Brain ¹⁸F-FDG PET of SIV-infected macaques after treatment interruption or initiation

Since its introduction in the early 1990s, antiretroviral therapy (ART) has significantly decreased the mortality and morbidity and improved the quality of life for people living with HIV (PLWH) [1]. The effectiveness and availability of ART have improved over time: by 2015, 46% of the 36.7 million PLWH had access to ART [2, 3]. While ART is increasingly accessible, adherence to drug regimens depends on a number of factors and can be disrupted by phenomena including poverty, stigma, lack of social support, comorbid psychiatric disorders, and adverse drug effects including GI discomfort, lipodystrophy, renal dysfunction, and sensory neuropathy [4‐10]. Due to these and other factors, ART adherence rates are estimated to be between 56 and 72%, with discipline declining over time, especially for populations with HIV-associated neurocognitive disorders (HAND) [7, 11‐14].

Research into treatment nonadherence and interruption has documented various negative consequences, including rebounds in the cerebrospinal fluid (CSF) and plasma HIV viral load (VL), increased CSF neurofilament protein and neopterin, lymphocytic pleocytosis, greater likelihood of neurocognitive disorders, and elevated rates of disease and death when compared to continuous treatment [15‐19]. Of note, despite the extensive evidence of benefit from ART, several studies have shown improvements in neuropsychological outcomes of PLWH after the interruption of ART treatment for periods as long as 96 weeks, adding nuance to our understanding of the effects of ART [16, 20, 21] and the balance between effects of HIV and side-effects of ART.

ART initiation generally suppresses HIV RNA in plasma and CSF, with the latter suppression being especially high for drugs with high central nervous system (CNS) penetration effectiveness (CPE) [22, 23]. High-CPE drugs, however, seem to have cognitive effects that are not fully understood. While some studies link them to improved neuropsychological outcomes, others have found a relationship between their use and an elevated risk of HIV dementia relative to drugs with low CPE [22‐24]. Furthermore, while the current generation of ART has been effective in reducing rates of HIV dementia compared to the pre-ART era, milder forms of HAND persist [25]. Certain antiretrovirals (ARVs), such as efavirenz, have been specifically implicated as adversely affecting neurocognitive performance, while increased usage of others has been associated with a reduction in white matter volume when measured by magnetic resonance imaging (MRI) [26, 27]. The mechanisms by which ART may be causing toxicity are not fully understood, but various ARVs have been found to inhibit telomerase activity, induce oxidative stress, and interfere with mitochondrial enzymes [28‐32].

In this study, we wanted to assess the longitudinal effects of ART initiation and interruption on brain glucose metabolism in a controlled setting. In doing so, we hoped to better understand how treatment interruption and resumption—whether medically sanctioned or non-prescribed—impact patients’ brains at the molecular level. Towards this goal, we longitudinally measured alterations in brain glucose metabolism, using positron emission tomography (PET) with ¹⁸F-Fluorodeoxyglucose (FDG), and correlated those with changes in CSF and plasma viral load, cytokines, and T lymphocyte counts as a result of treatment interruption and initiation. We anticipated that interrupting the treatment of simian immunodeficiency virus (SIV) infected monkeys would lead to increases in FDG uptake, CSF and plasma viral loads, and levels of inflammatory cytokines and to a decline in lymphocyte counts. We expected that the reverse would occur when treatment was initiated.

The current study has clearly demonstrated that discontinuation of ART is associated with increased metabolic activity in the brain as measured by FDG uptake, as early as 1 month after interruption. This is likely due to neuroinflammation secondary to SIV replication and is extremely relevant to patients taking “drug holidays” and to investigational treatment interruption studies. The interruption-induced increases in brain metabolism observed in our experiment were significantly correlated with increased CSF and plasma VL and cytokines and decreased CD4+ and CD8+ T-cell counts.

The concept that glucose metabolism reflects disease activity in HIV/SIV is not new. Although not necessarily concentrating on the brain, previous imaging studies have correlated HIV laboratory and clinical biomarkers to FDG PET data in the periphery: FDG uptake in the lymph nodes, for example, has been correlated with viremia in both treated and untreated patients, and inversely correlated with CD4+ T-cell counts [33‐35]. Another study performed with SIV-infected monkeys found that tissues with greater FDG uptake, such as the ileocecal lymph nodes, exhibited more productive SIV infection as measured by SIV RNA than did other lymph nodes with lower levels of FDG uptake [36]. Lymph node FDG uptake has also been found to correlate with immunological variables, including percentage of CD8+, CD38+, and RO+ T cells in untreated patients [37]. In the brain, in the pre-ART era, basal ganglia hypermetabolism was described as occurring in the early stages of infection, with eventual transition to cortical and subcortical hypometabolism in more chronic stages of the disease [38‐40]. This early-stage hypermetabolism was linked to worsened neuropsychological performance and was found in groups of patients with low CD4+ T-cell counts [41, 42]. Our data resemble those findings in that we show hypermetabolism in the setting of ART interruption and viral rebound, though the regional variations described above were not seen in our study, as evidenced by the strong agreement between whole brain and subcompartment uptake of FDG in this study (Fig. 3). This could be due to the short duration of follow-up we had (6 months). Our data, nonetheless, extend previous findings by correlating hypermetabolism in the brain with markers of disease course, most notably drops in CD4+ and CD8+ T-cell counts and increases in plasma and CSF viral loads.

Although we cannot assert neurologic damage in association with increased cerebral metabolism post-interruption, the phenomenon raises concerns about the potential effects of even brief periods of non-adherence to ART. FDG, while not a specific marker of inflammation, often correlates with the neuroinflammatory burden [43]. The hypermetabolism observed in this study, therefore, is likely the result of neuroinflammation in the setting of viral rebound. This possibility is supported by the positive correlations between post-interruption FDG uptake and levels of the CSF cytokines IL-2 and IL-15. Both IL-2 and IL-15 are pro-inflammatory cytokines which share two signaling subunits while having distinct functionality: IL-2 induces T cell, B cell, and NK cell proliferation and cytokine release while IL-15 causes increased activation of CD8+ T-cells [44‐49]. Levels of IL-2 have been found to be higher in patients with higher viral loads and those with higher CD4+ T-cell counts, while levels of IL-15 have been correlated with higher viral loads and lower CD4+ T-cell counts [50, 51]. The increased levels of these cytokines post-interruption therefore suggest an inflammatory response that is occurring in parallel to increased FDG uptake. The correlation of FDG uptake with CSF, but not plasma, concentrations of IL-2 and IL-15 further suggests that the CNS response may be somewhat disconnected from the peripheral response, and that peripheral measurements may not adequately represent cerebral disease course. Interestingly, there was no further correlation between time post-treatment interruption and change in SUV values beyond the first month of treatment cessation. These data indicate that the effects of interruption on glucose metabolism specifically might be most intense in the acute phase, and stabilize afterward.

The results observed in the treatment initiation group were more heterogeneous than those from interruption. Post-initiation, some animals showed decreases in brain metabolism at months one and three, and there was an average decrease in brain metabolism at month six, but these changes were not consistent or strong enough to reach statistical significance. This was despite a marked reduction, within 1 month, of plasma and CSF viral loads, an increase in CD4+ and CD8+ T-cell counts, and lowered levels of CSF cytokines including IL-1Ra and MCP-1, similar to what has been described previously in association with ART initiation [52‐56]. The lack of consistent changes in brain metabolism in the wake of treatment initiation and improved measures of disease course may suggest that the abatement or decrease of HIV/SIV-related neuroinflammation requires a longer period than our study spanned. Changes in brain chemistry and function are likely to lag behind changes of conventional disease measures like CD4+ T-cell counts and viral load. This pattern has been shown in other imaging modalities: a magnetic resonance spectroscopy study of ART initiation found that while 3 months of treatment improved the clinical biomarkers of HIV+ patients, levels of brain metabolites that measure brain injury remained abnormal [57]. Other PET studies have more directly monitored cerebral inflammation post-treatment: two groups assessed cerebral levels of the translocator protein (TSPO), a marker of microglial inflammation, finding elevations in TSPO binding for HIV+ individuals even when the subjects were optimally treated and neurocognitively asymptomatic [58‐60]. Interestingly, one of those studies failed to find an association between TSPO binding and levels of CSF cytokines including eotaxin, MIP-1β, IL-8, and IL-10 [61‐64]. This finding is notable, as our own experiment failed to find a correlation between FDG uptake and cytokine levels post-initiation, and suggests that treatment with ART may result in an unlinking of these variables under certain circumstances.

Our data should be interpreted with some qualifications. Our sample size is relatively small. The subjects, even within a single cohort, were relatively heterogeneous in terms of baseline cytokine levels which could have played a role in their subsequent reaction to treatment and/or treatment removal. While treatment regimens were carefully considered, one antiretroviral, Raltegravir, was mixed in with food and its intake could be affected by the animal’s eating habits, especially during sickness. Future experiments might address some of these issues, as well as incorporate longer follow-up periods and more frequent sampling, especially within the first month of treatment modification when the most dramatic changes tend to occur. Protocols might also seek to incorporate interruption and initiation guidelines which more closely resemble those used clinically, such as CD4+ T-cell count thresholds or viral load limits that are tied to treatment resumption. For a more direct assessment of inflammation, the usage of radioligands that bind directly to inflammatory markers like the translocator protein (TSPO), a mitochondrial membrane receptor known to be upregulated in activated microglia, would offer an additional perspective on inflammatory changes post-interruption and initiation of ART.

Our study showed that treatment interruption was associated with rapid increases in brain metabolism along with disease intensification, with correlations between imaging and laboratory data. We also showed that treatment initiation, while improving biomarkers of disease, does not moderate pre-treatment brain metabolism within the six-month time period studied. Based on our findings, we believe that even brief gaps in treatment can potentially lead to negative outcomes, both molecularly and functionally. As a whole, our data support modern treatment guidelines, which advocate treatment as early as possible and maximal adherence to ART regimens [65].