Gene expression patterns differed among patients who developed C-IRIS, survived or died
Our goal was to identify gene expression signatures in peripheral blood associated with the development of death or fatal C-IRIS among patients with CM. We performed next generation RNA sequencing of samples from 68 HIV-infected subjects with CM who participated in the COAT Trial. Of these, 39 subjects had been randomized to initiate ART within two weeks (earlier arm) and 29 subjects had been randomized to initiate ART after 5 weeks (deferred arm) [7]. All subjects received induction anti-fungal therapy with Amphotericin B and fluconazole, followed by consolidation and maintenance therapy with fluconazole as described previously [7]. Of the 68 subjects, 10 subjects in each arm did not develop C-IRIS or death (Control Group)), 10 subjects in each arm had C-IRIS and survived (C-IRIS Survivor Group), and 9 subjects in each arm died without C-IRIS (Death without C-IRIS Group). Ten subjects from the earlier arm had C-IRIS and died (Fatal C-IRIS Group; see Figure 1). Cases of C-IRIS included those classified based on predefined criteria in [6]. All cases of death or C-IRIS occurred within the first month of the ART treatment. Peripheral blood was collected before ART initiation (week 0), at weeks 1, 4 and 8 on ART, and at the time of C-IRIS (Figure 1). Groups had no significant differences in age, pre-ART HIV viral load, or CD4+ T cell count (Supplemental figure 1). The serum cryptococcal antigen titers, as measured by LFA (lateral flow immunoassay), were significantly higher in the C-IRIS Survivor Group, as compared to patients who died in the earlier study arm with or without C-IRIS (Supplemental figure 1).
We hypothesized that patients with CM would show distinct gene expression signatures in the blood that could distinguish between the four groups: 1) No C-IRIS or Death (control), 2) C-IRIS Survivor Group), 3) Fatal C-IRIS Group, and 4) Death without C-IRIS Group. We applied a principal component analysis (PCA), to visualize the differences in gene expression levels in the entire dataset (26398 expressed RNAs for each sample) (Figure 2A). Principal component analysis was used to separate gene expression in 211 individual samples from the 68 subjects. The results separated the samples into four clusters: samples from patients who survived (excluding samples collected at the time of C-IRIS events, yellow dots), samples from those who died (excluding samples collected at the time of C-IRIS events, blue dots), and samples collected at C-IRIS events (red, died; green, survived). As shown in figure 2A, samples collected from the C-IRIS Survivor Group at the time of C-IRIS events clustered separately from samples collected from the Fatal C-IRIS Group at the time of fatal C-IRIS events.
We performed additional PCA analysis in order to cluster samples based on expression of significant immune transcripts among samples collected at week 0 (pre-ART), from all patient groups (Figure 2B). The list of all expressed transcripts was truncated to 2200 immune genes, and the top 3 principal components were extracted. The immune gene expression at week 0 showed distinct separation comparing patients who later died of C-IRIS from those who survived (shown as PCA#1, black, grey and green dots, in figure 2B). These data show that gene expression in those who died could be readily distinguished from those who survived already at the time of ART initiation (week 0), suggesting that gene expression signatures could potentially be used as prognostic markers to define risk for death. These results prompted us to perform further data mining to reveal biological information in gene sets that may contribute to fatal events.
At week 0 transcriptome profiles differed between surviving patients who did or did not develop C-IRIS
Transcript expression was further compared between groups by ANOVA-REML, using criteria described in the methods. Comparing gene expression at week 0 in surviving patients who did not develop C-IRIS (Control Group) to those who developed C-IRIS (C-IRIS Survivor Group), C-IRIS survivors exhibited lower baseline expression of transcripts associated with antiviral responses, including interferon-induced protein family members, interferon-induced proteins with tetratricopeptide repeats, and oligoadenylate synthetases (IFIs, IFITs, ISGs, and OASs respectively) (Supplemental figure 2). This confirmed our previously published observation that lower expression of genes encoding interferon type 1- inducible transcripts at the time of CM diagnosis are associated with subsequent nonfatal C-IRIS events [9]. Additionally, we identified elevated expression of transcripts encoding markers of complement components (C1QA, C1QB, C1QC, CFD), HLA-DRB (1 and 5), and IL12 pathways in the C-IRIS Survivor Group at week 0, in comparison to the Control Group (p<0.05, no FDR correction, see Supplemental table 1). The complement components, HLA-DRs and IL12 play important roles in antigen presentation, which suggests that prior to ART commencement, monocytes and antigen-presenting cells are already activated in C-IRIS survivors. These results suggest that we can use gene expression biomarkers to identify patients at risk for the development of C-IRIS prior to their initiation of ART [9].
At week 0 transcriptome profiles differed between patients who survived or died
We compared gene expression at week 0 in ten patients from the earlier ART arm who subsequently died from C-IRIS (Fatal C-IRIS Group) to week 0 control patients (No C-IRIS or Death Group) in both study arms. Interestingly, the patients who died from C-IRIS did not exhibit significant deficiency in the expression of interferon-response pathway genes, as compared to controls. These results were different from the result described above where patients with C-IRIS who survived had lower expression of these genes. Using Ingenuity Pathway Analysis (IPA) software we combined transcript expression measurements over pathways to identify which pathways act as markers in predicting or explaining fatal outcomes. We ran a comparison analysis of canonical these pathways and built the heatmaps of the top up- and down-regulated pathways.
We identified twelve inflammatory immune pathways that were upregulated at week 0 in the Fatal C-IRIS Group as compared to the three other groups from either arm of study: No C-IRIS or Death Group (Control), C-IRIS Survivor Group, and Death without C-IRIS Group. Since the results were almost identical, we combined groups from both study arms together (see Figure 3A, columns 1, 2, 3, respectively). These 12 pathways included cytokines and other pro-inflammatory molecules such as CXCL1, CXCR1, ICAM1, IL6, IL8, IL11, interferons, and components of MAPK signaling (MAPK14, NFKBIA, IL1 receptors). No significantly downregulated pathways were identified, suggesting that patients who died due to C-IRIS, were not immunocompromised to a higher degree than patients from other groups based on lack of immune gene expression. Notably, patients from Death without C-IRIS Group did not show any statistically significant upregulation of immune gene expression, except for type 1 interferon signaling, when compared to the No C-IRIS or Death (Control), or the C-IRIS Survivor Group (week 0). The Death without C-IRIS Group, however, exhibited a trend towards a downregulated expression of various HLA-, Th1- and Th2- pathway encoded genes, but upregulation of PD1/PDL1 at the time of ART initiation, which may reflect the adaptive immune cell exhaustion (z scores <1). The fold change in expression values for these transcripts are presented in the Supplemental table 2. Together, these results suggest that gene expression at week 0 differed between the Fatal C-IRIS Group, the Death without C-IRIS group and those of survivors (Figure 3A).
Changes in gene expression caused by ART initiation, in earlier and deferred ART initiation arms has differed among those who died and survived
Using same pathway-based biomarker discovery approach, we assessed changes in pathway expression over time among patients in the earlier and deferred ART initiation arms. At one week of observation on ART, studied groups exhibit upregulation of similar components of pro-inflammatory pathways, when compared to week 0 within each corresponding arm (Figure 3B). Patients in the Death without C-IRIS Group, whether they were in the earlier or deferred ART arms, showed a distinct shift toward upregulation of transcripts involved in oxidative phosphorylation, high mobility group box 1 proteins (HMGB1), pro-inflammatory cytokine signaling (IL6, 8, 15), and Rho/GTPase pathways (Figure 3B, columns 1, 2). A similar trend was seen in Fatal C-IRIS Group at one week after ART initiation (Figure 3B, column 3). Interestingly, Ras homologous Guanosine diphosphate (GDP)-dissociation inhibitors, RhoGDIs, which negatively regulate Rho family GTPases, were downregulated in all Death Groups (with or without C-IRIS), at 1 week after ART initiation. Thus, No C-IRIS or Death Groups (with or without C-IRIS), showed distinct gene expression changes, yet, independently from the timing of ART initiation.
Control groups (in earlier and deferred arms) did not exhibit significant changes at week 1, in the above mentioned pathways, or other proinflammatory pathways, as compared to corresponding week 0 (Figure 3B, column 4, 5). Patients in the C-IRIS Survivor Group from the earlier ART arm exhibited more dramatic upregulation of proinflammatory gene expression as compared to those in the C-IRIS Survivor Group who were in the deferred ART arm (Figure 3B, column 6, 7).
Comparison between samples, collected at week one post-ART showed that the Death without C-IRIS Group exhibited the most pronounced changes in gene expression in earlier and deferred ART arms (Figure 3C). Numerous transcripts within granulocyte- activation pathways, such as N-formyl-Met-Leu-Phe (fMLP pathway), HMGB1, and Rho family GTPases (Rho GTPases), were upregulated in the Death without C-IRIS Group, as compared to the controls. The oxidative phosphorylation pathway was the foremost overexpressed, followed by Rho family GTPases and stress kinases (Figure 3C, column 1). The Fatal C-IRIS Group exhibited similar gene expression changes, when compared to controls (Figure 3C, column 2); however, when compared to Death without C-IRIS Groups (earlier and deferred arms combined), the oxidative stress and neutrophil involvement signatures were not present at week 1 of observation (Figure 3C, column 3). In comparison, C-IRIS survivors have shown significantly lower expression of 10 out of 12 pathways presented in figure 3 (see column 4) when compared to Death without C-IRIS Groups. These results indicate that the transcriptomic signature of Death without C-IRIS Group in the blood samples collected at the closest time point to death is different than in those of comparators.
Immune recovery over 8 weeks on ART in control patients without C-IRIS or death
Longitudinal analysis of immune reconstitution during 8 weeks on ART in the No C-IRIS or Death (control) Groups, who had favorable clinical recovery, revealed downregulation of interferon signaling and NF-kappa B signaling pathways, but upregulation of phagocyte maturation pathways (Figure 3D, columns 1, 2). This course of immune recovery on ART was similar to what we previously demonstrated in advanced stage HIV-infected patients without opportunistic infections after initiation of ART [10]. We also assessed changes in gene expression in the C-IRIS Survivor Group in the period following C-IRIS events. The C-IRIS Survivors Group exhibited downregulation of eight signaling pathways at week 8 on ART (most proximal post-C-IRIS time point) when compared to the gene expression at C-IRIS events (Figure 3D, columns 3, 4). For example, Toll-like receptor signaling, pro-inflammatory IL6, IL8, and T cell exhaustion pathways were downregulated, which may represent a sign of recovery from chronic antigen exposure and a prolonged stage of chronic inflammation [11]. Additionally, transcripts involved in Th1 and Th2 pathways showed a trend towards upregulation, but this trend did not pass the threshold of a significant z-score (Figure 3D, columns 3, 4). Perhaps, assessment of longer time points of evaluation, such as 12 and 26 weeks post-ART, would show a more significant recovery in expression of T cell pathways.
Dysregulated immune gene expression at the time of fatal C-IRIS
The major goal of this study was to identify altered gene expression at fatal C-IRIS events in order to better understand the molecular pathology of fatal C-IRIS. Analysis of changed gene expression at the time of fatal C-IRIS revealed many significantly upregulated acute phase response and oxidative stress pathways (e.g. IL1 and IL6, TLR, HMGB1, NRF2-mediated, NFkB, p38-MAPK), when compared to the most proximal longitudinal time point (W1) within the fatal C-IRIS group (Figure 3E column 1), or the No C-IRIS or Death (control) Group at week 1 (Figure 3E column 2).
Since fatal C-IRIS cases occurred within the 4 weeks of ART initiation and only in the earlier ART arm, we compared gene expression in the Fatal C-IRIS Group to the Control Group collected at week 4 post-ART in the earlier ART arm (Figure 3E column 3). Fatal C-IRIS showed significant upregulation of 14 inflammatory pathways, implying that during favorable immune reconstitution, effective antifungal treatment and ART, the acute inflammation accompanied by reactive oxygen species production are downregulated within the first month of treatment.
Transcripts involved in acute phase response pathways, Toll-like receptor signaling, IL1,6, p38 MAPK were upregulated in the fatal C-IRIS group as compared to C-IRIS Survivor Groups, during C-IRIS events (Figure 3E, column 4). The lower expression of transcripts involved in Th1, CD28 and other T cell- related pathways, but upregulation of transcripts involved in T cell exhaustion signaling pathways (PD1/PDL1), was observed during fatal C-IRIS events when compared to C-IRIS events in the C-IRIS Survivor Group (Figure 3E, column 4).
In comparison to Death without C-IRIS group, fatal C-IRIS showed the upregulation of the same acute phase response pathways, a similar expression of HMGB1 and T cell exhaustion signaling, but downregulation of complement and oxidative stress pathways (eNOS, NRF2-mediated, Rho family GTPases, fMLP, etc.) (Figure 3E, column 5). These results indicate a divergence between activation pathways that are associated with death without C-IRIS and death due to C-IRIS.
Transcriptomic biomarkers predict fatal C-IRIS or death
To identify transcripts that may be predictive biomarkers of fatal C-IRIS events, we used a probability modeling based on a partial least squares (PLS) computational algorithm, as described in the methods. PLS analysis was performed for samples from the C-IRIS Survivor Group and Fatal C-IRIS Group, with comparisons to the rest of the groups, including all time points within each group. The models were carried out on the list of 2200 molecules that are significantly enriched in immune pathways.
The model revealed that most of the top-ranked biomarkers were distinct and specific for either fatal or non-fatal C-IRIS events, and the rest were similar between these groups. For example, differential expression of transcripts encoding IL15, IL31, integrins (ITGA7, ITGB2), and SIGLECs (sialic acid binding immunoglobulin - like lectins) were ranked as highest importance for nonfatal C-IRIS events (C-IRIS Survivor Groups), but not for fatal C-IRIS event (see Table 1). Conversely, p38 MAPK signaling, IL1R, IL18R, TLR1,2,4, NLRP8,12, and transcripts encoding CLECs (C-type lectins), ranked as higher importance for fatal C-IRIS (Fatal C-IRIS Group), but not for C-IRIS survivors (Figure 4).
Table 1
The comparison of transcriptomic biomarkers between fatal and nonfatal C-IRIS.
Gene
|
Fatal C-IRIS Group
|
C-IRIS Survivor Group, earlier arm
|
C-IRIS Survivor Group, deferred arm
|
Symbol
|
R for VIP
|
VIP rank
|
R for VIP
|
VIP rank
|
R for VIP
|
VIP rank
|
LIMK1
|
-0.0116
|
-----
|
0.0054
|
++++
|
0.015
|
++++
|
ALOX5
|
-0.0101
|
----
|
0.0026
|
++
|
0.0102
|
+++
|
CTSC
|
-0.0087
|
----
|
0.0056
|
++++
|
0.0137
|
++++
|
IL17REL
|
-0.0092
|
----
|
0.0033
|
++
|
0.0074
|
++
|
TLR10
|
-0.0089
|
----
|
0.0027
|
++
|
0.012
|
+++
|
C3AR1
|
-0.0086
|
---
|
0.0058
|
++++
|
0.012
|
+++
|
CTSD
|
-0.0064
|
---
|
0.0038
|
+++
|
0.0054
|
++
|
CTSK
|
-0.0086
|
---
|
0.0045
|
+++
|
0.0103
|
+++
|
DEFB115
|
-0.007
|
---
|
0.0057
|
++++
|
0.0122
|
+++
|
IL17D
|
-0.0068
|
---
|
0.0036
|
++
|
0.0111
|
+++
|
ITGA7
|
-0.0077
|
---
|
0.0058
|
++++
|
0.0086
|
++
|
ITGB2
|
-0.0067
|
---
|
0.0052
|
+++
|
0.0086
|
++
|
JAG1
|
-0.0085
|
---
|
0.0034
|
++
|
0.011
|
+++
|
SIGLEC10
|
-0.0074
|
---
|
0.0065
|
++++
|
0.0118
|
+++
|
ALPL
|
-0.0044
|
--
|
0.0063
|
++++
|
0.0103
|
+++
|
CASP5
|
-0.0042
|
--
|
0.0045
|
+++
|
0.0121
|
+++
|
DEFA5
|
-0.0044
|
--
|
0.0047
|
+++
|
0.0047
|
+
|
ITGAX
|
-0.0061
|
--
|
0.0069
|
+++++
|
0.0106
|
+++
|
NOD2
|
-0.0043
|
--
|
0.0043
|
+++
|
0.0054
|
++
|
NOTCH2
|
-0.0045
|
--
|
0.0048
|
+++
|
0.0063
|
++
|
NOX1
|
-0.004
|
--
|
0.0048
|
+++
|
0.0044
|
+
|
SIGLEC5
|
-0.0044
|
--
|
0.0037
|
+++
|
0.0084
|
++
|
SIGLEC7
|
-0.0057
|
--
|
0.0047
|
+++
|
0.0091
|
+++
|
BEX1
|
-0.0033
|
-
|
0.0027
|
++
|
0.0026
|
+
|
CASP2
|
-0.0034
|
-
|
0.0038
|
+++
|
0.0063
|
++
|
ITGA5
|
-0.0034
|
-
|
0.0056
|
++++
|
0.0043
|
+
|
SIRPA
|
-0.0018
|
-
|
0.0038
|
+++
|
0.007
|
++
|
IL1RAP
|
0.0146
|
++++++
|
-0.002
|
-
|
-0.0071
|
--
|
MMP20
|
0.0117
|
+++++
|
-0.0037
|
--
|
-0.0086
|
--
|
ALOX5AP
|
0.0087
|
++++
|
-0.0035
|
--
|
-0.0073
|
--
|
CCL1
|
0.0103
|
++++
|
-0.004
|
---
|
-0.0079
|
--
|
CD55
|
0.0089
|
++++
|
-0.0005
|
|
-0.0041
|
-
|
CLEC4D
|
0.0102
|
++++
|
-0.0027
|
--
|
-0.0055
|
--
|
CLEC4E
|
0.0106
|
++++
|
-0.0049
|
---
|
-0.0068
|
--
|
CLEC5A
|
0.0091
|
++++
|
-0.0032
|
--
|
-0.004
|
-
|
F2RL2
|
0.0088
|
++++
|
-0.0011
|
-
|
-0.0058
|
--
|
FADD
|
0.0102
|
++++
|
-0.0037
|
---
|
-0.0071
|
--
|
FAM65B
|
0.0091
|
++++
|
-0.0006
|
-
|
-0.0029
|
-
|
IFNAR1
|
0.0094
|
++++
|
-0.0018
|
-
|
-0.0037
|
-
|
IL18R1
|
0.0106
|
++++
|
-0.0018
|
-
|
-0.0032
|
-
|
IL18RAP
|
0.0101
|
++++
|
-0.0016
|
-
|
-0.0049
|
-
|
IL1R1
|
0.01
|
++++
|
-0.0019
|
-
|
-0.0052
|
-
|
IRAK3
|
0.0088
|
++++
|
-0.0007
|
-
|
-0.0042
|
-
|
KLKP1
|
0.0102
|
++++
|
-0.0038
|
---
|
-0.0064
|
--
|
LILRB5
|
0.0106
|
++++
|
-0.0013
|
-
|
-0.006
|
--
|
TLR4
|
0.0096
|
++++
|
-0.003
|
--
|
-0.0046
|
-
|
CASP4
|
0.0065
|
+++
|
-0.003
|
--
|
-0.0021
|
-
|
CD59
|
0.0082
|
+++
|
-0.0003
|
|
-0.0003
|
|
CLEC2L
|
0.0063
|
+++
|
-0.0044
|
---
|
-0.0102
|
---
|
CXCR1
|
0.0065
|
+++
|
-0.0011
|
-
|
-0.0004
|
|
EXOSC4
|
0.0081
|
+++
|
-0.0018
|
-
|
-0.002
|
-
|
FAM188A
|
0.0076
|
+++
|
-0.0031
|
--
|
-0.0049
|
-
|
IFNGR1
|
0.0066
|
+++
|
-0.0002
|
|
-0.0017
|
|
IL17F
|
0.0086
|
+++
|
-0.0028
|
--
|
-0.0043
|
-
|
IL1R2
|
0.0084
|
+++
|
-0.0006
|
|
-0.0054
|
--
|
KLF2
|
0.0072
|
+++
|
0.0018
|
+
|
-0.0025
|
-
|
KLF6
|
0.0064
|
+++
|
-0.0008
|
-
|
-0.0022
|
-
|
MMP8
|
0.0068
|
+++
|
-0.0006
|
|
-0.006
|
--
|
CLEC4F
|
0.0039
|
++
|
-0.0036
|
--
|
-0.0065
|
--
|
IFRD1
|
0.005
|
++
|
0.0009
|
+
|
-0.0028
|
-
|
IL1RL1
|
0.0048
|
++
|
-0.0032
|
--
|
-0.0045
|
-
|
IL1RN
|
0.0049
|
++
|
-0.0028
|
--
|
-0.0016
|
|
MMP9
|
0.0057
|
++
|
0.0004
|
|
-0.0042
|
-
|
MOSPD2
|
0.0044
|
++
|
0.0001
|
|
-0.0019
|
-
|
TLR2
|
0.0059
|
++
|
-0.0007
|
|
-0.0006
|
|
HIV_RNA
|
0.0017
|
+
|
-0.0018
|
-
|
-0.0015
|
|
IL1RL2
|
0.0032
|
+
|
0.0023
|
++
|
-0.0011
|
|
SIGLEC9
|
0.0023
|
+
|
0.003
|
++
|
0.0038
|
+
|
IFNAR2
|
0.0008
|
|
0.0025
|
++
|
0.0019
|
+
|
NLRP3
|
-0.0009
|
-
|
0.0054
|
++++
|
0.0079
|
++
|
VIP, Variable Importance in the projection values.
R for VIP, Correlation Coefficient of Least Square Mean for VIP.
number of "-" is a visual representation for predictive importance strength that the following transcript's expression is negatively associated with occurrence of IRIS events. One "-" is equal 0.7 on VIP scale.
number of "+" is a visual representation for predictive importance strength that the following transcript's expression is positively associated with occurrence of IRIS events. One "+" is equal 0.7 on VIP scale.
|
PLS model identified the list of novel contributive transcripts with high importance for Death without C-IRIS Groups (earlier and deferred arms combined), but not for fatal or nonfatal C-IRISs. These included C1QTNF2 and 5, CD207, CD209, CXCL3 and 11, CCL28, MIP, defensins (DEFA6, DEFB107A, DEFB116, DEFB118, etc.) (Supplemental table 3).
Additionally, PLS model identified the high importance of many interferon-response genes, CLEC4F, NLRP5, CD4, CD1C, CCL21, CCR10, as week 0 predictors of subsequent nonfatal C-IRIS. This model also identified FN1 (fibronectin), LILRB4, SERPINE1, CD80, HLA-DQB2, FERM2 (fermitin 2), RND2, CCDC42 (Rho family GTPases), IL11, IL31, and many other transcripts within several stress response kinases pathways, to be week 0 predictors for subsequent fatal C-IRIS. Conversely, biomarkers such as NLRP6, C3, C5, CTSO, AGTR1, CDH9, etc. were identified as week 0 predictors of Death without C-IRIS Groups.
The complete listing of immune biomarkers with a quantitative estimation of the discriminatory power of each predictor transcript provided by means of VIP scores and the correlation coefficient (R for VIP) is presented in Supplemental Table 3. We were unable to provide pathways comparisons since pathways for many novel top-ranked biomarkers identified by PLS were not available in IPA. Overall, our results suggest that immune and inflammatory biomarkers that we discovered, could be useful for identifying patients with CM who are at risk for dying from C-IRIS.