0.50.60.70.80.91.0
Gene Dendrogram and Module Colors (Beta=12)
hclust (*, "average") as.dist(dissTOMPre)
Height
Module colors
HIV
Figure 3.5: Association of maternal HIV with co-expressed gene clusters (i.e., modules) identified by WGCNA. Association of maternal HIV with co-expressed gene clusters (i.e., modules) identified by Weighted Gene Co-expression Network Analysis (WGCNA). WGCNA was performed on the expression of 10,705 genes of 144 samples. This figure represents a gene dendrogram (a tree representation of the co-expressed genes) at top, a module colour band at the right after the tree and a gene-phenotype (i.e., maternal HIV) colour band at the bottom. The module colour band indicates 14 different modules and the gene-phenotype colour band shows the association of maternal HIV with different modules. In gene-phenotype colour band red represents strong positive association and blue represents strong negative association with maternal HIV exposure.
analysis using the hub genes from M1 and M14 are reported in Figures 3.8(a) and 3.8)(b). These results demonstrate the relationship of some of the hub genes with the offspring lung function at a later part of their life, which is observed through these genes’ impacts on some immune system oriented REACTOME pathways.
ME−Trait Relationships
−0.2
−0.1 0 0.1 0.2
gestation_deliv ery
HIVn ptsd_n
gender_nalcohol_n batch_n ethnicity_n MEblue
MEgreen MEturquoise MEblack MEbrown MEmagenta MEcyan MEpink MEred MEgreenyellow MEsalmon MEpurple MEyellow MEgrey
0.111 (0.2)
−0.0615 (0.5)
0.0157 (0.9)
−0.154 (0.07)
−0.075 (0.4)
−0.137 (0.1)
−0.0424 (0.6)
0.00685 (0.9)
0.0911 (0.3)
0.125 (0.1)
−0.082 (0.3)
0.101 (0.2)
−0.157 (0.06)
−0.024 (0.8)
0.0743 (0.4)
−0.0171 (0.8)
0.0134 (0.9)
−0.029 (0.7)
0.0524 (0.5)
−0.0313 (0.7)
−0.031 (0.7)
−0.0736 (0.4)
0.159 (0.06)
0.0818 (0.3)
−0.00149 (1)
0.0722 (0.4)
0.0725 (0.4)
−0.0267 (0.8)
−0.0684 (0.4)
−0.0218 (0.8)
0.114 (0.2)
0.0187 (0.8)
−0.142 (0.09)
−0.0583 (0.5)
−0.0363 (0.7)
−0.107 (0.2)
0.0363 (0.7)
0.126 (0.1)
0.064 (0.4)
−0.085 (0.3)
−0.0521 (0.5)
−0.0177 (0.8)
−0.0717 (0.4)
0.0672 (0.4)
0.15 (0.07)
0.0589 (0.5)
0.0506 (0.5)
−0.635 (1e−17)
−0.106 (0.2)
−0.0177 (0.8)
−0.00843 (0.9)
0.0835 (0.3)
0.156 (0.06)
−0.0161 (0.8)
−0.421 (2e−07)
−0.0204 (0.8)
0.0588 (0.5)
0.00199 (1)
−0.233 (0.005)
0.109 (0.2)
0.117 (0.2)
0.162 (0.05)
0.0875 (0.3)
0.053 (0.5)
0.0387 (0.6)
−0.169 (0.04)
−0.0875 (0.3)
0.0821 (0.3)
0.617 (2e−16)
0.134 (0.1)
−0.0427 (0.6)
−0.00822 (0.9)
−0.139 (0.1)
−0.0535 (0.5)
0.028 (0.7)
0.759 (3e−28)
0.144 (0.09)
−0.0234 (0.8)
0.137 (0.1)
−0.0642 (0.4)
−0.0891 (0.3)
−0.0336 (0.7)
0.104 (0.2)
−0.101 (0.2)
−0.0915 (0.3)
−0.122 (0.1)
0.0606 (0.5)
0.0905 (0.3)
−0.0797 (0.3)
−0.0839 (0.3)
0.0464 (0.6)
0.0286 (0.7)
−0.00329 (1)
−0.146 (0.08)
0.0738 (0.4)
0.0956 (0.3)
0.105 (0.2)
0.0537 (0.5)
Figure 3.6: Association WGCNA modules with phenotypes. Each square represents the association of a phenotype with a module with red representing strong positive association and blue strong negative association. The correlation coefficient and p-value are shown in each square.
for biological processes that regulate immune responses. These changes were also associated with lung function at 6 weeks and 2 years of age. Previously some studies shown that HEU infants had altered lung function at 6 weeks and 2 years of age in the DCHS [39, 48] and these findings suggest a potential mechanism underlying this association.
One gene, SRXN1, encoding Sulfiredoxin 1, was significantly differentially expressed between HEU and unexposed infants after correction for multiple testing. Sulfiredoxin 1 regulates perox- iredoxins which are involved into many different physiological and pathological functions including inflammatory diseases [53]. In addition, peroxiredoxins was reported to play a vital role in lung
Figure 3.7: Significant modules for maternal HIV. (a) This figure shows the significant modules for maternal HIV using the p-values of the differentially expressed genes and the modules identified through WGCNA. This bar plot presents the comparison of the modules to each other with respect to the genes that are differentially expressed for HIV. The height of the bars represents the mean value of p-values across the modules. The black dashed horizontal line represents the first quartile.
The black, cyan, green, purple, and yellow modules are associated with maternal HIV exposure.
(b) This bar plot represents the over-representation of differentially expressed genes between HEU and unexposed infants in the WGCNA modules showing –log10 of FDR for each module. The red dotted horizontal line indicates those modules that cross FDR <0.05. The yellow (M14) and black (M1) modules were over-represented for genes differentially expressed in HEU.
physiology [54] and identified to be a vitamin D regulated pathway in lung development [55]. In- terestingly, HIV Tat protein, known to alter cellular redox environment, regulates SRXN1 gene expression [56]. The DEGs for maternal HIV exposure were significantly enriched for GO biological processes and KEGG pathways involving immune function or inflammatory diseases.
WGCNA identified two modules, M1 and M14, that were strongly correlated with maternal HIV exposure. These modules were enriched for genes shown to be differentially expressed in HEU compared to unexposed. Three genes (ADIPOR1, FOXO4, and HEMGN) from M14, and four genes (LOC100127993, LOC643433, LOC646766, and LOC647856) from M1, were identi- fied as the most informative central (‘hub’) network genes significantly associated with maternal HIV exposure. Genetic variation inADIPOR1, an inflammatory and immune response gene [57],
Figure 3.8: Bar plot of the REACTOME pathways enriched in the maternal HIV associated highly co-expressed genes from two significant modules sorted by the ranking score. Length of the bar represents the degree of significant enrichment of the pathway. (a) Bar plot of the REACTOME pathways enriched in the maternal HIV associated highly co-expressed genes from the black (M1) module sorted by the ranking score. (b) Bar plot of the REACTOME pathways which were enriched by the maternal HIV associated highly co-expressed genes from yellow (M14) module.
has been previously associated with lung function [58]. Enrichment analysis of genes in these two modules identified enrichment in immune system oriented REACTOME pathways including Immune System, Adaptive Immune System, Innate Immune System, and Cytokine Signaling in Immune System. These results, coupled with the DGE analysis, clearly demonstrate that immune development is clearly altered in response to maternal-HIV exposure.
The expression of two genes (LOC728428 andLOC729500) from M14, and three genes (LOC730455, CTNNAL1,and ARL4A) from M1, were shown to be significantly associated with increased tidal volume at 6 weeks of age after adjustment for confounding variables. Expression of 42 genes from M14 were significantly associated with increased tidal volume at 2 years of age (Appendix B - Table B.7), although it should be noted in our previous studies lung function at 2 years was altered in the sub-group with severe maternal disease [48].
There are some limitations to this study. Firstly, the sample size (N=144) limits detection of small differences in gene expression between HEU and unexposed infants. In the DGE analysis, only the expression of one gene, SRXN1, was significantly different in HEU infants. However, the advantage of fGSEA is that it considers the entire ranked list of genes that reflects real biology rather than focusing on arbitrary levels of significance. Furthermore, the use of WGNCA facilitates the interpretation of large data (e.g. by focusing on modules with likely biological origins). Multiple socioeconomic, environmental, and biological factors influence child development. The effects of
HIV exposure on cord gene expression could be confounded by other factors that are affected by maternal health. In addition deciphering the role (if any) of antiretroviral neurotoxicity from HIV exposure is challenging [59].
Another limitation is that gene expression was measured in whole cord blood. However, the result of the CIBERSORT cell de-convolution showed that there were no significant differences in immune cell subtypes between HEU and unexposed offspring, suggesting that differences in cell type proportions is unlikely to account for the differences in gene expression observed. Finally, assessment of gene expression in cord blood does not necessarily provide insight into alterations in gene expression other tissues. Thus, association of gene expression in cord blood with lung function may not represent a causal relationship, rather cord blood gene expression may simply represent a biomarker of adverse in utero development. However, cord blood represents neonatal blood and has been widely studied to identify potential dysregulation of the in utero environment [61].