that the origins of many
   long-term adult conditions may be in part established during embryonic
   or fetal development^[124]35,[125]36, we undertook detailed
   transcriptomic analysis of monosomy X fetal samples from several key
   tissues of interest between 11-15 weeks post conception (wpc). Although
   we were limited by sample availability to some extent, our aim was to
   better understand the transcriptomic events associated with monosomy X
   in early human development, and to obtain a unique perspective on human
   X chromosome biology and possible disease mechanisms in Turner
   syndrome, and to investigate the effects of monosomy X in the placenta
   in relation to pregnancy loss.

Results

Global transcriptomic differences and key sex chromosome genes

   In order to identify global transcriptomic differences associated with
   a single X chromosome during development, we performed bulk RNA
   sequencing (bulk RNA-seq) using human monosomy X fetal samples (n = 20)
   between 11 and 15 wpc and compared these to tissue- and age-matched
   46,XX (n = 20) and 46,XY controls (n = 20) (Fig. [126]1a, Supplementary
   Fig. [127]1 and Supplementary Data [128]1). Pancreas, liver, kidney,
   skin, and a mixed sample group (comprised of brain, heart, lung and
   spleen) were chosen, as these are biologically relevant to the clinical
   features associated with TS in childhood or in later life. All 60
   tissue samples underwent single nucleotide polymorphism (SNP) array
   analysis on simultaneously extracted DNA. This approach confirmed the
   expected karyotype in all cases, except for one 45,X fetus where
   low-level mosaicism for a 46,XY cell line was seen (Supplementary
   Fig. [129]2).

Fig. 1. Experimental design and principal component analysis of tissues.

   [130]Fig. 1
   [131]Open in a new tab

   a Overview of age stages and tissues used for the main study.
   Karyotypes are indicated by the key. One 13 week post conception (wpc)
   pancreas was bisected longitudinally and both parts processed
   independently. b Principal component analysis (PCA) of the whole
   multi-tissue dataset (n = 60). PC, principal component. c PCA of
   pancreas samples (n = 12) (Note: the two 45,X samples at 13 wpc showed
   biological variability). d PCA of liver samples (n = 12). e PCA of
   kidney samples (n = 11; one outlier removed due to low level ( < 10%)
   adrenal contamination). f PCA of skin samples (n = 11; one outlier
   removed due to low level ( < 10%) muscle contamination). Age and
   karyotypes of tissues are shown in the key. g Heat map of key sex
   chromosome genes for each set of tissue samples; XIST for the X
   chromosome and DDX3Y, KDM5D, RPS4Y1 and USP9Y for the Y chromosome.
   Expression intensity is shown in the key.

   Using principal component analysis (PCA) for the entire dataset
   (n = 60), samples clustered together based primarily on tissue of
   origin rather than karyotype, as expected (Fig. [132]1b, Supplementary
   Fig. [133]3 a, b). In contrast, PCA of individual tissues showed an
   influence of karyotype on clustering for pancreas (Fig. [134]1c) and
   liver (Fig. [135]1d), some effect of karyotype in kidney, and limited
   effect in skin (Fig.[136]1 e, f and Supplementary Fig. [137]3 c, d).

   Heatmap normalized expression of several important sex chromosome genes
   is shown in Fig. [138]1g. XIST, the principle X chromosome regulator of
   X inactivation, was strongly expressed in all 46,XX tissues, consistent
   with the presence of two chromosomes. XIST was not expressed in 46,XY
   samples nor in 45,X samples, both of which have a single X chromosome
   and do not undergo X inactivation. Y chromosome genes (e.g., DDX3Y,
   KDM5D, RPS4Y1, USP9Y) were strongly expressed in 46,XY samples and not
   in 46,XX samples nor 45,X samples, except at low level in the two
   mosaic samples (pancreas, heart/mixed) outlined above (Fig. [139]1g).

   In order to identify genes with consistently lower or higher expression
   in monosomy X tissues compared to controls, differential gene
   expression (DGE) analysis was undertaken. Volcano plots comparing 45,X
   samples (n = 4) with either 46,XX or 46,XY matched control samples
   (n = 4 each group) for each tissue are shown in Fig. [140]2
   (Supplementary Data [141]2–[142]25). Genes with higher expression in
   45,X samples have a positive log[2] fold change (log[2]FC), whereas
   genes with a higher expression in either 46,XX or 46,XY samples have a
   negative log[2]FC. As expected, the X inactivation genes XIST and TSIX
   showed higher expression in 46,XX samples compared to 45,X samples,
   whereas Y chromosome genes were differentially expressed in 46,XY
   samples compared to 45,X (Fig. [143]2).

Fig. 2. Volcano plots showing differential expression of genes between the
45,X and either 46,XX or 46,XY matched control samples.

   [144]Fig. 2
   [145]Open in a new tab

   Data are shown for: (a) Pancreas; (b) Liver; (c) Kidney; (d) Skin; (e)
   Mixed group (brain, heart, lung, spleen). Four samples were included in
   each group (see experimental design Fig. [146]1a). Comparison of 45,X
   versus 46,XX is shown in the left-hand panel and comparison of 45,X
   versus 46,XY is shown on the right-hand panel. The top ten most
   differentially expressed genes in each dataset are labeled, based on
   adjusted p-value and where log[2] fold change (FC) is greater than
   +/−0.5. Genes with higher expression in 45,X tissues have a positive
   log[2]FC and those with higher expression in control samples have a
   negative log[2]FC. The significance level of highlighted points is
   shown in the key.

Genes with lower expression in monosomy X tissues

   Our initial analysis focused on genes that were lower in both 45,X
   versus 46,XX and 45,X versus 46,XY datasets, and consistent across each
   of the five different tissues studied (n = 10 datasets)
   (log[2]FC < -0.5, adjusted p-value (p-adj) < 0.05; Supplementary
   Fig. [147]4). This relatively low threshold for log[2]FC was chosen to
   identify subtle but potentially meaningful differences in gene dosage,
   especially related to haploinsufficiency effects in monosomy X, and
   within the context of small sample sizes. By intersecting these groups,
   no genes were shared among all tissues studied with this cut-off,
   although four genes (SLC25A6, AKAP17A, GTPBP6, ZBED1) did have lower
   monosomy X expression in multiple tissues (Fig. [148]3a–c and
   Supplementary Fig. [149]4). These genes showed haploinsufficiency of
   expression in monosomy X tissues in violin plots (Fig. [150]3b and
   Supplementary Fig. [151]5). As these genes are all located in the PAR1
   region of the X chromosome (Fig. [152]3d), investigation of the
   expression of all PAR genes was undertaken for 45,X versus 46,XX or
   46,XY tissues (Fig. [153]3e). This analysis showed consistently lower
   expression of many PAR1 genes in 45,X tissues compared to controls,
   where log[2]FC = -1.0 represents haploinsufficiency in 45,X, and where
   log[2]FC = 0 represents similar expression in control and 45,X samples
   (Fig. [154]3e). Several PAR1 genes did not show differential
   expression; these genes generally had low transcript counts, or were
   not detected in all tissues (e.g., SHOX, CRLF2, P2RY8, ASMT, XG)
   (Supplementary Fig. [155]5 and Supplementary Data [156]26 and [157]28).
   Of note, the three PAR2 region genes (SPRY3, VAMP7, IL9R) also did not
   show marked differential expression (Fig.[158]3e, Supplementary
   Fig. [159]6, and Supplementary Data [160]27 and [161]28), as shown in
   samples from adults with monosomy X^[162]20,[163]21,[164]32.

Fig. 3. Genes with lower expression in monosomy X when compared to 46,XX or
46,XY controls.

   [165]Fig. 3
   [166]Open in a new tab

   a Five-way Venn diagram showing overlap of genes with lower expression
   in monosomy X when compared to control karyotypes across all tissues. b
   Graph to show genes with lower expression in monosomy X across all
   tissues with inset of violin plot showing normalized count data of
   SLC25A6 across all tissues (from panel a) according to karyotype (n = 4
   in each group). Pseudoautosomal region (PAR) genes are shown in bold
   and dark bars. c Heatmap of SLC25A6 expression across all tissues used
   in study. d Schematic of the X chromosome and showing PAR1 and PAR2
   genes. e Differential expression of genes located on the PAR regions
   for 45,X compared to 46,XX or 46,XY samples. Individual mean data
   points for each tissue group are shown, as indicated in the legend. The
   bars represent the mean of the different tissue groups with standard
   error of the mean shown. Log[2] fold change (FC) = −1.0 represents half
   the expression in 45,X samples (i.e., haploinsufficiency), whereas
   log[2]FC = 0 represents similar expression in 45,X samples and 46,XX
   and 46,XY controls (n = 4 for each karyotype in each tissue group).

   In addition to PAR genes, recent reports have identified core sets of
   genes that likely escape X inactivation or influence phenotype in
   monosomy X adult tissues^[167]20, as well as ten X chromosome genes
   that are implicated as key mediators of biological sex
   differences^[168]27. Analysis of these genes in monosomy X tissues
   during development compared to controls is shown in Fig. [169]4a–c, and
   Supplementary Data [170]29, [171]30. Notably, the histone demethylase
   genes, KDM5C (also known as JARID1C/SMCX) and KDM6A (UTX) had lower
   expression in 45,X compared to 46,XX control tissues, likely due to X
   inactivation escape. As KDM5C and KDM6A are X-chromosome genes with
   Y-chromosome homologs (KDM5D and UTY, respectively), we extended our
   analysis to look at differential expression of homologous, ancestral
   X-Y gene pairs (as defined by Godfrey et al. 2020)^[172]37
   (Fig. [173]4d, e and Supplementary Data [174]31–[175]33). Most X-Y gene
   pairs showed a decrease in both 45,X versus 46,XX and in 45,X versus
   46,XY datasets, with the most consistent and marked lower expression in
   monosomy X for EIF1AX/EIF1AY, ZFX/ZFY, DDX3X/DDX3Y, KDM6A/UTY (also
   known as KDM6C), KDM5C/KDM5D (also known as JARID1D/SMCY), and
   RSP4X/RSP4Y1. Many of these genes play a role in transcription,
   translation and histone methylation status, which may be altered in
   45,X and could influence developmental processes.

Fig. 4. Additional key genes proposed to have lower expression in monosomy X,
including X-Y homolog pairs.

   [176]Fig. 4
   [177]Open in a new tab

   a Schematic showing the genomic location of key genes of interest on
   the X chromosome (see Fig. 4b, c). PAR genes are not shown. b
   Differential expression in our datasets of selected “core” X chromosome
   genes that have been linked to monosomy X, including those proposed to
   show X chromosome dosage (from Viuff et al. 2023). Individual mean data
   points for each tissue group are shown, as indicated in the legend. The
   bars represent the mean of these different tissue groups with standard
   error of the mean shown. Log[2] fold change (FC) = −1.0 represents half
   the expression in 45,X samples (i.e., haploinsufficiency), whereas
   log[2]FC = 0 represents similar expression in 45,X samples and 46,XX
   and 46,XY controls (n = 4 for each karyotype in each tissue group).
   Note XIST and TSIX show strong differential expression in all datasets
   but are omitted from the graphic. c Differential expression in our
   datasets of selected X chromosome genes linked to “sex differences”
   (from San Roman et al.)^[178]27. Data are presented as described above.
   Note the different y-axis scale compared to Fig. 4b. d Differential
   expression in our datasets of key “ancestral X-Y homolog” gene pairs
   (from Godfrey et al. 2020)^[179]37. This panel shows the X chromosome
   gene of the pair, with the corresponding Y chromosome homolog gene
   below it (Fig. 4e). Data are presented as described above. e
   Differential expression in our datasets of key “ancestral X-Y homolog”
   gene pairs (from Godfrey et al. 2020)^[180]37. This panel shows the Y
   chromosome gene of the pair, with the corresponding X chromosome
   homolog gene above it (Fig. 4d).

   Next, we analyzed genes that showed global lower expression in monosomy
   X across multiple tissues (total n = 10; of which n = 5 were
   45,X < 46,XX and n = 5 were 45,X < 46,XY). We hypothesized that these
   genes could also have important biological functions, beyond the
   PAR/XCI model. In addition to the PAR genes and XCI escape genes
   described above, several notable autosomal genes (e.g., ALDH1A3, RELN,
   LDLR, DKK2) and other X chromosome genes (e.g., AGTR2) emerged as
   having lower expression in 45,X tissues (mean log[2]FC below −0.4 for
   45,X compared to controls shown in Fig. [181]5a and Supplementary
   Data [182]34). Furthermore, when a log[2]FC below −0.35 cut off was
   applied to identify key genes for disease enrichment pathway analysis
   (n = 74 genes), biological functions emerged related to dissecting
   aneurysm, aneurysm of the ascending thoracic aorta and connective
   tissue disorders (Fig. [183]5b). These functions resulted largely from
   lower expression of core connective tissue genes in monosomy X (i.e.,
   FBN1, COL5A1, COL3A1, COL1A1, MMP2, COL5A2, COL1A2) (Fig. [184]5c). The
   mean differences in gene dosage were subtle, and not all values were
   significant when adjusted for multiple comparisons, but more marked
   differences were seen in the skin where collagen genes are innately
   expressed (Fig. [185]5c). We hypothesize, therefore, that lower
   expression of these genes could confer a risk for the development and
   progression of aortic root dilatation and aortic dissection in women
   with TS, especially as loss of FBN1 is found in Marfan Syndrome where
   thoracic aortic aneurysm can be a key feature^[186]38.

Fig. 5. Additional genes that have lower expression in monosomy X samples and
related pathways.

   [187]Fig. 5
   [188]Open in a new tab

   a Differential expression data of other non-PAR X chromosome and
   autosome genes that are lower in 45,X compared to 46,XX or 46,XY
   samples. Individual mean data points for each tissue group are shown,
   as indicated in the legend. The bars represent the mean of these
   different tissue groups with standard error of the mean shown. Log[2]
   fold change (FC) = −1.0 represents half the expression in 45,X samples
   (i.e., haploinsufficiency), whereas log[2]FC = 0 represents similar
   expression in 45,X samples and 46,XX and 46,XY controls (n = 4 for each
   karyotype in each tissue group). Data are ordered according to the mean
   of the 45,X versus 46,XX and 45,X versus 46,XY datasets. b Dot plot
   showing disease enrichment analysis for genes lower in 45,X compared to
   mean controls (n = 74), with log[2]FC cut-off below −0.35. c
   Differential expression in our dataset of the key genes contributing to
   the thoracic aortic aneurysm disease enrichment pathway. Data are
   presented as described above. See Fig. 5a for legend.

Genes with higher expression in monosomy X tissues

   Although lower expression of many X chromosome genes was expected in
   monosomy X tissues, especially in the PAR1 region, we also addressed
   whether any X chromosome or autosome genes show consistently higher
   expression in monosomy X tissues compared to matched controls.

   In order to achieve this, a tissue specific analysis was first
   performed to identify genes that were higher in both 45,X versus 46,XX
   and 45,X versus 46,XY datasets (log[2]FC > 0.5, p-adj ≤ 0.05;
   Supplementary Fig. [189]7). Next, the intersect of these genes across
   all five tissue groups was generated (Fig. [190]6a, b).

Fig. 6. Genes with higher expression in monosomy X when compared to 46,XX or
46,XY controls.

   [191]Fig. 6
   [192]Open in a new tab

   a Five-way Venn diagram of differentially expressed genes with higher
   expression in monosomy X tissues. b Graph to show genes with higher
   expression in 45,X with inset showing violin plot of normalized count
   data of OVCH1-AS1 across all tissues according to karyotype (n = 4 in
   each group). c Heat map of OVCH1-AS1 expression across all tissues used
   in the study. d Differential expression data of genes that are higher
   in 45,X compared to 46,XX or 46,XY samples. Individual data points for
   each tissue are shown, as indicated in the legend. The bar represents
   the mean of these data points with the standard error of the mean
   shown. Here, log[2] fold change (FC) = 1.0 represents twice the
   expression in 45,X tissues, whereas log[2]FC = 0 represents similar
   expression in 46,XX and 46,XY control and 45,X samples. (n = 4 for each
   karyotype in each tissue group). e Quantitative real-time PCR
   (polymerase chain reaction) showing relative expression of OVCH1-AS1 in
   fetal liver (n = 4 samples in each group). Data are presented as
   scatter dot plots with each datapoint being the mean of triplicate
   experiments. The mean and standard error of the mean of these
   datapoints is also shown. f Quantitative real-time PCR showing relative
   expression of OVCH1-AS1 in fetal kidney (n = 4 samples in each group).
   Data are presented as scatter dot plots with each datapoint being the
   mean of triplicate experiments. The mean and standard error of the mean
   of these datapoints is also shown. g Violin plots of OVCH1-AS1
   expression (normalized counts) in 45,X placenta compared to 46,XX and
   46,XY controls (n = 6 in each group). h Violin plots of OVCH1-AS1
   expression (normalized counts) in 45,X muscle compared to 46,XX and
   46,XY controls (n = 4 in each group). *, p-value < 0.05; **,
   p-value < 0.01; ns, not significant.

   Remarkably, only one gene was consistently identified across all
   tissues (Fig. [193]6a). This gene, OVCH1 Antisense RNA 1 (OVCH1-AS1)
   had higher expression in all monosomy X tissues, including the mixed
   group, when compared to 46,XX and 46,XY controls (Fig. [194]6b−d; see
   also Fig. [195]2, Supplementary Fig. [196]7, and Supplementary
   Data [197]35 and [198]36). These findings were confirmed using qRT-PCR
   (Fig. [199]6e, f), and in replication datasets of bulk RNA-Seq from
   placental tissue and muscle (Fig. [200]6g, h).

   OVCH1-AS1 is a long non-coding RNA (lncRNA) gene located on chromosome
   12 (UCSC, chromosome12:29,389,642-29,487,473, human GRCh38/hg38)
   (Fig. [201]7 a, b). OVCH1-AS1 has been identified as a potential
   antisense transcript of the protein coding gene, ovochymase 1 (OVCH1),
   and maps to a locus containing FAR2 (forward strand) and TMTC1 and
   ERGIC2 (reverse strand) (Fig. [202]7a).

Fig. 7. OVCH1-AS1 locus and conservation.

   [203]Fig. 7
   [204]Open in a new tab

   a Overview of chromosome 12 (p.11.22) and genes located in the region
   of OVCH1-AS1. Arrows indicate direction of transcription. The red arrow
   indicates OVCH1-AS1 (non-coding RNA gene) and blue arrows indicate
   protein coding genes. b Multiz alignments of species across the
   OVCH1-AS1 locus. A single line indicates no transcript in the aligned
   species. Double lines show aligning species having one or more
   unalignable transcripts in the gap region. c Relative expression of all
   chromosome 12 genes in 45,X tissues compared to 46,XX (orange) and
   46,XY (green). The higher expression of OVCH1-AS1 in monosomy X tissue
   is indicated as log2 fold change (FC) in relation to control tissues. d
   Violin plot of OVCH1 in the different tissues (normalized counts from
   bulk RNA-sequencing) (left panel) (n = 4 samples in each group).
   Quantitative real-time PCR (polymerase chain reaction) showing relative
   expression of OVCH1 in fetal kidney (right panel) (n = 4 samples each
   group). Data are presented as scatter dot plots with mean and standard
   error of the mean also shown. e Violin plot of ERGIC2 in different
   tissues (bulk RNA-seq) (left panel) (n = 4 samples in each group).
   Quantitative real-time PCR showing relative expression of ERGIC2 in
   fetal pancreas (right panel) (n = 4 samples each group). ns, not
   significant.

   Relatively little is known about OVCH1-AS1, but new studies are
   emerging^[205]39. In the GTEx database of adult tissue gene expression,
   OVCH1-AS1 generally has low level expression in adult tissues, as is
   often the case with lncRNAs (Supplementary Fig. [206]8). Conservancy
   analysis in vertebrates using Multiz alignments (UCSC genome browser)
   showed that OVCH1-AS1 is not highly conserved beyond higher primates;
   for example, it is not identified in mouse (Fig. [207]7b). Although a
   potential open reading frame exists between nucleotides 231-825 (203
   codons) using the Coding-Potential Assessment Tool (CPAT) (RNAcentral,
   transcript variant 1 (URS000075D789_9606; CPAT coding probability
   0.824, cut-off of 0.364), most other algorithms suggest OVCH1-AS1
   (RefSeq ID: [208]NR_073172) is a lncRNA rather than a protein coding
   gene (Supplementary Table [209]1).

   As lncRNA and anti-sense RNA transcripts can potentially modulate gene
   transcription, we analyzed the relative expression of all chromosome 12
   genes in our dataset (45,X versus 46,XX and 45,X versus 46,XY,
   independently). We did not identify consistently altered gene
   expression in this region (Fig. [210]7a–c). Next, more granular
   analysis of expression of genes in the OVCH1-AS1 locus was undertaken
   (OVCH1, ERCIG2, FAR2, TMTC1) using bulk RNA-seq data for each tissue as
   well as qRT-PCR. No significant differences in the relative expression
   of these genes were observed between monosomy X samples and the 46,XX
   and 46,XY controls in these tissues studied (Fig. [211]7d, e, and
   Supplementary Fig. [212]9 a, b, and Supplementary Data [213]37). Thus,
   OVCH1-AS1 is clearly a consistently upregulated autosomal long
   non-coding RNA associated with monosomy X in multiple tissues, but its
   potential biological role as a mediator of phenotypes in women with TS
   requires further investigation.

Multi-tissue analysis

   Given the relatively small sample size in each tissue studied (n = 4
   for each karyotype), an additional “multi-tissue” analysis was
   undertaken that combined the pancreas, liver, kidney, skin and mixed
   groups (n = 20 for each karyotype). Despite having more power to detect
   small changes, a similar group of genes was identified that were
   consistently lower (AKAP17A1, ASMTL, GTPBP6, PLCXD1, POLRMTP1, SLC25A6,
   ZBED1) or higher (OVCH1-AS1, PAEP) in monosomy X tissues (log2FC > 0.5,
   p-adj < 0.05) (Supplementary Figs. [214]4f, [215]7f; Supplementary
   Data [216]38-[217]41). By correlating the log[2]FC for 45,X versus
   46,XX with the same gene in 45,X versus 46,XY datasets, the
   distribution and magnitude of differentially expressed genes could be
   clearly seen (Fig. [218]8; see Supplementary Figs. [219]10 and [220]11
   for individual tissue correlation plots). As the multi-tissue analysis
   contains several different tissue samples from the same fetus
   (Supplementary Fig. [221]1a), further analysis was undertaken using a
   linear mixed effects model with individual as a random effect. No major
   differences were seen. Fewer significant results were found due to a
   reduction in false positives, as expected for an interaction test
   (Supplementary Data [222]42).

Fig. 8. Correlation plot of gene expression in 45,X tissues compared to 46,XX
(y-axis) and 46,XY (x-axis) showing global patterns of differential changes.

   [223]Fig. 8
   [224]Open in a new tab

   Each data point represents a specific gene in each of five different
   tissue groups (pancreas, liver, kidney, skin, mixed group). Genes with
   higher expression in monosomy X are shown in the upper, right quadrant
   (e.g., OVCH1-AS1). Genes with lower expression in monosomy X are shown
   in the lower, left quadrant. Key clusters of Y chromosome genes and
   X-inactivation genes (e.g. XIST) are also shown.

Monosomy X and the developing placenta

   Although monosomy X is the only chromosomal monosomy compatible with
   survival in humans, it has been estimated that most monosomy X
   pregnancies are lost^[225]5–[226]7. Several hypotheses have suggested
   that this could be due to fetal anomalies (such as lymphedema or
   hydrops fetalis), defects in placental development and function, or
   mechanisms involving the maternal-fetal
   interface^[227]5,[228]40,[229]41. Detailed transcriptomic analysis of
   human monosomy X placenta is lacking, especially during the early
   stages of fetal development when loss of many monosomy X pregnancies
   occurs.

   In order to investigate this further at a biologically relevant time
   point (late first trimester/early second trimester), samples were taken
   from monosomy X placentas (n = 6) between 11 and 15 wpc, together with
   matched 46,XX (n = 6) and 45,XY (n = 6) control placental tissues
   (Fig. [230]9a, Supplementary Fig. [231]12a).

Fig. 9. Differential expression of genes in the 45,X placenta compared to
46,XX or 46,XY controls.

   [232]Fig. 9
   [233]Open in a new tab

   a Overview of placental age stages used for the study. Karyotypes are
   indicated by the key. wpc, weeks post conception. b Expression of the X
   inactivation regulator XIST, bulk RNA-sequencing (RNA-seq) counts in
   the 45,X placenta group (n = 18) compared to 46,XX placenta (n = 6) and
   46,XY placenta (n = 6). c Principal component analysis (PCA) of 45,X,
   46,XX and 46,XY placental samples used in the study. PC, principal
   component. d Volcano plot showing differential expression of genes
   between the 45,X and 46,XX matched control samples (n = 4 in each
   group). The top ten most differentially expressed genes in each dataset
   are labeled, based on adjusted p-value (p-adj) and where log[2] fold
   change (FC) is greater than +/−0.5. Genes with higher expression in
   45,X samples have a positive log[2]FC and those with higher expression
   in control tissues have a negative log[2]FC. The significance level of
   highlighted points is shown in the key. e Volcano plot showing
   differential expression of genes between the 45,X and 46,XY matched
   control samples. f Venn diagram with the common intersection showing
   placenta-specific genes that are consistently lower in 45,X. Data were
   generated for each group (n = 6 versus n = 6) with a differential
   expression cut-off of log[2]FC > 0.5 and p-adj < 0.05. g Violin plot of
   CSF2RA expression (normalized counts) in the placenta (bulk RNA-seq)
   (n = 6 each group). h Violin plot of AADACL3 expression (normalized
   counts) in the placenta (bulk RNA-seq) (n = 6 each group). i Heat map
   of AADACL3 and CSF2RA expression across all placental samples (n = 6
   samples in each group).

   As placental mosaicism for a fetal 46,XX cell line has been proposed as
   a mechanism that can “rescue” or modify placental disruption^[234]41,
   we initially undertook SNP array analysis using DNA derived from two to
   four independent areas of each placenta in the 45,X group (n = 6, total
   n = 18 areas), as well as 46,XX and 46,XY controls. As expected, a low
   level 46,XX component was identified in all samples (45,X; 46,XY),
   consistent with the presence of limited maternal decidual tissue
   (Supplementary Fig. [235]13). Strong enrichment of a 46,XX line was not
   seen in 45,X samples compared to 46,XY controls. These findings were
   validated by investigating the expression of XIST in bulk RNA-seq data
   from 45,X samples, where similar transcript levels to 46,XY controls
   were seen (Fig. [236]9b and Supplementary Fig. [237]12b). Taken
   together, these data suggest that placenta mosaicism for a 46,XX cell
   line is not common in 45,X placenta, at least at this stage of
   gestation.

   We then analyzed bulk RNA-seq data to identify genes that were lower in
   45,X placenta (n = 6) compared to controls, which could help to
   elucidate potential mechanisms of pregnancy loss associated with
   monosomy.

   Global transcriptomic analysis using PCA showed that placental samples
   clustered together compared to control tissues as expected
   (Supplementary Fig. [238]14), and that within the placenta samples an
   influence of karyotype and age was seen (Fig. [239]9c). Volcano plots
   comparing 45,X with 46,XX and 46,XY placental samples again revealed
   the expected pattern of XIST and Y chromosome gene changes, with higher
   expression of OVCH1-AS1 also seen in 45,X tissue (Fig. [240]9d, e, and
   Supplementary Data [241]43–[242]46). To investigate placenta-specific
   genes that were consistently lower in 45,X placenta, a four-way
   analysis was undertaken using datasets derived from 46,XX or 46,XY
   versus 45,X placenta (“45,X lower” genes), as well as from 46,XX or
   46,XY placenta versus 46,XX or 46,XY control tissues
   (“placenta-specific” genes) (Fig. [243]9f, Supplementary Data [244]47,
   [245]48). Using this approach, the PAR1 gene CSF2RA (UCSC,
   chrX:1,268,814-1,309,935 41,122, human GRCh38/hg38) was identified as
   the principal placenta-specific gene that is lower in monosomy X
   (Fig. [246]9f–i, and Supplementary Data [247]49). Lower expression of
   an autosomal gene AADACL3 (UCSC, Chr1:12,716,110-12,728,760, human
   GRCh38/hg38) was also seen, although expression levels across samples
   were lower and more variable (Fig. [248]9f, h, i, and Supplementary
   Data [249]49).

   AADACL3 encodes arylacetamide deacetylase-like3, a potential membrane
   protein with hydrolase activity. This gene has low level expression in
   adult tissues and is only found in placenta, skin and breast (Human
   Protein Atlas Consensus RNA-seq) (Supplementary Fig. [250]15a).
   Consistent with this, analyses of available single cell RNA-sequencing
   (scRNA-seq) datasets^[251]42–[252]44 show low level expression of
   AADACL3 in the extravillous trophoblast of early placenta, and in
   cytotrophoblasts, stromal cells, endothelial cells and decidual cells
   in late third trimester and pre-term tissue (Supplementary
   Fig. [253]16c and Supplementary Fig. [254]17b)^[255]42–[256]44. The
   role of AADAC3L and its protein in the placenta is unknown.

   In contrast, CSF2RA shows much stronger and more specific expression in
   placenta, as well as in adult haematopoietic/immune regulating tissues
   (bone marrow, lymph node, tonsil, spleen) (Human Protein Atlas
   Consensus RNA-seq) (Supplementary Fig. [257]15b). CSF2RA encodes the
   Colony Stimulating Factor 2 Receptor Subunit Alpha (also known as
   Granulocyte-Macrophage Colony-Stimulating Factor (GM-CSF) Receptor
   Subunit Alpha or CD116). This protein forms part of a heterodimeric
   cytokine receptor complex that mediates the effects of colony
   stimulating factor 2 (CSF2)^[258]45. CSF2 (GM-CSF) acts through this
   low affinity receptor via STAT5 to stimulate the proliferation,
   differentiation and functional activation of hematopoietic
   cells^[259]45. In available single cell RNA sequencing (scRNA-seq) data
   from first trimester placenta, CSF2RA expression was observed in
   placental villous cytotrophoblast, extravillous trophoblast and
   syncytiotrophoblast cells, and well as in decidual macrophages
   (Fig. [260]10a and Supplementary Fig. [261]16)^[262]42. More granular
   analysis of term placenta by scRNA-seq shows high expression of CSF2RA
   in many hematopoietic cell lineages (macrophages, NK-cells, activated
   and resting T-cells), as well as in non-proliferative interstitial
   cytotrophoblasts (Supplementary Fig. [263]17)^[264]43.

Fig. 10. Expression of CSF2RA/CSF2RA in the placenta and related pathways.

   [265]Fig. 10
   [266]Open in a new tab

   a Consensus first trimester single cell data Uniform Manifold
   Approximation and Projection (UMAP) for clusters based on cell type
   (left panel) and feature plot for CSF2RA (right panel). These data were
   generated by the Vento-Tormo/Teichmann groups at the Wellcome Sanger
   Institute, Hinxton, UK and can be accessed using CZ CELLxGENE from
   [267]https://maternal-fetal-interface.cellgeni.sanger.ac.uk/ (Vento-Tor
   mo, R., Efremova, M., Botting, R.A. et al. Single-cell reconstruction
   of the early maternal–fetal interface in humans. Nature 563, 347–353
   (2018); 10.1038/s41586-018-0698-6)^[268]42. This graphic is generated
   under a Creative Commons Attribution-BY 4.0 International License
   ([269]https://creativecommons.org/licenses/by/4.0/). DC dendritic
   cells, dM decidual macrophages, dS decidual stromal cells, Endo
   endothelial cells, Epi epithelial glandular cells, EVT extravillous
   trophoblast, f fetal, F fibroblasts, HB Hofbauer cells, ILC innate
   lymphocyte cells, l lymphatic, m maternal, M3 maternal macrophages, PV
   perivascular cells, p proliferative, SCT syncytiotrophoblast, VCT
   villous cytotrophoblast. b H&E staining of 45,X placenta at 11 weeks
   post conception (wpc) (upper panel), 46,XY placenta at 14 wpc (center
   panel) and 45,X placenta at 14 wpc) (lower panel) (scale bars 50 μm).
   Arrowheads show regions of irregular villus border in the 14wpc 45,X
   sample. c Immunohistochemistry of CSF2RA (Colony Stimulating Factor 2
   Receptor Subunit Alpha) in 46,XX, 46,XY and 45,X placenta at 14wpc at
   lower magnification (left panels) (scale bars 100 μm) and at higher
   magnification (right panels) (scale bars 50 μm). Arrowheads show
   regions of irregular border and diffuse CSF2RA staining in the 14wpc
   45,X villus. Inset images show higher magnification of regions
   indicated by an asterix. d Biological process enrichment analysis for
   genes lower in 45,X placenta compared to 46,XY controls (bulk
   RNA-sequencing n = 6 each group; 266 genes identified) with a
   differential expression cut-off of log[2]FC > 0.5 and adjusted
   p-value < 0.05. The number of differentially expressed genes in 45,X
   placenta compared to 46,XX controls was insufficient for pathway
   analysis.

   To obtain more direct evidence for the role of CSF2RA in the monosomy X
   placenta, we undertook immunohistochemistry (IHC) for CSF2RA in 45,X
   placenta compared to 46,XX and 46,XY control samples. Between 11wpc and
   15wpc the monosomy X placenta shows increasing non-specific
   inflammatory changes as well as irregular villus tufting
   (Fig. [270]10b). CSF2RA localized to the syncytio- and
   cyto-trophoblastic layers in all samples as expected, but in the 45,X
   placenta at 14wpc showed less defined surface staining (Fig. [271]10c).
   Furthermore, pathway enrichment analysis of genes (n = 266;
   Supplementary Data [272]46) that are lower in monosomy X placenta
   compared to control (46,XY) identified processes linked to leukocyte
   activation and adhesion, immune response, and cytokine signaling
   (Fig. [273]10d).

   Taken together, these data provide some of the first direct evidence
   that CSF2RA/CSF2RA is altered in monosomy X placenta, and these
   findings suggest that dysregulation of immune and inflammatory
   processes may contribute to the placental dysfunction associated with
   loss of an X chromosome.

Discussion

   Monosomy X is the most common chromosomal aneuploidy in humans and can
   be associated with a broad range of clinical features in Turner
   syndrome. Some of these features have a developmental origin. The exact
   mechanisms that give rise to phenotypes in women with TS are still
   poorly understood, as are the mechanisms responsible for the high
   incidence of pregnancy loss. We therefore undertook a detailed
   transcriptomic analysis of monosomy X tissues during early human
   development, to better understand the molecular basis of Turner
   syndrome, with the ultimate aim of developing personalized approaches
   to long-term management.

   We first assessed genes with lower expression in monosomy X, to be
   consistent with the long-standing hypothesis that haploinsufficiency of
   genes in the X chromosome may drive phenotypic features. As expected,
   many of these genes were pseudoautosomal genes (e.g., SLC25A6, AKAP17A,
   GTPBP6, ZBED1, ASTML, DHRSX, PLCXD1) expressed from the PAR1 region.
   Not all PAR genes showed lower expression, mostly due to lower
   background transcript levels or greater variability in expression
   between different tissues (e.g., SHOX). A lack of SHOX differential
   expression has been found previously^[274]46. Notably, we also found
   PAR2 genes had no differences in expression between monosomy X and
   control karyotypes. This finding has also been shown in other recent
   studies^[275]20,[276]21,[277]32, suggesting that the dynamics of X
   chromosome PAR2 regulation may be different to PAR1.

   Several other studies have looked for key X chromosome genes that have
   lower expression in adult Turner
   tissues^[278]20,[279]21,[280]26,[281]29,[282]32,[283]47,[284]48. These
   studies have mainly analyzed blood leukocytes and more recently muscle
   and fat, or derived cell lines^[285]20,[286]32–[287]34. In addition to
   several PAR1 genes, the study by Viuff et al.^[288]20 found a subset of
   key genes that were differentially expressed by X chromosome dosage
   (45,X versus 46,XX and 47,XXY versus 46,XY), including genes that
   escape X inactivation (e.g., PUDP, ZFX, KDM6A, JPX, TSIX and XIST).
   These genes were also generally found to be differentially expressed in
   our dataset. Another recent study by San Roman et al.^[289]27 proposed
   a core group of X chromosome genes that are linked to sex differences.
   Several of these genes were found to be lower in our datasets, such as
   DDX3X, KDM5C, KDM6A, SMC1A and ZFX. By analysing ancestral X-Y gene
   pairs^[290]37, we showed consistently lower expression in monosomy X
   for EIF1AX/EIF1AY, ZFX/ZFY, DDX3X/DDX3Y, KDM6A/UTY, KDM5C/KDM5D and
   RSP4X/RSP4Y1. Although these genes can influence transcription and
   translation, the role of histone demethylases (e.g., KDM5C/KDM5D,
   KDM6A/UTY) may be particularly important. For example, pathogenic
   variants in KDM6A cause Kabuki syndrome 2 (KABUK 2, X-linked) (OMIM
   300867), which has a high prevalence of TS-associated features such as
   horseshoe kidney, short stature and hearing loss^[291]49. Here, we
   report, for the first time, core groups of PAR and XCI genes, and
   ancestral X-Y gene pairs that show lower expression in key
   biologically-relevant fetal monosomy X tissues such as pancreas,
   kidney, liver and skin, rather than the typical approach of studying
   adult samples, usually blood leukocytes.

   Extending our analysis to other genes with lower expression
   (log[2]FC < -0.35) in monosomy X identified several autosomal genes
   which may be clinically relevant. For example, lower expression of the
   low-density lipoprotein receptor gene (LDLR), which regulates
   cholesterol homeostasis, could contribute to the risk of dyslipidemia
   in Turner syndrome^[292]50. Previously, increased LDLR expression has
   been reported in amniotic fluid RNA in a small number of monosomy X
   fetuses^[293]46. Furthermore, pathway enrichment analysis identified
   biological processes related to ascending/thoracic aortic aneurysm and
   connective tissue disease^[294]38,[295]51. Aortic root dilatation and
   thoracic aortic aneurysm (TAA) is rare in the population but can be a
   significant risk in women with TS^[296]17,[297]52. Aortic root
   dilatation is now regularly monitored in
   adulthood^[298]3,[299]16,[300]18,[301]51 and TAA has the highest odds
   ratio for mortality in the adult TS group (standardized mortality
   ratio, =23; 95% CI 13.8–37.8)^[302]3,[303]53. The main genes identified
   in this pathway were collagen (COLA) genes and fibrillin 1 (FBN1).
   Defects in FBN1 cause Marfan syndrome, which is also associated with
   TAA^[304]38. This finding raises the possibility of a pathogenic link
   between the TS and Marfan syndrome. Further studies with larger sample
   numbers, and looking more specifically at developing aortic tissue, are
   needed.

   We also hypothesize that this risk of TAA could be compounded by lower
   expression of AGTR2 (angiotensin II receptor type 2, AT2R) in monosomy
   X. AGTR2 typically antagonizes the effects of the angiotensin II type 1
   receptor (AT1R). Lower AGTR2 could lead to higher blood pressure and
   vascular proinflammatory effects^[305]54. Several studies have shown
   higher plasma renin activity in women with TS^[306]55,[307]56. In
   addition, monosomy X augments the severity of angiotensin II-induced
   aortopathies in mice^[308]57. Notably, AT1R blockade has been shown in
   recent systematic reviews to reduce the risk of thoracic aortic
   aneurysm in Marfan syndrome^[309]58. Further studies are needed to
   assess whether the interplay of these factors could lead to increased
   TAA risk in women with TS and be an avenue for new therapeutic
   approaches.

   Although most of the effects of TS might be expected to be due to a
   reduction of gene dosage, one gene (OVCH1-AS1) was consistently higher
   in all datasets studied. OVCH1-AS1 encodes a long non-coding RNA that
   is expressed at relatively low transcript level, which is typical for
   lncRNAs, but with at least two-fold increase in expression in all 45,X
   tissues assessed. The potential role of OVCH1-AS1 is not known. Recent
   studies have suggested different expression levels of leukocyte
   OVCH1-AS1 in relation to frailty with age in humans^[310]39, as well as
   higher OVCH1-AS1 expression in biopsies from Crohn’s disease^[311]59.
   Furthermore, several studies have identified OVCH1-AS1 as a
   higher-expressed gene in leukocytes, fat, and muscle from adult
   women with TS ^[312]20,[313]21,[314]26,[315]29,[316]47. This finding
   was rarely highlighted in detail, but here we have conclusively
   demonstrated that OVCH1-AS1 is higher in monosomy X in 14 different
   sub-studies from multiple different tissues.

   As antisense lncRNAs sometimes modulate sense strand RNA, or genes in
   the region, we undertook more detailed analysis of the OVCH1-AS1 locus
   of chromosome 12 to investigate potential cis effects, but we did not
   see an obvious effect on the expression of OVCH1 nor of other local
   genes. Of note, like many lncRNAs, this gene is present in humans but
   not in the mouse. We hypothesize that, if OVCH1-AS1 is biologically
   relevant, this feature could contribute to marked differences in
   phenotype between monosomy X in humans and the murine model of a single
   X chromosome^[317]60.

   Some of the biggest questions to address are how monosomy X is such a
   common cause of pregnancy loss and also whether there are any specific
   features in those pregnancies that survive to term? It has been
   estimated that many monosomy pregnancies are lost mostly towards the
   end of the first trimester and that this contributes to a substantial
   proportion of spontaneous miscarriage in the general
   population^[318]5,[319]41,[320]61.

   One long-standing hypothesis is that term monosomy X pregnancies have
   placental mosaicism for a fetally-derived 46,XX line, but direct data
   are limited^[321]41,[322]62. Here, we did not identify significant
   46,XX mosaicism using both RNA (XIST counts) and DNA (SNP arrays)
   technologies in the six 45,X placental samples studied, suggesting that
   placenta mosaicism is not likely to be a common cause of fetal rescue
   in TS, at least at this gestational age, although placental sampling
   was limited to between two to four distinct regions of each 45,X
   placenta. Furthermore, because the placental material was obtained
   following termination in the early second trimester, we do not know
   whether spontaneous loss might still have occurred. Using a
   trophoblast-like model derived from X-monosomic human induced
   pluripotent stem cells, Ahern et al. showed impaired secretion of
   placental growth factor and human chorionic gonadotropin^[323]63. We
   did not observe significantly lower expression of the genes encoding
   these growth factors in our datasets. Nevertheless, whether there could
   be a tropic growth advantage for any small mosaic population of 46,XX
   cells to expand with time throughout gestation remains to be seen.

   Another question relates to the mechanistic basis of pregnancy loss.
   Several studies have proposed that PAR genes such as CSF2RA are
   implicated, but direct data are limited^[324]2,[325]7. We therefore
   developed a systematic approach to identify placenta-specific genes
   with lower expression in monosomy X placenta compared to control
   placenta. We found two genes of interest: AADACL3 and CSF2RA.

   AADACL3 encodes arylacetamide deacetylase-like 3, a putative membrane
   hydrolase expressed at low level, primarily in the extravillous
   trophoblast. AADACL3 expression was more variable between monosomy X
   samples and controls. The functional role of this gene in the placenta
   is unclear, although higher AADACL3 expression has been reported in
   46,XY preterm and term placental tissue compared to 46,XX^[326]64.

   More importantly, the PAR1 gene CSF2RA was observed in our dataset as
   having consistently lower expression in monosomy X compared to both
   46,XX and 46,XY control placenta samples. In scRNA-seq analysis, CSF2RA
   is clearly expressed in early trimester and term placenta, as well as
   in macrophages, NK-cells and activated and resting T-cells, consistent
   with its expression in placenta and hematopoietic system-derived
   tissues in adult panels^[327]42,[328]43,[329]65. CSF2RA encodes the
   alpha subunit of the CSF receptor, which mediates the effects of CSF2
   (GM-CSF) on hematopoietic differentiation, immunity, and inflammation.
   Immune regulation within the fetal-placental interface is emerging as a
   key mechanism in pregnancy maintenance, and CSF2RA has been proposed as
   a candidate gene for early lethality in 45,X embryos^[330]2,[331]7.
   CSF2RA has been shown to have 10-fold lower expression in 45,X-derived
   human embryonic stem cells compared to 46,XX-derived controls, and
   reduced expression in BDCA4+ dendritic cells from women with
   TS^[332]7,[333]66. CSF2RA is implicated in placental
   development^[334]67, and studies in bovine cultured embryos suggest
   CSF2 promotes embryo survival through antiapoptotic effects and by
   signaling of developmental pathways^[335]68. However, direct analysis
   of monosomy X placenta has not been reported. Of note, Csf2ra is
   autosomal in the mouse (chromosome 19).

   Here, we show morphological changes in the monosomy X placenta during
   development. Using IHC, we have shown that CSF2RA localizes to
   syncytial trophoblast and basal trophoblast cells, with more diffuse
   patterning in monosomy X placenta at 14wpc at a time of degeneration of
   the brush border. Coupled with this, pathway analysis of
   lower-expressed genes in the monosomy X placenta showed enrichment of
   key biological processes, such as leukocyte activation, immune
   responses, and inflammation, involving key genes such as XCR1,
   interleukin 1 and interleukin family receptors. Taken together, these
   findings suggest that CSF2RA is likely to be a key modulator of
   fetal-placental dysfunction in monosomy X pregnancies and might have
   implications for our understanding of pregnancy loss in general.
   Further work to elucidate the potential mechanistic role of CSF2RA is
   needed, such as in model systems of monosomy X placental development
   using organoids, and in mediating immune defence mechanisms.
   Furthermore, as we studied placental material at 11-15wpc, we do not
   know if the samples we analysed harbored any specific features that
   protected against pregnancy loss at an earlier stage (first trimester),
   or whether the histological and transcriptomic changes seen could still
   have been associated with pregnancy loss at a later gestational age.

   This study has several limitations. Firstly, access to tissues of
   interest was limited so only a relatively small number of samples were
   available and within a relatively narrow developmental time window, and
   one 45,X pancreas sample (13 wpc) was bisected in order to have four
   samples in this group. However, having samples within a 4-week period
   and a carefully balanced study design reduced marked variations due to
   organ development, increasing the power to detect more subtle sex
   chromosome-related effects. Furthermore, a combined “multi-tissue”
   analysis provided additional data. Second, one fetus had low level
   mosaicism for a Y chromosome line. The overall effects of this
   variability on our findings were likely to be minimal. Third,
   alternative approaches such as scRNA-seq might have provided more
   granular detail of single cell transcriptomics, but risked sampling
   bias and limited detection of low-level RNA transcripts. Thus, a bulk
   RNA-seq strategy was used with robust experimental design, replicating
   data with both 46,XX and 46,XY control groups, and validating findings
   in multiple different tissue groups. This is important as gene dosage
   changes linked to loss of a single X chromosome are generally subtle
   and of a lower magnitude than in most studies of developmental
   differential gene expression. Finally, sampling a greater number of
   regions of the placenta would have provided more data to exclude
   significant 46,XX confined placenta mosaicism, although none was seen
   in the 18 samples analysed from 6 placentae in the study.

   To the best of our knowledge, this is the first detailed transcriptomic
   study of human monosomy X in key clinically relevant tissues (e.g.,
   pancreas, kidney, liver, skin, and placenta). This work has highlighted
   several important findings with implications for sex chromosome
   biology, the underlying mechanisms that could contribute to TS and
   potential targeted therapeutics. These findings will be important for
   facilitating research in the future.

Materials And Methods

Tissue samples

   Human fetal tissue samples included in this study were obtained with
   informed consent and with ethical approval (REC references: