Reduced levels of pro-inflammatory chemokines in FXS patients
Compared to healthy individuals, Ashwood et al. found that FXS patients with ASD exhibited higher plasma levels of IL-1a and IL-12p40 and reduced levels of CCL2, CCL5, CCL11 and CXCL10, but similar levels of IL-1b, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12p70, IL-13, IL-15, GM-CSF, IFN-g, TNF-a and CCL3 [17]. They also found that FXS patients without ASD exhibited higher plasma levels of IL-1a and reduced levels of CCL5 and CXCL10, but similar levels of the other biomarkers [17]. While seven interleukins, i.e. IL-1a, IL-1b, IL-2, IL-4, IL-5 and IL-13 were readily detected in plasma samples by Ashwood et al. using a different immunoassay, these IL were below the lower level of detection in our serum samples. In agreement with this previous study, we found that IL-6, IL-7, IL-8, IL-10, IL-12p40, IL-12p70, IL-15, GM-CSF, IFN-g and TNF-a were present at similar levels in FXS patients and healthy controls. However, we did not highlight an increase in IL-12p40 in FXS patients and we show a decrease in CCL3. These apparent discrepancies could be explained by the type of matrix used (plasma versus serum), the analytical platform used (bead-array versus ECL assay) or differences in the statistical methods used. However, both our study and the one by Ashwood et al. suggest that FXS patients do not exhibit a clear low-grade pro-inflammatory profile that would be characterized by higher levels of multiple pro-inflammatory cytokine and acute phase proteins such as CRP. In agreement with this observation in FXS patients, serum levels of the two pro-inflammatory cytokines TNF-α and IFN-γ were reported to be identical in Fmr1-KO mice and their non-transgenic littermates [25].
Chemokines are well known for their ability to induce directed chemotaxis in nearby responsive cells. Here, we have found that FXS patients exhibit reduced serum levels of CCL2, CCL3, CCL4, CCL11, CCL13, CCL17, CCL22, CCL26 and CXCL10. Furthermore, CCL2, CCL3, CCL11, CCL22, CCL26 and CXCL10 were negatively associated with FXS diagnosis in a robust regression model in which we adjusted for age, BMI and time of sampling. Five of these chemokines (CCL2, CCL11, CCL22, CCL26 and CXCL10) were also retained in a dataset reduced to male subjects, strengthening our findings for these chemokines. Finally, CCL2, CCL3, CCL11, CCL22, CCL26 and CXCL10 are also selected by our K-sparse clustering as features discriminating the clusters, in combination with 10 additional biomarkers. Importantly, a previous report highlighted that the plasmatic levels of CCL2, CCL11 and CXCL10 were significantly decreased in a larger sample of male FXS patients [17]. All these data support the fact that dysregulation of at least a subset of specific pro-inflammatory chemokines could contribute to FXS.
The subset of chemokines exhibiting reduced levels in FXS patients indicated possible decreased signalling from six chemokine receptors: CCR1, CCR2, CCR3, CCR4, CCR5 and CXCR3. This could possibly impact the chemotaxis of a number of cell types and pathways ontologically related to “inflammatory response”, including “response to viral infection”, “response to LPS” or “response to toxic insults”. CXCR3 is expressed primarily by activated NK cells and T lymphocytes, and by epithelial cells [26]. Th1 cells co-express CXCR3 and CCR5 while Th2 cells express CCR3 and CCR4. CXCR3 ligands that attract Th1 cells can concomitantly block the migration of Th2 cells in response to CCR3 ligands, thus enhancing the polarization of effector T cell recruitment [26]. Reduced serum levels of CXCL10 might dampen CXCR3 signalling, while decreased levels of CCL3 and CCL5 could reduce CCR3 and CCR4 signalling. This could ultimately cause T cell dysfunction in FXS patients. CCL2 is the most strongly dysregulated chemokine in FXS patients as compared to controls. Through signalling via CCR2 and CCR4 receptors, CCL2 mediates chemotaxis of monocytes and dendritic cells, as well as memory T cells to the sites of inflammation upon tissue injury or infection [27, 28]. Decreased levels of circulating CCL2 could therefore contribute to reduce the local immune response to infection in FXS patients.
A general health survey on FXS patients highlighted an increased occurrence of ear, throat and nose infections such as sinusitis and otitis in FXS patients [9, 10]. However, anatomical particularities and malformations of the ear conducts and sinuses are likely to explain the specificity of these infections [9], as FXS patients do not appear more sensitive to other types of infections. This would have been expected, if there was a general defective chemokine receptor signalling during infection in FXS patients. Nevertheless, it may be envisioned that reduced chemokine levels in FXS patients may be associated with a reduced capacity of the immune system of FXS patients to respond to specific pathogens. Of note, one study in the Drosophila melanogaster model of FXS showed that the dfmr1 mutants exhibit increased sensitivity to bacterial infection and decreased phagocytosis of bacteria by systemic immune cells [29], suggesting that dfmr1 gene is required for the activation of phagocytic immune cells and therefore for their immune responsiveness.
Although immune responsiveness to LPS and PHA of PBMC from FXS patients did not differ from those of healthy controls [18], a few studies have been performed in asymptomatic individuals carrying the FMR1 premutation. Notably, an increase in IL-10 secretion by PBMC in the absence of immune challenge was described in FMR1 premutation carriers [30]. Furthermore, a decreased cytokine secretion response was observed in LPS-stimulated PBMC from individuals carrying the FMR1 premutation [31]. This suggests that dysregulation in the FMR1 gene could alter immune responsiveness under yet to be defined conditions. FMRP, the FMR1 gene product, is an RNA-binding protein and a translational regulator [6]. Although majorly described as a translational repressor, some studies suggest that FMRP can activate the translation of some of its mRNA targets [6]. Further work would be required to determine whether the mRNAs encoding the six dysregulated cytokines are bona fide mRNA targets for translational activation by FMRP or whether their decreased levels indirectly result from compensatory or adaptive mechanisms.
Possible neuro-immune alterations in FXS
Interactions between the nervous and the immune system are critical not only during early neurodevelopment but also in adolescence and adulthood [32]. As a consequence, immune dysfunction may cause changes in brain connectivity associated with neurodevelopmental disorders and this has been mostly studied in the context of ASD [33]. Chemokine and their cognate receptors are widely expressed in the developing and adult CNS and disruption of their patterns of expression have been involved in CNS disorders, including ASD [34]. Notably, CCL2, CCL3, CCL4, CCL5 and CCL11 are required, via signalling through CCR2, CCR3 and CCR5 receptors, for microglia chemotaxis [35-37]. In addition, CCL2 regulates migration of neural stem cells in the brain [38]. Although peripheral levels of those chemokines might not reflect their actual levels in the CNS, alterations in CCL2, CCL3, CCL4, CCL5 and CCL11 signalling could participate to the defects in CNS patterning observed in FXS patients. In addition, alterations in chemokine secretion or immune response could enhance the sensitivity of FXS patients to neurological damages induced by environmental sources, including infections and xenobiotics exposure. In line with this, and although chemokines were not assessed, one study has shown that cortical astrocytes derived from the brain of FXS mouse model secreted more IL-6 in response to LPS stimulation. Furthermore, the authors provide evidence that the abnormal elevation of IL-6 in the cortex of Fmr1-KO mouse could be linked to the synaptic phenotypes [39].
Immune-related biomarkers enable clustering of samples in FXS and control cases
The K-sparse clustering strategy we applied to the biomarkers’ dataset enabled discrimination of FXS samples from control samples relying principally on a combination of 16 immune-related markers. A number of efficient methods already exist to perform unsupervised classification of samples based on datasets encapsulating biological features. It is common practice to use PCA k-means to perform clustering analysis. k-means does not perform both clustering and feature selection, therefore providing minimal insights into the discriminating features and therefore into the underlying biology. The K-sparse method we have used alternates k-means with projection-gradient minimization to promote sparsity and enable features selection, with significant improvements in clustering performances as compared to k-means standard algorithm in terms of clustering performances [40]. By considering the weighted linear combination of the selected features, the clustering computed by K-sparse also presented the advantage of handling extreme values and outliers which are frequently observed in biomarker datasets. The K-sparse clustering highlighted that 8 of the 9 significantly dysregulated chemokines contribute to separation between controls and FXS cases. It further identified additional immune-related molecules with discriminative abilities: CCL19, CXCL11, IL-12p40, IL-12p70, IL-17F, IL-21, INF-g and SAA. K-sparse highlights the contribution of biomarkers which do not differ in univariate analysis and regression analysis. This can be explained by the fact that K-sparse relies on weighted linear combination of the selected features and not on the individual distribution of values among each class. Correlations are observed between a subset of the 16 biomarkers selected by K-sparse, showing that relationships between biomarkers can be of biological relevance. Our study supports that K-sparse clustering can complement classical univariate and regression analyses to identify relevant biomarkers of disease. It paves the way for the use of K-sparse clustering analysis for the identification of combinations of disease biomarkers, but also for the stratification of patients in homogenous subtypes bearing similar biological patterns.
Limitations of the study
Our study has some limitations. First, the sample size is relatively small (n=25 FXS patients; n=29 sex and age-matched controls), which may lead to an overestimation of the effect size owing to reduced power, but also limit the generalization of our findings. Second, the methodological choices we have made based on out dataset constraints, e.g. the use of robust Elastic Net regression instead of classical logistic regression, may limit the interpretation of the identified associations. Indeed, coefficients are shrunk and computation of confidence interval and asymptotically valid p-values are not yet available in the robust regression framework. However, among the six chemokines that we have shown here to be negatively associated with FXS, three (CCL2, CCL11, CXCL10) have already been demonstrated to be present at lower levels in an independent cohort of FXS patients (n=64) compared to healthy controls (n=19) [17], therefore strengthening the validity of our conclusions. Third, although we have adjusted the associations between specific chemokines and FXS diagnosis for a number of covariates which could impact serum cytokines (age, BMI and time of sampling), we cannot rule out the contribution of additional unmeasured covariates. Fourth, we acknowledge that the present study provides limited insight into possible underlying mechanisms. However, it is noteworthy that almost nothing is known on possible immune-related dysfunctions in FXS patients. Exploratory studies are therefore needed to provide the rationale for new studies investigating specific underlying mechanisms.