3D conformation of the TLR7-TLR8 region on the X chromosome
We first investigated chromatin conformation over the TLR7-TLR8 region of chromosome X by leveraging available chromosome conformation capture (Hi-C) datasets (64, 68). Data from the GM12878 human transformed lymphoblastoid female cell line allowed allelic discrimination of the conformations of the Xa, of maternal origin in this line, and the paternal Xi (65). The overall conformation of X chromosomes confirmed the erosion of topologically associating domain (TAD) structures on the Xi, substituted as expected by two mega-domains (data not shown), but no major conformational differences were observed at 5kb-resolution between the Xa and the Xi over the TLR7-TLR8 region, confirming that escape regions retain the typical TAD topology (69) (Fig. 1C and Fig S1). This suggested that global non-allelic Hi-C analyses of female X chromosomes can be informative about the typical conformation of the TLR7-TLR8 region in both the Xa and Xi, enabling a comparison with the single X chromosome of male cells.
High 1kb-resolution, non-allelic Hi-C data from GM12878 cells (65) revealed that TLR7 and TRL8 are located in the same domain of interactions (Fig. 1D). However, on a close look at this region, it is striking that the TLR7 promoter makes a strong interaction with the adjacent PRPS2 gene, denoted by a characteristic loop signal (Fig. 1D, dark blue circle), whereas TLR8 seems to form a sub-domain of its own (Fig. 1D, yellow triangle). Similar conformations were found in male monocytes (66) and T cells (67) (Fig. 1E,F). These data on the 3D conformation of the TLR7-TLR8 genomic region suggested different patterns of transcriptional regulation for TLR7, on one hand, and for TLR8, on the other, and prompted us to examine the expression of primary transcripts from either gene using RNA FISH on immune cell types where both receptors were known to be co-expressed.
TLR7 and TLR8 evade X chromosome inactivation in female monocytes and CD4 + T cells
We used RNA FISH probes to detect both TLR7 and TLR8 primary transcripts on the Xa and Xi of CD14+ monocytes and CD4+ T lymphocytes from women. To discriminate the Xa and Xi chromosomal origins of TLR7 or TLR8 primary transcripts, early RNA FISH experiments involved a previously validated probe for the long non-coding RNA XIST (reference (48), and Table S1), but this probe failed to paint the Xi territory in female monocytes (not shown). Similar occurrences of XIST non-detection have been noted by others with regard to resting B and T lymphocytes (70–72). Our alternative strategy was to differentiate instead the Xa of female cells. For this, we searched the scientific literature and relevant databases for X-linked genes subject to XCI that were well-expressed in monocytes. We developed RNA FISH probes for three such genes, MSN, PGK1 and CFP (Table S1), respectively encoding moesin, phosphoglycerate kinase 1, and complement factor properdin. Preliminary hybridizations with each probe separately confirmed widespread but strictly mono-allelic transcription of these genes in female human monocytes (Fig S2A,B), and we used the pooled probes as a positive marker of the Xa in subsequent RNA FISH experiments. A fraction of female monocytes exhibited two transcriptional foci for TLR7 (Fig. 2A) or TLR8 (Fig. 2C), denoting bi-allelic expression of both genes. The frequency of nuclei with a positive signal for the Xa probes was 35% in average (Fig S2C), whereas the frequency of nuclei positive for either TLR7 or TLR8 primary transcripts was > 40% in monocytes (Fig S2D). The percentages of monocyte nuclei with biallelic expression of TLR7 or TLR8 among total positive cells for either TLR7 or TLR8 signals were 10% (95% CI: 4%-16%) and 17% (95% CI: 8%-27%), respectively. Additionally, we ascribed transcription to the Xi in those cells where a single TLR7 or TLR8 transcriptional focus was observed separate from a patent Xa as detected by the Xa probe (Fig. 2B and 2D). By these criteria, we observed escape from XCI for TLR7, as expected, but also for TLR8, in all the donors of our female study group (n = 6; S1A Data). The frequency of TLR7 transcription on the Xi varied donor to donor between 5% and 32% of cells, with a group mean of 13% (Fig. 2E). For TLR8, transcription on the Xi concerned 10–34% of cells, with a group mean of 17% (Fig. 2F).
Because TLR7 and TLR8 are also expressed in human CD4+ T cells, we investigated whether our findings on XCI escape in CD14+ monocytes could be extended to this lymphocyte class. For this, naive CD4+ T lymphocytes were stimulated with IL-2, and then activated through CD3 and CD28 (Fig S3B). We verified activation on the third day using flow cytometry analysis for CD25 and CD69 levels (Fig S3C). As described earlier for monocytes, we concluded to XCI escape in CD4+ T cells from women based on the presence of TLR7 (Fig S3D) or TLR8 (Fig S3E) transcripts on both X chromosomes of a cell or, alternatively, on identifying a single transcriptional signal on the Xi (not shown). TLR7 evaded XCI in 5–11% of T cells, with a group mean of 8% (Fig. 2G and S1 Data). For TLR8, the frequency of escape cells varied over a 4–20% range, with a group mean of 8% (Fig. 2H and S1 Data). On average, XCI escape for these genes was decreased 2- to 3-fold in T cells relative to monocytes. By contrast with monocytes, the XIST hybridization signal characteristic of the Xi could be visualized by RNA FISH in CD4+ T cells three days after stimulation through CD3 and CD28 (72). We were thus able to detect TLR7 and TLR8 transcripts just next to the inactive Xi territory covered with XIST RNA, further confirming the XCI escape of the two genes (Fig. 2I,J). Intra-individual levels of XCI escape for TLR7 and TLR8 were not significantly correlated in monocytes (Pearson correlation coefficient r = − 0.42; p = 0.40; Fig S4A) nor in CD4 T cells (r = − 0.06; p = 0.92; Fig S4C).
TLR7 and TLR8 evade XCI in monocytes from 47,XXY KS males
In a previous study, we established TLR7 escape from XCI in monocytes from four 47,XXY men with Klinefelter syndrome (KS) (48). We have studied here five further KS men by RNA FISH. As with female monocytes, we were able to establish XCI escape in all the individuals of this study group based on transcriptional TLR8 or TLR7 foci present on both X chromosomes (Fig. 3A, top and bottom right), or on a single signal ascribable to the Xi (Fig. 3A, top and bottom left). TLR7 evaded XCI in 10–17% of KS male monocytes (a narrower range than in women), with a group mean of 13% (Fig. 3B and S1 Data). For TLR8, the frequency of escape cells varied over an unexpectedly wide range, 8–49%, with a group mean of 23% (Fig. 3C). As in females, intra-individual levels of TLR7 and TLR8 escape were not correlated (r = − 0.03; p = 0.96; Fig S4B).
The frequencies of TLR7 and TLR8 transcriptional foci are sex-biased
Regardless of the Xa or Xi chromosome of origin, the proportion of monocytes that exhibited TLR7 or TLR8 transcripts exhibited wide inter-individual variation, and was higher overall in the women group (n = 6) than among euploid men (n = 7). There were substantial women-versus-men differences in regard to TLR7 (Fig S5A), with 31% positive cells in women versus only 17% in euploid men, and 18% for men with KS (p < 0.0001). This represents a difference between groups dependent on sex, not on the number of X chromosomes. For TLR8, by contrast, between-groups differences were modest (Fig S5B), with a female group mean of 30% versus 24% for XY males, and 31% for XXY KS males (n = 5). For women and XY men, intra-group variation in the frequency of TLR8+ cells was distinctly wider than the difference of means between the two groups (Fig S5B).
We next determined transcript detection frequencies on the Xa specifically, by considering only those cells positive for the Xa marker probe, and the TLR7 or TLR8 foci co-localizing with the Xa signal. Here again, we observed a clear difference between women and XY men, with group means of 48% and 33% of TLR7 positive cells, respectively (p < 0.0001) (Fig S5C). With regard to TLR8 transcripts, the greatest contrast occurred between the groups of XY and XXY KS men, who exhibited 45% and 61% of positive monocytes, respectively (p < 0.0001), with the women’s group at an intermediate value of 54% (Fig S5D).
The analysis for TLR7 and TLR8 transcripts in CD4+ T cells, circumscribed to women and XY men, showed a sex bias in the same direction as in monocytes (Fig S6A and S6B, S6 data). The women’s group mean frequency of TLR7-positive cells was 19% versus 15% for XY men. For TLR8, the divergence was more marked with 31% of positive cells in women versus 20% in XY men (p < 0.0001). These observations on monocytes and T cells dovetail with our previous observation that, on average, women’s mononuclear blood cells express more TLR7 protein than the cells from normal men (48). The female bias was also clearly visible in the parallel RNA FISH analysis of the transcripts from the Xa specifically (Fig S6C and S6D), with group means of 48% in women versus 30% in XY men for TLR7, and 63% in women versus 39% in XY men for TLR8. These sex-dependent divergences were significant for both genes at the p < 0.0001 level.
Overall, the higher counts of TLR7- and TLR8-positive cells in monocytes and T cells from women relative to the male cells is likely to arise not only from the contribution of escape transcripts of the Xi alleles but also from greater transcriptional frequencies on the Xa in women. This strongly suggests that the single X chromosome of XY men and the Xa of women are functionally non-equivalent as regards the X-linked TLR loci. These observations prompted us to expand our RNA FISH study of TLR7 and TLR8 together to quantify the suggested sex bias.
The combined transcription profile of TLR7 and TLR8 is sex-biased in monocytes
Hybridizations combining the TLR7 and TLR8 probes hinted at a co-transcriptional sex bias in monocytes, as individual cells from XY men were positive for either TLR7 (Fig. 4A) or TLR8 RNA signals alone (Fig. 4B), but only rarely for both genes at the same time (Fig. 4C). By contrast, co-occurring TLR7 and TLR8 signals were readily observable in women’s monocytes, where signals from TLR7, TLR8 or both genes occurred in a variety of combinations reflecting the presence of two potential source X chromosomes, Xa and Xi (Fig. 4D–G). Monocytes from XXY KS males exhibited RNA FISH patterns similar to those of the female cells (Fig. 4G). In XY men, 95% of signal-positive monocytes, where the two genes are necessarily in cis, exhibited transcripts from either TLR7 (Fig. 4A,G) or TLR8 alone (Fig. 4B,G), and only a consistently small minority (< 2% of all cells) transcribed both genes at the same time (Fig. 4C,G and Fig. 5A). Among women’s monocytes, TLR7+TLR8+ cell numbers reached 14% on average, a seven-fold increase relative to XY men (Fig. 4D-G and Fig. 5A; p < 0.0001). Monocytes from XXY KS men exhibited an intermediate group mean (9% of TLR7+TLR8+ cells Fig. 4G and Fig. 5A).
This contrast between women and XY men persisted when we next compared the single X of men with the Xa of women (Fig. 5B). The frequency of simultaneous transcription of TLR7 and TLR8 was again 7-fold greater on the Xa of women (group mean 31%) than on the X of XY men (4.1%; 95% CI: 2.4–6.0%), while the overall frequency of positive events (for TLR7, TLR8, or both genes together) was similar between women, normal men and men with Klinefelter syndrome (Fig S5E). Reciprocally, the frequency of monogenic expression (either TLR7 or TLR8) was two-fold enriched on the X of XY men (group mean 69%) compared with the Xa of women (38%), with XXY KS men in between the two other groups (53%) (Fig S5F). These observations indicate that the Xa of women and XXY KS men is non-equivalent with the single X of XY men regarding the combined transcription of TLR7 and TLR8 in monocytes.
TLR7 and TLR8 are transcriptionally non-independent in monocytes
The low frequency of monocytes from XY men where TLR7 and TLR8 were transcribed at the same time suggested that the two genes are transcriptionally non-independent from each other. To investigate this possibility, we cross-classified the cell counts in the RNA FISH data as 2×2 contingency tables, i.e., monocytes were stratified depending on the presence of signals for both TLR7 and TLR8, either gene alone, or neither gene (S3 Data). This allowed the theoretical (expected) cell counts in each table cell to be calculated under the null hypothesis of independence between the two genes. We used this information to compute the observed-to-expected ratio (obs/exp) for double-positive events, namely the ratio of observed TLR7+TLR8+ cell counts to the expected number of cells in this stratum assuming transcriptional independence between TLR7 and TLR8. Non-independence would be denoted by a deviation from the critical value, obs/exp = 1, as a trend for either mutually exclusive (obs/exp < 1) or co-dependent (obs/exp > 1) transcription. As shown in Fig. 5C, monocytes from euploid males comprised only one-half of the expected number of cells displaying TLR7 and TLR8 transcripts simultaneously, suggesting mutually exclusive transcription (obs/exp = 0.49; 95% CI: 0.35–0.62). Remarkably, we observed an excess of comparable magnitude for double-positive cells among monocytes from females (obs/exp = 1.62; 95% CI: 1.60–1.70) and KS males (obs/exp = 1.70; 95% CI: 1.58–1.84), indicating a trend for co-dependent transcription.
In parallel, we used a classical measure of association between two nominal variables, Yule’s Q coefficient of association in 2×2 tables (61). Q can be intuitively interpreted by analogy with a correlation coefficient: Q = − 1 would denote mutually exclusive transcription of TLR7 and TLR8 in single cells; Q = 0, independent transcription at either locus; and Q = 1, that both genes are always either on or off at the same time. We computed Yule’s Q for each individual, together with the corresponding meta-analytical group summaries for this statistic (Fig. 5D). Q = − 0.48 (95% CI: −0.61 to − 0.35) in euploid men; Q = 0.54 (95% CI: 0.47–0.61) in women; and Q = 0.53 (95% CI: 0.45–0.60) in KS men. Deviations from independence (Q = 0) were significant for all groups (p < 0.0001 in all instances). This analysis confirmed the transcriptional non-independence between TLR7 and TLR8 expression in women and KS men (Q > 0), and the opposite patterns of transcriptional association in euploid men (Q < 0).
The preceding analyses scored the cells without regard to the chromosome of origin of the transcripts in women and KS men (“Any X” data), but we carried out a parallel scoring procedure (S4 Data) restricted to Xa+ cells and considering only the alleles on the Xa. This analysis revealed a similar excess of TLR7-TLR8 co-transcription among women with an elevated obs/exp ratio (obs/exp = 1.24; 95% CI: 1.14–1.32) relative to XY men (obs/exp = 0.31; 95% CI: 0.20–0.43) (Fig. 5E). The values for Yule’s Q coefficient of association in the Xa-specific data paralleled the patterns for the obs/exp ratio, and confirmed a negative association between the transcription of TLR7 and that of TLR8 in euploid men (Q = − 0.82; 95% CI: −0.89 to − 0.73), and a positive association in women (Q = 0.50; 95% CI: 0.31–0.66) (Fig. 5F). Consistent with obs/exp ≈ 1, the group value for Yule’s Q among KS men was not significantly different from Q = 0 denoting independence, but we concluded that our group of five 47 XXY KS males non-homogeneous, because it encompassed euploid male-like, female-like, and neutral patterns of TLR7-TLR8 co-transcription on the Xa (Fig. 5F).
Taken together, these observations indicate the occurrence of mutually exclusive transcription of the genes TLR7 and TLR8 in cis in the monocytes from euploid men, in parallel with co-dependent transcription of these genes on the Xa of the monocytes from women, and a heterogenous phenotype in 47,XXY men.
TLR7-TLR8 transcriptional dependency differs between the monocytic Xa and Xi
Because only a fraction of the female monocytes analyzed exhibited XCI escape, the cross-classified cell counts based only on the Xi signals encompassed fewer positive cells, with instances of zero events in the double-positive (TLR7+ TLR8+) stratum (S2 Data). To analyze these sparse data for Yule’s Q, we pooled the cell counts group-wise to increase statistical power (S2 Data). Fig S7A shows the pooled data as 2×2 contingency tables of observed frequencies, in a comparison with the frequencies expected under the null hypothesis of independent transcription at either Xi locus. The 95% CIs for Q straddle the critical value, Q = 0 in both the women and KS men groups, and the corresponding p-values from Monte Carlo χ2 tests were non-significant (Fig S7B). This result points to the absence of transcriptional association between the two genes on the Xi (Fig S7B). Next, we contrasted these TLR7-TLR8 co-transcription data for the Xi with the data for the Xa of women and KS men (Fig. 5E,F and S5 Data). We performed Monte Carlo χ2 tests on 3×2 tables of cell counts to formally compare the relative proportions of TLR7+ TLR8+, TLR7− TLR8+, and TLR7+ TLR8− cells in the Xa and the Xi of individual donors. The tests demonstrated significant differences between the Xa and the Xi for all individuals (S5 Data), and there is therefore a conclusive divergence between the marked TLR7-TLR8 transcriptional co-dependency on the Xa (Fig. 5E,F) and the trend for non-dependency observed on the Xi alleles (Fig S7A,B).
TLR7 and TLR8 are transcriptionally non-independent in CD4 + T cells
Further to the study of monocytes, we applied a similar strategy to CD4+ T cells from women and euploid men. We scored first the cells regardless of the Xa marking (Any X data), and Fig. 6A shows that, similar to monocytes, TLR7+TLR8+ events among female T cells were more frequent than among the male cells: 13% of cells versus only 2% in males, a 6:1 ratio consistent with the 7:1 ratio observed earlier in monocytes (S6 Data). There was a clear excess of TLR7+TLR8+ cell counts among the female T cells (obs/exp = 2.27; 95% CI: 2.15–2.40) (Fig. 6B), a pattern even more pronounced here than in monocytes. The male cells again displayed a shortfall in the expected number of TLR7+TLR8+ events (obs/exp = 0.77; 95% CI: 0.56–1.00). Yule’s Q coefficient confirmed the strong positive association in women (Q = 0.79; 95% CI: 0.75–0.84) (Fig. 6C). The parallel analysis restricted to the Xa (i.e., of transcripts for a TLR7-TLR8 gene pair in an obligate cis topology) revealed a similar contrast (Fig. 6D) between the excess of double positive cells in women: obs/exp = 1.19; 95% CI: 1.15–1.23) and the shortfall in euploid men (obs/exp = 0.55; 95% CI: 0.42–0.69). On the Xa, the negative association was conclusive (Q = − 0.56; 95% CI: −0.68 to − 0.43; p < 0.0001) and of a similar magnitude to the association observed earlier in male monocytes (Fig. 5E). The analysis for Yule’s Q corroborated the positive transcriptional association of the two Xa genes in women’s T cells (Fig. 6F). In summary, the observations in both CD14+ monocytes and CD4+ T cells outline a pattern of mutually exclusive transcription for the adjacent TLR7 and TLR8 genes in the cells from euploid men, and an opposite pattern of co-dependent transcription of these genes in the cells from women. This is further proof of functional non-equivalence for this locus between the single X chromosome of men and the active X chromosome of women.
Female immune cells express higher levels of TLR8 protein than male cells
We previously reported that female PBMCs expressed higher levels of TLR7 protein by western blot, including the full-length (140-kDa) and proteolytically mature (75-kDa) forms of the protein (56, 73). We performed a similar analysis by comparing TLR8 expression between male and female PBMCs using a highly specific TLR8-specific antibody (6). Normalized TLR8 protein expression was significantly higher in female than in male PBMCs (Fig. 7A,B), despite similar proportions of monocytes between either sex (Fig. 7C, Fig S8AB). Because monocytes were found to express the highest level of TLR8 protein compared with other immune cell populations (not shown), we next analyzed TLR8 expression by flow cytometry within monocyte subsets defined by the expression profile of the CD14 and CD16 markers (Fig. 7D, Fig S8A). Whereas CD14+CD16− classical and CD14+CD16+ intermediate monocytes all expressed TLR8 protein to variable degrees, only 60% on average of non-classical CD16+CD14− monocytes positively stained for TLR8 (Fig. 7D). Geometric mean fluorescence intensities (GMFIs) of TLR8+ cells revealed higher TLR8 protein expression for all subsets of female monocytes than in their male counterparts (Fig. 7E-G). Together, using two different highly specific mAbs (6, 40), our data provide evidence for higher TLR8 protein expression in female than in male leukocytes, including all monocyte subpopulations.