Persistent Microglial HIV Infection Drives Neuroinflammation Despite Viral Suppression: Insights from Rapid Postmortem Brain Biopsies

Reviewer #1 (Public review):

Summary:

HIV can persist in brain microglia despite ART and is linked to ongoing neuroinflammation and altered cellular function.

Strengths:

The authors demonstrate an innovative cell-type-specific analysis of human postmortem brain tissue from aviremic and viremic people with HIV. It uses FANS, bulk and single-nucleus RNA-seq to show that HIV DNA is concentrated mainly in microglia and that inflammatory and synaptic abnormalities persist despite ART.

Weaknesses:

The evidence is exploratory, based on a small, heterogeneous postmortem cohort. Therefore, the findings are suggestive rather than definitive.

The study would be stronger if a larger cohort, particularly more HIV-negative controls, were included. Also, less heterogeneity would reduce confounding from co-infections and terminal illness. The findings would also be better supported by longitudinal or matched peripheral data, protein-level validation, and direct evidence of viral activity rather than proviral DNA alone. A larger sample size would improve statistical power and make the cell-type differences more reliable.

Reviewer #2 (Public review):

Summary:

The authors use FANS of rapidly obtained postmortem brain tissue from DPWH, seven aviremic, four viremic and three HIV-negative controls to characterize the CNS HIV reservoir and cell-type-specific transcriptional changes.

Strengths:

The study addresses a genuinely important and understudied question: the effect of viral suppression specifically on the CNS reservoir and transcriptome using a rare and well-characterized specimen set.

Weaknesses:

I have some reservations about the conclusions, because of the confounders, mechanistic narrative, and the data itself.

(1) With n = 3 negative, n = 4 viremic, and n = 6 aviremic (post-H5 exclusion), every DEG and enrichment result rests on very few individuals. Rather than HCA reporting effect-size distributions and per-gene sample support, the authors should consider sensitivity/leave-one-out analyses to show that results are not driven by single donors. To me, it is as in Figure 3: major changes in the DGE are between the viremic vs aviremic, interestingly not with the negative control.

(2) HIV-negative controls were significantly older (74,76 & 83, inflammaging) and entirely male (sex-based immune differences). Both bias the immune comparisons that anchor the paper. PCA reassurance with n = 3 is weak. The authors should address this quantitatively, e.g., age/sex as model covariates, or explicit discussion of directionality of bias for each key pathway.

(3) HIV DNA was detectable in only 5/11 DPWH, and the microglial reservoir signal comes from ~3 individuals. The 10³-10⁴ copies/million figure and "dominant reservoir" claim should be framed against this limited detection and the focal distribution of infection.

(4) Only two participants had documented cognitive symptoms, and histopathology showed no neuropathology in anyone. The transcriptome-to-HAND link is currently asserted rather than demonstrated. The authors should state this limitation prominently and avoid implying an established relationship.

(5) A large fraction of DPWH had TB (one TBM), and controls had SARS-CoV-2. TBM alone causes microglial activation. Excluding H5 does not remove the broader TB signal. The authors should analyze/discuss TB status as a potential driver of the microglial immune signature in the retained cohort.

(6) Sorting on IRF5 cannot distinguish microglia from perivascular macrophages, as correctly stated by the authors in the discussion, so the "microglial" reservoir may include other myeloid populations. The authors should change the cell-type attribution accordingly.