TLR8 escapes X chromosome inactivation in human monocytes and CD4⁺ T cells

Abstract

Background

Human endosomal Toll-like receptors TLR7 and TLR8 recognize self and non-self RNA ligands, and are important mediators of innate immunity and autoimmune pathogenesis. TLR7 and TLR8 are, respectively, encoded by adjacent X-linked genes. We previously established that TLR7 evades X chromosome inactivation (XCI) in female immune cells. Whether TLR8 also evades XCI, however, has not yet been explored.

Method

In the current study, we used RNA fluorescence in situ hybridization (RNA FISH) to directly visualize, on a single-cell basis, primary transcripts of TLR7 and TLR8 relative to X chromosome territories in CD14⁺ monocytes and CD4⁺ T lymphocytes from women, Klinefelter syndrome (KS) men, and euploid men. To assign X chromosome territories in cells lacking robust expression of a XIST compartment, we designed probes specific for X-linked genes that do not escape XCI and therefore robustly label the active X chromosome. We also assessed whether XCI escape of TLR8 was associated with sexual dimorphism in TLR8 protein expression by western blot and flow cytometry.

Results

Using RNA FISH, we show that TLR8, like TLR7, evades XCI in immune cells, and that cells harboring simultaneously TLR7 and TLR8 transcript foci are more frequent in women and KS men than in euploid men, resulting in a sevenfold difference in frequency. This transcriptional bias was again observable when comparing the single X of XY males with the active X of cells from females or KS males. Interestingly, TLR8 protein expression was significantly higher in female mononuclear blood cells, including all monocyte subsets, than in male cells.

Conclusions

TLR8, mirroring TLR7, escapes XCI in human monocytes and CD4⁺ T cells. Co-dependent transcription from the active X chromosome and escape from XCI could both contribute to higher TLR8 protein abundance in female cells, which may have implications for the response to viruses and bacteria, and the risk of developing inflammatory and autoimmune diseases.

Highlights

• TLR8, like TLR7, escapes X chromosome inactivation in immune cells from women and 47,XXY men with Klinefelter syndrome.

• The frequency of double-positive cells for TLR7 and TLR8 primary transcripts is sevenfold higher in women than in men.

• TLR7 and TLR8 form a co-regulated gene cluster on the human X chromosome, with sex-specific, divergent transcriptional patterns observable in monocytes and CD4⁺ T lymphocytes.

• Co-dependent transcription of the TLR7 and TLR8 genes on the active X was observed in women and in men with Klinefelter syndrome, contrasting with mutually exclusive transcription in euploid men.

• Blood mononuclear cells, including monocyte subsets, expressed higher levels of TLR8 protein in females than in euploid males.

Plain Language summary

Human endosomal Toll-like receptors TLR7 and TLR8, encoded by two adjacent X-linked genes, recognize self and non-self RNA ligands, and are important mediators of innate immunity and autoimmune pathogenesis. We previously reported that TLR7 evades X chromosome inactivation (XCI) in female immune cells, correlating with enhanced functional properties in B cells harboring biallelic expression of this gene. Here, we conducted a comprehensive single-cell resolution analysis of the transcriptional regulation of both TLR7 and TLR8, in CD14⁺ monocytes and CD4⁺ T lymphocytes. We unequivocally demonstrated that TLR8, like TLR7, escapes XCI in immune cells from female and Klinefelter syndrome males. When we analyzed TLR7 and TLR8 transcripts together, cells from women and KS men exhibited higher frequencies of cells co-transcribing the two genes. Surprisingly, these differences were attributable not only to the ability of TLR7 and TLR8 to be expressed on the Xi, but also to the joint transcriptional behavior of the TLR7–TLR8 gene pair on the active X chromosome specifically. This contrasted with a striking pattern of mutually exclusive transcription on the single X of euploid men. Corroborating our RNA FISH results, we found higher TLR8 protein expression in female than in male leukocytes, including all monocyte subpopulations. In summary, our data suggest that sex-biased co-regulation of the Toll-like receptor locus and XCI escape of TLR8 contribute to the sexual dimorphism in TLR8 expression, which may have important consequences for the functional make-up of monocyte and T cell populations.

Background

Human Toll-like receptors 7 and 8 (TLR7, TLR8) are essential components of the innate immune response to microbial pathogens. These paralogue receptors recognize RNA degradation products from viruses, intracellular bacteria, fungal and protozoan pathogens, and also endogenous sources [1,2,3,4,5,6,7,8,9,10,11]. TLR7 and TLR8, along with receptors for CpG-unmethylated DNA (TLR9) and double-stranded RNA (TLR3), make up the subfamily of endosomal nucleic acid-binding TLRs [12]. Crystal structures have revealed TLR7 and TLR8 to be dual sensors possessing two ligand-binding sites working in synergy, one for small agonists, namely guanosine in TLR7 and uridine in TLR8, and a separate site for short single-stranded RNA molecules: uridine-rich RNA in TLR7, and U- and G-containing RNA in TLR8 [13,14,15].

TLR7 and TLR8 exhibit different expression landscapes among human leukocytes [16, 17]. TLR7 is found primarily in plasmacytoid dendritic cells (pDCs), monocytes, and B lymphocytes, whereas TLR8 is preferentially expressed in monocytes, myeloid dendritic cells, and neutrophils. Vigorous production of type I interferon (IFN) by pDCs upon TLR7 stimulation is a key component of the antiviral response, whereas TLR7 and TLR8 signaling leads to the secretion of proinflammatory cytokines in cells of the monocyte/macrophage lineage. TLR7 engagement, however, is also known to elicit anergy and cell death in CD4⁺ T cells during chronic infection [18, 19], and TLR8 engagement is reported to reverse Foxp3⁺ Treg cell function through inhibition of glucose uptake and glycolysis [20, 21]. Accordingly, ligand sensing by TLR7 or TLR8 initiates a variety of processes across immune cell populations, as illustrated by single-stranded viral RNA fragments from SARS-CoV-2, which elicit either an antiviral or a proinflammatory response in pDCs, myeloid dendritic cells and lymphocytes [22, 23].

Human endosomal TLRs have evolved under stringent purifying selection owing to their non-redundant role in preserving host fitness [24], and only of late have null mutations of the X-linked genes, TLR7 and TLR8 been discovered. During the spread of SARS-CoV-2, deficiency in type I interferon-dependent antiviral immunity caused by rare loss-of-function or hypomorphic TLR7 mutations emerged as a determinant of life-threatening COVID-19 in men under 60 [25,26,27,28,29], even though TLR7 engagement in pDCs may also promote the macrophage-induced cytokine storm in COVID-19 patients [23]. Excess TLR7 or TLR8 activity, by contrast, can lead to sterile inflammation and autoimmunity, and is an important contributor to the pathogenesis of autoimmune syndromes [30, 31]. Its role in the development of autoimmunity was established with the aid of murine models, where the expression of two copies of Tlr7 or Tlr8 was sufficient to induce lupus-like manifestations [16, 31,32,33,34]. Introgression of TLR7 deficiency into genetic backgrounds predisposing to lupus conversely protected the mice against autoimmunity [35, 36]. Remarkably, dendritic cells of TLR8-deficient mice overexpressed Tlr7 and were hyperresponsive to TLR7 ligands, and the animals developed autoimmunity [37, 38]. These animal models prefigured the phenotypes for human mutations that have been described only recently. A missense mutation (p.Tyr264His) increasing TLR7 affinity for guanosine was carried by a female child suffering from severe lupus, and this dominant gain-of-function variant of TLR7 was sufficient to cause autoimmunity in both male and female transgenic mice [39]. Another study identified a missense TLR8 mutation leading to partial TLR8 deficiency in monozygotic twin boys with severe autoimmune hemolytic anemia and a TLR7-dependent autoinflammatory phenotype [40]. These human and animal studies call attention to TLR7 and TLR8 as bidirectionally dosage-sensitive genes underpinning a tonic level of immune activity between the extremes of defective antiviral defense and severe autoimmune pathogenesis.

The TLR7 and TLR8 genes lie within a narrow, 56-kb interval on the short arm of chromosome X (Fig. 1A, B). These paralogue genes arose from an ancestral autosomal gene that duplicated in the vertebrate line before the divergence of tetrapods from fishes ∼400 million years ago [41, 42]. In eutherian mammals, one of the two X chromosomes of female cells is randomly inactivated during early embryonic development to equalize the dosage of gene expression between the sexes. An early step of this process is the coating in cis of the X chromosome to undergo inactivation by the long noncoding RNA, XIST [43]. In men with Klinefelter syndrome (KS), who carry an aneuploid karyotype encompassing one or more supernumerary X chromosomes, all but one X are similarly inactivated [44]. Around 15% to 23% of human X-linked genes escape X chromosome inactivation (XCI) to a variable extent. Some of these genes escape XCI in a constitutive manner, whereas a majority are defined as facultative escapees as they evade XCI in certain cells, in some tissues or in some individuals [45,46,47]. We previously demonstrated that the TLR7 gene can escape XCI and is therefore transcribed in bi-allelic fashion in a fraction of female immune cells in all individuals tested [48, 49] but also in immune cells from 47,XXY KS males [48]. Interestingly, KS men have an equivalent risk to women to develop relatively rare immune disorders such as systemic lupus erythematosus (SLE) [50], Sjögren’s syndrome [51] or systemic sclerosis [52], suggesting a dominant role of X-linked genetic effects over sex hormones in autoimmune disease susceptibility. Whether TLR8 also evades XCI, however, has not yet been explored. The close proximity between the two genes, and the homology and similarity of function between the encoded receptors, warranted investigating whether TLR8, mirroring TLR7, could be transcribed on the inactive X chromosome (Xi) of the immune cells of women and Klinefelter syndrome men.

Fig. 1

3D conformation of the TLR7–TLR8 genomic region in male and female cells. A Idiogram of the human X chromosome with the positions of the adjacent genes, TLR7 and TLR8, and three marker genes, CFP, MSN and PGK1, that do not evade X chromosome inactivation. B Map of the TLR7, TLR8-AS1 and TLR8 locus on Xp22.2; tel, cen denote the telomeric and centromeric ends of the map. C Hi-C map of interactions around the TLR7–TLR8 region on the active (Xa, top panel) or inactive (Xi, bottom panel) X chromosome at 5-kb resolution in GM12878 human female cells. D Hi-C map of interactions around the TLR7–TLR8 region at 1-kb resolution in GM12878 human female cells (datasets in C and D from Rao et al. [65]). The dark blue circle denotes a peak of interaction (loop anchor) between TLR7 and PRPS2 as determined by the Peak tool in the Juicebox software. Blue and yellow dashed lines represent domains of preferential interactions as defined by the Contact Domain tool in Juicebox. E, F Hi-C maps of interactions around the TLR7–TLR8 region at 5-kb resolution in human male monocytes E, F, top; dataset from Phanstiel et al., [66]) versus GM12878 human female cells (E, bottom, dataset from Rao et al., [65]) or male T cells (F, bottom panel, dataset from Zhang et al., [67]). Blue dashed lines represent domains of preferential interactions as defined by the Contact Domain tool in Juicebox

We have studied here CD14⁺ monocytes and CD4⁺ T lymphocytes, where both TLR7 and TLR8 are expressed [18, 53,54,55], using RNA fluorescence in situ hybridization (RNA FISH) to visualize on a single-cell basis the primary transcripts of TLR7 and TLR8 relative to X chromosome territories. We provide evidence that TLR8, like TLR7, escapes XCI in CD4⁺ T cells and monocytes. In addition, we have gathered evidence for co-dependent transcription of the TLR7 and TLR8 genes on the active X chromosome (Xa) of the cells from women and KS men, in a comparison with euploid men (46,XY). Importantly, we found TLR8 protein expression to be up-regulated in female immune cells, including monocytes, suggesting that both XCI escape and co-dependent transcription could enhance TLR8 protein abundance in females.

Materials and methods

Donors and ethical compliance

This study complied with the ethical principles of the Declaration of Helsinki, and with applicable French regulations. Peripheral blood mononuclear cells (PBMCs) of anonymous, healthy blood donors at the Toulouse blood transfusion center (Etablissement Français du Sang) were from an in-house biobank approved by the competent ethics board (Comité de Protection des Personnes Sud-Ouest et Outre-Mer II, Toulouse) under reference 2–15-36. Klinefelter syndrome males presented with a 47,XXY karyotype, and were aged 16 to 44 at the time of sample collection. These patients were enrolled at the Toulouse University Hospital, and their study was approved by the aforementioned ethics board under reference 1–16-28. Written consent was obtained from each patient or, for child participants, from the legal guardian.

Isolation of CD14⁺ monocytes and CD4⁺ T lymphocytes

Upon thawing, PBMCs from healthy (female n = 6; males n = 7) or Klinefelter syndrome donors (n = 5) were cultured overnight in RPMI 1640 medium supplemented with 2 mM L-glutamine, 0.5 mM sodium pyruvate, 5 mM HEPES, and non-essential amino acids (all from Life Technologies); 100 U/ml penicillin and 100 µg/ml streptomycin (BioWhittaker); 5% human AB serum (Gemini Bio-Products); and, for T cell work, 10 ng/ml human recombinant IL-2. Classical CD14⁺ monocytes and CD4⁺ T cells were, respectively, isolated with EasySep Human CD14 Positive Selection and Human CD4 Negative Selection kits from STEMCELL Technologies. Monocytes were cultured in 96-well polypropylene plates to avoid non-specific activation arising from adhesion to polystyrene. Naive CD4⁺ T cells were seeded into 24-well plates (10⁶ cells/well), and stimulated for three days in the presence of 10 ng/ml recombinant human IL-2 and 12.5 µl/ml ImmunoCult Human CD3/CD28 T Cell Activator (STEMCELL Technologies).

Flow cytometry

Human immune cells were stained with fluorescent mouse monoclonal antibody (mAb) conjugates against specific surface markers to study CD4⁺ T cell activation: PE-Cy5 anti-human CD69, clone NF50; and BV421 anti-human CD25, clone M-A251 (both from from BD Biosciences). Flow cytometry analysis was performed on a BD Biosciences LSR II flow cytometer or FACSAria II cell sorter, and the data were processed with the FlowJo software (FlowJo LLC).

For intracellular TLR8 staining, freshly thawed PBMCs (5 × 10⁶ cells/ml) from age-matched male and female healthy subjects from our in-house biobank were surfaced stained with PE-Vio615 conjugated Lin-specific mAb (anti-CD3, cloneBW264/56; anti-CD19, clone LT19; and anti-CD56, clone REA196; all from Miltenyi Biotec), anti-CD14-PB (clone REA599, Miltenyi Biotec) and anti-CD16-AF700 (clone 5C3, BD Biosciences). Cells were then fixed and intracellularly stained using fixation and permeabilization buffers (#00-5123-43 and #00-8333-56, from eBioscience) with anti-TLR8-PE and anti-TLR8-APC mAb (clone S16018A, BioLegend). Flow cytometry analysis was performed on a BD Biosciences Fortessa instrument, and monocyte subsets defined as Lin⁻ CD14⁺CD16⁻ (classical monocyte), CD14⁺CD16⁺ (intermediate monocyte), and CD14⁻CD16⁺ (non-classical monocyte). Data were processed with the FlowJo 10.9 software.

Quantification of TLR8 protein expression by western blotting

PBMCs were thawed and cultured at 37 °C for 2 h before counting, cell lysate preparation, and western blotting as below. Cell lysates were prepared in Laemmli sample buffer (Invitrogen), sonicated, and total protein quantified by a bicinchoninic acid protein assay (Pierce). Samples were heated for 10 min at 70 °C in the presence of a reducing agent (Invitrogen). Protein (20–25 µg per lane) were fractionated by SDS-PAGE on precast 4–15% gradient Stainfree gels (Bio-Rad), and transferred to Amersham Hybond 0.45-µm PVDF membranes (GE Healthcare). The membranes were blocked with 5% skim milk, 0.1% Tween-20 in PBS, probed overnight with the anti-N-terminus (LRR1)-specific anti-human TLR8 mAb (rabbit monoclonal IgG, clone D3Z6J #11886, Cell Signaling Technology), and finally incubated with suitable peroxidase-conjugated secondary antibodies (anti-rabbit IgG, #7074, Cell Signaling Technology). Chemiluminescent detection was carried out with Amersham ECL Prime reagent (GE Healthcare), and densitometric analysis performed with the Image Lab 5.0 software (Bio-Rad). TLR8 protein quantification was performed as described [56]. The densitometric signals of the full-length TLR8 120-kDa forms were normalized to total protein and then to an internal standard PBMC lysate that was loaded in each gel for inter-gel data normalization as described [56].

RNA FISH probes

Probes for RNA FISH were prepared by PCR amplification of human genomic DNA fragments using the primer sets in Additional file 1: Table S1, targeting exon 1 of XIST (MIM *314670), and both exon and intron regions of TLR7 (Xp22.2; MIM *300365), TLR8 (Xp22.2; MIM *300366), CFP (Xp11.23; MIM *300383), MSN (Xq12; MIM *309845) and PGK1 (Xq21.1; MIM *311800). The XIST and TLR7 probes have been described previously [48]. To exclude repetitive DNA from the PCR amplimers, genomic sequences were filtered in silico with the RepeatMasker tool [57] prior to primer design. Probes were fluorescently labeled using the Vysis Nick Translation kit (Abbott) according to the manufacturer’s instructions, and any of the following dUTP conjugates: aminoallyl-dUTP-ATTO-655, aminoallyl-dUTP-ATTO-550, or aminoallyl-dUTP-XX-ATTO-488 (all from Jena Bioscience). The Xa marker genes, CFP, MSN and PGK1, were identified as non-escape genes in the scientific literature [47, 58], supported by additional molecular genetics information from the OMIM database at https://omim.org/, and monocyte expression data from the EBI-EMBL Expression Atlas at https://www.ebi.ac.uk/gxa/home.

RNA FISH

RNA FISH was performed as described in our earlier reports [48, 59]. Briefly, spreads of monocytes or T lymphocytes on poly-L-lysine-coated coverslips were fixed for 10 min with 3% paraformaldehyde at room temperature, and permeabilized for 7 min in ice-cold cytoskeletal buffer containing 0.5% Triton X-100 and 2 mM vanadyl-ribonucleoside complex (New England Biolabs). The cells were dehydrated through successive ethanol baths, air-dried briefly, and incubated with the labeled probes overnight at 42 °C. The coverslips were rinsed twice with 50% formamide in 2 × SSC (saline sodium citrate) and thrice with 2 × SSC alone, and nuclei were counterstained with DAPI in phosphate buffered saline. The coverslips were slide-mounted using Dako fluorescence mounting medium before microscopy on a Leica TCS SP8 or Zeiss LSM710 microscope using a 63 × oil immersion objective. Image data were processed with the Fiji software (https://fiji.sc/).

Cell scoring

In RNA FISH experiments, individual cells were scored as positive or negative for the gene of interest depending on the detection or non-detection of primary transcript foci under microscopic examination. To quantitate the escape from XCI for TLR7 or TLR8, we counted the cells displaying either bi-allelic transcriptional foci or a single signal on the Xi, and expressed the percentage of escape cells with reference to the number total of cells positive for the gene under consideration. In estimating the overall frequencies of TLR7, TLR8, and joint TLR7 and TLR8 transcriptional foci, cells were scored by two alternative approaches: either (i) in random microscopic fields, by considering all cells regardless of signals from the Xa marker probe, or (ii) by including only those cells marked by the Xa probe.

Statistical analysis

For TLR7 and TLR8 considered together, cell counts were cross-classified on a per-donor basis as a 2 × 2 contingency table with TLR7 and TLR8 signals as nominal variables with positive or negative outcomes as above. Statistical analysis was then performed in the R computing environment [60]. From each 2 × 2 table classifying N cells total,

	TLR7+	TLR7−
TLR8+	a	b
TLR8−	c	d

we derived descriptive statistics, i.e., the percentage of scored cells in a stratum of interest, and also a measure of association between TLR7 and TLR8 transcriptional signals, namely Yule’s Q coefficient of association in 2 × 2 tables [61] as given by Eq. (1).

$$\begin{array}{c}Q=\frac{ad-bc}{ad+bc}\end{array}$$

(1)

In parallel, we considered the ratio of the observed frequency of TLR7⁺ TLR8⁺ cells to the frequency expected under the null hypothesis of independent transcription of the two genes. We computed this obs/exp ratio according to Eqs. (2) and (3), where the probability p_7·8 of observing transcriptional foci for TLR7 and TLR8 together is given, under independence, by the product of the respective probabilities of observing TLR7 and TLR8 signals.

$$\begin{array}{c}{p}_{7\cdot 8}={p}_{7}{p}_{8}=\frac{a+c}{N}\cdot \frac{a+b}{N}\end{array}$$

(2)

$$\begin{array}{c}\frac{obs}{exp}=\frac{a}{{Np}_{7\cdot 8}}=\frac{aN}{\left(a+c\right)\left(a+b\right)}\end{array}$$

(3)

We summarized the RNA FISH data for a given group by a statistical procedure built around R library gmeta [62], which performs meta-analysis based on the mathematics of confidence distributions (CDs) [63]. We generated a CD for each statistic under consideration and for each individual, and fed the individual CDs together to the gmeta function for summarization under a random effects model. CDs for a proportion (such as percent cells evading XCI, or positive for both TLR7 and TLR8 transcripts) were computed by a binomial approach; CDs for the analytical statistics, Yule’s Q and the obs/exp ratio, were estimated non-parametrically by bootstrap resampling with 10,000 or 50,000 replications. The software computed a point estimate and 95% CI from each individual CD, and pooled the individual CDs into a summary CD to derive a group’s meta-analytical mean and its 95% CI.

To test for independence in 3 × 2 tables encompassing sparse cells (Additional file 1: S5 Data), we performed χ² tests in R by a Monte Carlo procedure with 10⁶ replications. When carrying out between-groups comparisons, e.g., males versus females, we computed a p-value to test the difference between the group summaries for the statistic under consideration. For this, we determined the probability for the group-A meta-analytical mean according to the CD for group B, and the probability for the group-B meta-analytical mean according to the CD for group A; we took the greater of the two one-tailed p-values, and multiplied it × 2 to approximate a two-tailed hypothesis test. We computed a similar, two-tailed CD-based p-value to test whether a group’s summary for a measure of association (Yule’s Q, or the obs/exp ratio) diverged from the null value denoting independence, by performing a single comparison between the corresponding null value (Q = 0, or obs/exp = 1) and the group’s summary CD for this statistic. Our R scripts performing two- and three-group meta-analytical summarization, hypothesis tests, and plotting are available from the Zenodo repository under digital object identifiers (DOIs) https://0-doi-org.brum.beds.ac.uk/10.5281/zenodo.6580369 and https://0-doi-org.brum.beds.ac.uk/10.5281/zenodo.6580378. We performed the Pearson correlation analyses and statistical tests in Additional file 2: Fig. S2, Fig. S4 and Fig. S8, and in Fig. 7 with GraphPad Prism 7.0a from GraphPad Software.

Hi-C dataset analysis

We used the Juicebox software [64], including the Contact Domains and Peaks calling tools, at http://www.aidenlab.org/juicebox/ to access and visualize the following chromatin conformation datasets: (i) data for human female lymphoblastoid cell line GM12878 from Rao and coll. [65], Gene Expression Omnibus (GEO) Accession GSE63525; (ii) data for male human monocytic cell line THP-1 from Phanstiel and coll. [66], Sequence Read Archive (SRA) accession PRJNA385337; and (iii) data for non-activated, human primary CD3⁺ T cells from Zhang and coll. [67], GEO accession GSE104579.

Results

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 5-kb 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 Additional file 2: 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 (1-kb) 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 noncoding RNA XIST (reference [48], and Additional file 1: 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 [70,71,72]. 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 (Additional file 1: 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 (Additional file 2: 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% on average (Additional file 2: 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 D). 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; Additional file 1: 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% to 34% of cells, with a group mean of 17% (Fig. 2F).

Fig. 2

TLR7 and TLR8 evade X chromosome inactivation in the monocytes and CD4⁺ T cells of women. A–D RNA FISH analysis of CD14⁺ monocytes from female donors. The images show confocal microscopy planes of cell nuclei after hybridization with fluorescent probes for transcripts arising from TLR7 (green), TLR8 (red), and the marker gene triad transcribed from the Xa only (PGK1, CFP, MSN; pink). Nuclei are counterstained with 4ʹ,6-diamidino-2-phenylindole (DAPI; blue). The arrow heads in A, C indicate transcript foci from the two alleles of TLR7 or TLR8. In B, D, the two genes are transcribed only from the Xi (arrowheads). The TLR7 or TLR8 hybridization pattern is schematized to the right of each row. E–H Quantification of the escape from XCI for TLR7 and TLR8 from monocytes (E, F) and CD4⁺ T cells (G, H). The forest plots show the percentage of XCI escape in monocytes or T cells and its 95% confidence interval (CI) in individual female donors (n = 6); the red diamond denotes the meta-analytical group mean and 95% CI. I, J Confocal microscopy planes of cell nuclei after RNA FISH with TLR7 (I, pink), TLR8 (J, red), and XIST (green, painting the Xi) probes. Nuclei are counterstained with DAPI (blue). The TLR7 and TLR8 transcriptional foci adjacent to the Xi territory confirm that both genes evade XCI

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 (Additional file 2: Fig S3B). We verified activation on the third day using flow cytometry analysis for CD25 and CD69 levels (Additional file 2: Fig S3C). As described earlier for monocytes, we concluded to XCI escape in CD4⁺ T cells from women based on the presence of TLR7 (Additional file 2: Fig S3D) or TLR8 (Additional file 2: 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% to 11% of T cells, with a group mean of 8% (Fig. 2G and Additional file 1: S1 Data). For TLR8, the frequency of escape cells varied over a 4% to 20% range, with a group mean of 8% (Fig. 2H and Additional file 1: 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; Additional file 2: Fig S4A) nor in CD4 T cells (r =  − 0.06; p = 0.92; Additional file 2: 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% to 17% of KS male monocytes (a narrower range than in women), with a group mean of 13% (Fig. 3B and Additional file 1: 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; Additional file 2: Fig S4B).

Fig. 3

TLR7 and TLR8 evade X chromosome inactivation in the monocytes of men with Klinefelter syndrome. A RNA FISH analysis of CD14⁺ monocytes from men with KS (47,XXY). Confocal microscopy planes of cell nuclei hybridized with fluorescent probes for transcripts arising from TLR8 (red), TLR7 (green), and the Xa marker genes (pink). Nuclei are counterstained with DAPI (blue). The arrow heads indicate TLR8 (top) or TLR7 (bottom) transcript foci occurring either on both X chromosomes (right) or on the Xi only (left). B, C Quantification of XCI escape for TLR7 and TLR8. The forest plots show the percentage of XCI escape and its 95% CI in KS men (n = 5). The meta-analytical group mean and 95% CI is denoted by the red diamond

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 (Additional file 2: 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 (Additional file 2: 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 (Additional file 2: 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) (Additional file 2: Fig S5C). With regard to TLR8 transcripts, the greatest contrast occurred between the groups of euploid 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% (Additional file 2: 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 (Additional file 2: Fig S6A and S6B, Additional file 1: 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 (Additional file 2: 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 sevenfold increase relative to XY men (Figs. 4D–G and 5A; p < 0.0001). Monocytes from XXY KS men exhibited an intermediate group mean (9% of TLR7⁺ TLR8⁺ cells; Figs. 4G and 5A).

Fig. 4

Examples of TLR7 and TLR8 combined transcription profiles in euploid male and female monocytes. RNA FISH analysis of CD14⁺ monocytes from male (A–C) and female (D–F) donors. Confocal microscopy planes of cell nuclei after RNA FISH for TLR7 (green), TLR8 (red), and the Xa marker (pink). Nuclei are counterstained with DAPI (blue). A, B Detection of TLR7 or TLR8 transcripts from the single X of male cells. C, D Simultaneous transcription of TLR7 and TLR8 on the male X or on the female Xa. E, F TLR7 and TLR8 co-transcription on the female Xa with concomitant TLR7 (E) or TLR8 (F) transcriptional activity on the Xi. G Schematics of the different patterns of TLR7 and TLR8 transcription in our RNA FISH data. The percentages of nuclei (mean ± SD) in females (n = 6) and males (n = 7) for each RNA FISH profile are shown

Fig. 5

TLR7 and TLR8 are transcriptionally non-independent in monocytes. Quantitative analysis of RNA FISH experiments on monocytes from XX women, XY men, and XXY KS men. For each donor, individual cells were scored positive or negative for TLR7 and for TLR8 transcripts, and cell counts accordingly cross-classified as a 2 × 2 contingency table. Descriptive and analytical statistics were computed from each table (Additional file 1: S3 Data), and summarized group-wise by meta-analysis. The forest plots display the statistics of interest for each donor and its 95% CI (dots and whiskers), and the meta-analytical group means (diamonds) and their 95% CIs (whiskers and shaded areas). In A, C and E, cells were scored regardless of the chromosome of origin of the TLR7 and TLR8 transcripts (Any X). In B, D and F, only cells positive for the Xa probe (Xa⁺) were counted, and the data are restricted to TLR7 and TLR8 signals observed on Xa⁺ male or female X chromosomes. A, B Percentage of cells positive for both TLR7 and TLR8 transcripts. C, D Analysis for the obs/exp ratio of the observed number of double-positive cells to the number of cells in this category expected under the hypothesis of independent transcription of TLR7 and TLR8. E, F Analysis for Yule’s Q coefficient of association in 2 × 2 tables. The p-values in A, B test the differences between group means; in C–F, each p-value tests the divergence of a group summary value relative to the critical value, obs/exp = 1 or Q = 0; the two-tailed p-values were derived from the group CDs in the meta-analytical summarization. Women and XY men display significant deviations from Q = 0, of opposite signs, signifying transcriptional non-independence of TLR7 and TLR8

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 sevenfold 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 (Additional file 2: Fig S5E). Reciprocally, the frequency of monogenic expression (either TLR7 or TLR8) was twofold 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%) (Additional file 2: 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 (Additional file 1: 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 (Additional file 1: 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 was 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 (Additional file 1: S2 Data). To analyze these sparse data for Yule’s Q, we pooled the cell counts group-wise to increase statistical power (Additional file 1: S2 Data). Additional file 2: 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 (Additional file 2: Fig S7B). This result points to the absence of transcriptional association between the two genes on the Xi (Additional file 2: 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 Additional file 1: 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 (Additional file 1: 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 (Additional file 2: 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 (Additional file 1: 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.

Fig. 6

TLR7 and TLR8 are transcriptionally non-independent in CD4⁺ T cells. Quantitative analysis of RNA FISH experiments on stimulated CD4⁺ T cells from women and XY men. Cell counts from RNA FISH experiments were processed as in Fig. 4. The forest plots display the statistic of interest for each donor and its 95% CI (dots and whiskers), together with the meta-analytical group means (diamonds) and their 95% CIs (whiskers and shaded areas). In A–C, T cells were scored regardless of the chromosome of origin of the TLR7 and TLR8 transcripts (Any X). In D–F, only cells positive for the Xa probe (Xa⁺) were counted, and the data are restricted to TLR7 and TLR8 signals observed on those Xa⁺ male or female chromosomes. A, D Percentage of T cells positive for both TLR7 and TLR8 transcripts. B, E Analysis for the obs/exp ratio of the observed number of double-positive cells to the number of cells in this category expected under the hypothesis of independent transcription of TLR7 and TLR8. C, F Analysis for Yule’s Q coefficient of association in 2 × 2 tables. The p-values in A, D test the differences between group means; in B, C, E, and F, each p-value tests the divergence of a group summary value relative to the critical value, obs/exp = 1 or Q = 0; the two-tailed p-values were derived from the group CDs in the meta-analytical summarization. Women and euploid men display deviations of opposite signs from Q = 0, signifying transcriptional non-independence of TLR7 and TLR8

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 [48, 56]. We performed a similar analysis by comparing TLR8 expression between male and female PBMCs using a highly specific anti-TLR8 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, Additional file 2: Fig S8A, B). 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, Additional file 2: 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.

Fig. 7

TLR8 protein is expressed at a higher level in female than in male immune cells. A Western blot analysis of TLR8 expression (upper panel) and total protein staining as control (lower panel). B Densitometric signals from the full-length (120-kDa) form of TLR8 were normalized to total protein and then to an internal standard PBMC lysate that was loaded in each gel for inter-gel data normalization. Results for male (n = 24) and female (n = 21) donors are shown, and data were pooled from four independent experiments. C Cytometrically determined frequency of CD14⁺ monocytes among PBMCs from the same donors. D Classical monocytes were defined as Lin⁻ CD14⁺ CD16⁻, intermediate Lin⁻ CD14⁺ CD16⁺ and non-classical Lin⁻ CD14⁻ CD16⁺ cells as shown in Additional file 2: Fig. S8A. TLR8 expression was measured by intracellular co-staining with anti-TLR8-PE and anti-TLR8-APC antibodies as shown in Additional file 2: Fig. S8A. D, E Anti-TLR8-PE staining intensity, gated on double TLR8-positive cells was calculated and expressed as ΔGMFI with background staining subtracted. D Representative histogram profiles from male and female donors are shown. E Data from individual male (n = 10) and female (n = 10) donors were pooled from two independent experiments. Statistical differences between groups were analyzed using the Mann and Whitney test; actual p-values are shown

Discussion

Here, we conducted a comprehensive single-cell resolution analysis of the transcriptional regulation of the X-linked genes encoding RNA-specific Toll-like receptors, TLR7 and TLR8, in CD14⁺ monocytes and CD4⁺ T lymphocytes, where both receptors are expressed. By analyzing for primary transcript expression relative to the Xa and Xi chromosome territories, we unequivocally demonstrated that TLR8, like TLR7, escapes XCI in female and in 47, XXY KS male monocytes. We show also that both genes escape XCI in a substantial proportion of T cells despite lower mRNA expression levels than in monocytes [16, 73], with TLR7 and TLR8 bi-allelic cells observable in all our female and KS male donors. When we analyzed TLR7 and TLR8 transcripts together, distinct co-expression profiles emerged between our three study groups. Surprisingly, these differences were attributable not only to the ability of females and KS males to express TLR7 and TLR8 on the Xi, but also to the joint transcriptional behavior of the TLR7–TLR8 gene pair on the active X chromosome specifically. Monocytes from women and KS men exhibited higher-than-expected frequencies of cells co-transcribing the two genes, in contrast to a striking trend against simultaneous transcription of the TLR7 and TLR8 genes on the single X of euploid men. Corroborating our RNA FISH results, we found higher TLR8 protein expression in female than in male leukocytes, including all monocyte subpopulations. In summary, our findings provide compelling evidence for sex-specific transcriptional regulation of the X-linked TLR locus on the active X of healthy subjects, which may have important consequences for the functional make-up of monocyte and T cell populations.

The key findings of the present work are thus the marked association between the transcription of TLR7 and that of TLR8 in single cells, and that this association is of positive sign in women and negative in normal men. Simultaneous transcription of TLR7 and TLR8 was disproportionally frequent in female monocytes and T cells, and disproportionally scarce in the cells from normal men, and we traced this back to the behavior of the alleles carried in cis on the Xa of women and on the single X chromosome of euploid men. In some of our KS patients, by contrast, there was discordance between the whole-cell and Xa-specific patterns of TLR7–TLR8 transcription, which could have arisen from undetected 46,XY/47,XXY mosaicism [74]. Overall, our findings imply that the Xa in XX and XXY karyotypes, and the single X of normal men are functionally non-equivalent as regards the TLR genes. Besides, transcriptional non-independence between TLR7 and TLR8 indicates that these genes, which encode the two isoforms of the human TLR for single-stranded RNA, form a co-regulated gene cluster. We propose to name this cluster the X-linked TLR locus to acknowledge this previously unsuspected layer of transcriptional regulation.

Because RNA FISH provides a single-cell snapshot of the transcriptional state of each allele at the time of cell harvest, our data must be interpreted within the notion that genes are not transcribed continuously but in bursts or pulses separated by silent periods [75]. Alleles permissive for transcription but analyzed between pulses may thus appear negative, and hybridization-positive cell counts probably underestimate the proportion of expressing cells or the frequency of escape from XCI. This likely explains that, in female monocytes, the average frequency for TLR7 escape determined here by RNA FISH is lower (13.3% versus 30%) than in our previous study, which employed instead reverse transcription and PCR amplification on the mRNA pool of individual cells [48]. Regardless of the technique employed, our results are compatible with the notion that the TLR7 and TLR8 genes on the Xi may not be poised for transcription in all cells, and that this status may be relatively stable within cells as previously suggested [48, 76]. In favor of this hypothesis, we previously provided evidence for a causal link between X chromosome dosage, XCI escape and increased functional TLR7-driven responses in human B cells [48] and pDCs [77], where bi-allelic TLR7 expression at single-cell resolution was strongly correlated with enhanced responsiveness to cognate ligands. Along the same line, female pDCs with bi-allelic TLR7 transcription displayed heightened basal levels of mRNA for IFN-α and IFN-β, which suggests that baseline TLR7 sensing of endogenous ligands contributes to the functional heterogeneity of pDCs [77, 78]. This mechanism and its effects should be investigated in further human immune cell populations, such as monocytes and macrophages, in respect of TLR7 but also TLR8. A relevant study in mice has recently identified a robust sexual dimorphism in tissue macrophages [79] by using transcriptome and ATAC (assay for transposase-accessible chromatin) profiling of untreated and IFN-induced immune cells. This study suggested that female macrophages exhibited a higher basal potential of innate immunity pathways prior to immune challenge, and that the resulting enhancement of immune alertness made females less susceptible to infectious diseases [79]. Future work should delve deeper into the heterogeneity of these and further immune cell populations, such as T lymphocytes, and possibly define the role of endogenous X-linked TLR signaling in the acquisition of this sex-specific functional heterogeneity.

We hypothesize that the versatile sex-specific transcriptional patterns in the X-linked TLR locus arise from either cis or trans-acting factors, or both. The significant difference between the Xa and the Xi of individual females regarding TLR7–TLR8 joint transcriptional patterns clearly argues for cis-acting factors. The analyses of available Hi-C datasets suggest that although TLR7 and TLR8 lie very close to each other on the X chromosome, and fall within the same domain of interactions, the two genes seem to form different sub-domains of interactions that may explain the versatility of the regulation of their expression, with conditions of co-regulation, or co-exclusion or independence, which could be influenced by sex-specific factors, genetic predisposition or the inflammatory environment.

A limitation of the present work is that RNA FISH is not informative on the past or future evolution of the cell, and also that the RNA FISH readout cannot at present be readily correlated with protein quantitation on a single-cell basis, whether by flow cytometry or by emerging techniques such as single-cell proteomics [80]. Of note, we previously reported greater levels of TLR7 protein in the PBMCs of women relative to men [48, 56], and the present study shows a similar trend regarding TLR8 protein expression, which positively correlates with the greater frequency of RNA FISH signals observed in female cells. Whether this difference translates into sex-specific responses to TLR8-specific ligands warrants further investigation. Although RNA FISH showed sexual karyotype-biased patterns of transcription at the X-linked TLR locus, this did not translate into specialized cell subpopulations with predominant expression of either TLR7 or TLR8 in males, as the frequencies of TLR8⁺ cells were similar between males and females. Another intriguing question, prompted by evidence of functional attenuation of TLR7 by TLR8 in mouse and human mutants [37, 38, 40], is whether the different patterns of co-transcription observed here determine a sex bias with regard to TLR8:TLR7 protein ratios in individual cells, and hence possibly different outcomes of the exposure to relevant ligands.

Perspectives and significance

This work constitutes, to our knowledge, an unprecedented single-cell exploration of the transcriptional activity of the X-linked TLR locus according to sex and the sexual karyotype. Together with our earlier study of the TLR7 gene [48], it provides proof that sex is a critical factor to consider in studying the function and biological significance of the two Toll-like receptors for single-stranded RNA. In addition to the immense worldwide toll of 76 autoimmune disorders [81], and many disabling and life-threatening microbial pathogens, these new insights into the biology of TLR7 and TLR8 are of translational significance to vaccine design, where development of TLR7 and TLR8 agonists as vaccine adjuvants constitutes a promising field of research [82].