Gene expression patterns associated with neurological disease in HIV infection

To provide new insight into the pathogenesis of neurocognitive impairments (NCI) in HIV infection, we used the Gene Set Enrichment Analysis (GSEA) algorithm to analyze pathway dysregulations in gene expression profiles of HIV-infected patients with or without NCI and HIV encephalitis (HIVE). While HIVE was characterized by widespread inflammation and tissue damage, gene expression evidence of induction of interferon (IFN), cytokines and tissue injury was apparent in all brain regions studied before the emergence of NCI. Various degrees of white matter changes were present in all HIV-infected subjects and were the primary manifestation in patients with NCI in the absence of HIVE. The latter showed a distinct pattern of immune activation with induction of chemokines, cytokines, β-defensins, and limited IFN induction. Altogether results indicate that significant neuroinflammation and neuronal suffering precede NCI. Patients with NCI without HIVE showed a predominantly white matter dysfunction with a distinct pattern of immune activation.

Altogether results indicate that significant neuroinflammation and neuronal suffering 39 precede NCI. Patients with NCI without HIVE showed a predominantly white matter 40 dysfunction with a distinct pattern of immune activation. 41 Introduction 54 55 While the prevalence of severe HIV-associated dementia (HAD) has decreased 56 since the introduction of combination antiretroviral therapy (cART), milder and chronic 57 forms of neurocognitive impairment (NCI) including asymptomatic neurocognitive 58 impairment (ANI) and HIV-associated neurocognitive disorders (HAND) as well as HIV-59 associated major depressive disorder remain high (1-7). HIV encephalitis (HIVE) is 60 considered to be the main neuropathological substrate of HAD (8-10). NCI in the setting 61 of cART is associated with synaptodendritic degeneration (7,11,12). While the brain 62 represents a sanctuary where HIV can persist due to suboptimal penetration of 63 antiretroviral drugs (13), various studies highlighted the occurrence of NCI even in the 64 setting of viral suppression (14,15). Chronic neuroinflammation is believed to drive 65 neurodegeneration in cART-era HAND (7,9,16,17). However, the pathogenic 66 mechanisms behind HAND remain unclear. 67

68
To identify gene expression correlates of neurological disease progression in 69 HIV, we analyzed pathway dysregulations in brain regions of patients in the National 70 NeuroAIDS Tissue Consortium (NNTC) gene expression profile dataset. The NNTC 71 dataset consists of samples from 3 different brain regions (white matter, basal ganglia, 72 prefrontal cortex) of control and HIV-infected patients with or without NCI and HIVE 73 (18). For pathway analysis we used the Gene Set Enrichment Analysis (GSEA), a 74 computational method to assess whether a priori defined sets of genes show 75 statistically significant differences between biological states (19). GSEA was used in 76

Identification of pathways differentially regulated in HIV-infected patients with NCI 158 without HIVE vs. uninfected controls (C-A comparison). 159
HIV-infected patients with NCI and no HIVE (group C), showed significant changes oxidative stress and cytochrome P450 enzymes (KEGG DRUG METABOLISM 166 CYTOCHROME P450, REACTOME BIOLOGICAL OXIDATIONS), matrix 167 metalloproteases (MMPs) (NABA MATRISOME ASSOCIATED), and downregulation of 168 genes related to RNA transcription and processing (e.g., REACTOME RNA POL II 169 TRANSCRIPTION, KEGG SPLICEOSOME), (Supplementary Table 3 Two pathways were concordantly differentially regulated between groups B and C in all 174 brain regions. These pathways are indicative of type I IFN activation in HIV-infected 175 patients without NCI (group B), as indicated above. We also identified 47 pathways 176 specifically differentially regulated in the prefrontal cortex. Among the pathways 177 upregulated in group B as compared to C in the prefrontal cortex, were pathways 178 indicative of tissue damage (e.g., REACTOME REGULATION OF APOPTOSIS), RNA 179 transcription and processing (e.g., REACTOME METABOLISM OF RNA, KEGG 180 RIBOSOME), and pathways related to protein degradation (e.g., KEGG 181 PROTEASOME, REACTOME AUTODEGRADATION OF THE E3 UBIQUITIN LIGASE 182 COP1, REACTOME APC C CDC20 MEDIATED DEGRADATION OF MITOTIC 183  (Fig. 6). These pathways include genes indicative of trophic 215 interaction, protein misfolding and mitochondrial function. We also identified 216 downregulated pathways related to mitochondria and energy metabolism were 217 decreased in group D in all brain regions at FDR < 0.2 (e.g., REACTOME TCA CYCLE 218 AND RESPIRATORY ELECTRON TRANSPORT, REACTOME PYRUVATE 219 METABOLISM AND CITRIC ACID TCA CYCLE, REACTOME GLYCOLYSIS), 220

PROTEINS), (Supplementary
(Supplementary Table 5  the latter may also reflect compensatory changes as passively administered IFN-β 286 impaired spatial memory in mice in another study (36). A recent study suggested a role 287 for IFN-γ in shaping fronto-cortical connections and social behavior (37), which is 288 consistent with a potential role of excessive IFN activation in the pathogenesis of NCI. 289 In a recent large multi-center trial, depression was not significantly increased by IFN-290 βtreatment for multiple sclerosis (MS) (38). The early induction of IFN in patients of the 291 NNTC dataset (patients with HIV without NCI, group B) is reminiscent of previous 292 studies in which IFN induction was not closely correlated with NCI, e.g., (39), and raises 293 the possibilities that either protracted IFN dysregulation may be required to produce NCI 294 or that it may be a co-factor in NCI pathogenesis. 295 296 Also evident in HIV-infected patients without NCI was the activation of 297 mechanisms indicative of tissue injury, such as expression of matrix metalloproteases 298 (MMP) and complement-related genes in the white matter. MMP expression by HIV-1 299 infected monocytes and macrophages is recognized as a pathogenic mechanism in 300 neuroAIDS (40). Elevated MMP levels can contribute to microglial activation, infiltrate 301 through cleavage of adhesion molecules, neuronal and synaptic injury, as well as blood-302 brain barrier disruption (41-44). MMP increases were present in the white matter in HIV-303 infected patients with NCI and no HIVE. In patients with HIVE, induction of MMPs was 304 also evident in the prefrontal cortex and basal ganglia. 305 Another key finding in the study is that patients with NCI without HIVE (group C) 306 in the NNTC cohort did not show significant activation of IFN, unlike patients in groups B 307 and D. This discordant regulation of IFN signaling did not appear to be associated with 308 antiretroviral therapy as patients with NCI and no HIVE include both patients treated 309 with antiretrovirals and untreated patients. Conversely, patients with NCI without HIVE 310 Evidence also suggests a role for defensins in the chronic inflammation associated with 326 degenerative brain diseases, and in particular Alzheimer's disease (52, 53). Defensins-327 related pathways were also induced in HIVE, but showed no consistent regulation in 328 HIV-infected patients with no NCI, suggesting a possible contribution to the 329 pathogenesis of NCI. 330 331 In HIV without NCI, genes related to neurotransmission were also downregulated 332 in the prefrontal cortex and basal ganglia while genes related to apoptosis, such as 333 calpain-related mechanisms, which contribute to neurodegeneration in HIV (54) The anatomical distribution of the gene expression programs dysregulated in the 342 NNTC dataset appears to reflect brain-region specific dynamics in neurological disease 343 progression in HIV/AIDS. In particular, we observed some degree of white matter 344 alteration of gene expression in all HIV-infected groups with and without NCI and HIVE. 345 However, gene expression changes in patients with NCI without HIVE (group C) were 346 localized to the white matter and had a specific gene expression profile. Lack of gene 347 expression changes suggestive of neuronal injury in the prefrontal cortex and basal 348 ganglia in patients with NCI without HIVE (group C) suggests that they may not be 349 accompanied by significant neuronal atrophy, but that white matter pathology likely 350 drives NCI in these patients. Prominent white matter gene expression changes were 351 also present in HIVE, which was also characterized by considerable gene expression 352 changes in the prefrontal cortex and basal ganglia. White matter damage correlating 353 with the severity of cognitive manifestations has been observed since the early days of 354 the HIV pandemic (8, 56). Evidence of white matter injury in HIV-infected patients with 355 and without NCI is also demonstrated in recent imaging studies (57,58). In addition to 356 white matter changes, gene expression in HIVE was characterized by considerable 357 changes in the prefrontal cortex and basal ganglia. This is also in apparent agreement 358 with the association of NCI with progression of functional abnormalities involving the 359 basal ganglia and the prefrontal cortex as well as with generalized white matter damage 360 (56,(59)(60)(61)(62). 361

362
The present study has several limitations. Primarily, the NNTC dataset groups 363 are of small sample size that was further reduced as part of the quality control analysis. 364 Larger studies will be needed to better understand the pathogenesis and progression of 365 neurological disease and to adequately represent all possible variants of central Interestingly, the group of HIV-infected patients with NCI without HIVE showed a 376 preeminently white matter dysfunction characterized by a distinct pattern of immune 377 activation with low IFN. Larger studies are necessary to better understand the 378 pathogenesis of neurological disease and its progression, to evaluate the impact of 379 therapy on various HIV disease conditions, and to identify better therapeutic targets and 380 strategies for NCI in HIV. 381

Competing interests 382
The authors declare that they have no competing interests" 383 384 Authors