Keywords

1 Introduction

Despite viral suppression with antiretroviral therapy (ART), no cure for HIV exists, and comorbid disease now represents the major challenge for people living with HIV (PLWH). HIV-associated neurocognitive disorders (HAND) affect approximately 18–69% (variability related to characteristics of study population with differing risk factors) of PLWH despite viral suppression with ART (Heaton et al. 2010; Simioni et al. 2010; Sacktor et al. 2016; dos Santos-Silva et al. 2017; Robertson et al. 2007; Cysique and Brew 2011) and are independent risk factors of mortality (Ellis et al. 1997; Tambussi et al. 2000). HAND in the context of virally suppressive ART mainly impairs cognitive function, although mild motor and behavioural deficits are still common. HAND induces brain injury including total and regional brain volume reduction (Cardenas et al. 2009; Küper et al. 2011) as well as neurodegeneration at least in its broadest sense (Boban et al. 2017). However, the complete neuropathological characterisation of HAND with viral suppression is still evolving (Gelman 2015). Thus, HAND may dramatically affect a person’s ability to function and live independently. Typical examples include impacting a person’s ability to drive and use the Internet which may result in social withdrawal and relationship breakdowns (Woods et al. 2017). Furthermore, there is strong evidence to suggest that individuals with HAND are more prone to poor medication adherence (Hinkin et al. 2002; Kamal et al. 2017) which may cause antiviral drug resistance and HIV disease progression. Together with cardiovascular disease and non-AIDS-cancers, HAND represent the major burden that continue to affect virally suppressed PLWH.

HAND is subdivided into three categories (by increasing cognitive and functional severity): (1) asymptomatic neurocognitive impairment (ANI), (2) mild neurocognitive disorders (MND) and (3) HIV-associated dementia (HAD) (Antinori et al. 2007). Effective viral suppression with current ART regimens has reduced the incidence of HAD (once a major cause of HIV-associated mortality) from ~20% to 2–4% of PLWH (Heaton et al. 2010). However, the incidence of ANI and MND remains high in virologically suppressed individuals resulting in impairments that can predominantly affect learning and/or memory functions and complex attention (Heaton et al. 2010, 2011; Robertson et al. 2004; Harezlak et al. 2011; Cysique et al. 2004). Furthermore, the presence of ANI has also been linked to a twofold to sixfold increased risk of earlier development and more rapid onset of symptomatic HAND (Grant et al. 2014), highlighting the potential progressive nature of the disease. Thus, HAND is a significant health, social and financial burden affecting PLWH worldwide.

Traditionally, the pathogenesis of HAND, especially HAD, has been strongly associated with HIV viremia (i.e. low nadir CD4+ T cell count; high plasma and/or CSF viral load). However, HAND in the context of virally suppressive ART is likely related to other factors such as low-level viral replication of whole virus or viral components, epigenetic changes and even systemic inflammation driven by gastrointestinal dysfunction, factors which may present greater prognostic/diagnostic value in this era of suppressive ART. Furthermore, HAND does not affect all PLWH, indicating that some individuals are, for as yet undefined reasons, more susceptible to disease possibly implying that host cell factors may also play a role in disease pathogenesis. Here, we assess the potential mechanisms driving the pathogenesis of HAND in chronically HIV-infected virally suppressed individuals and how they may be targeted to improve health outcomes for people living with HIV (Fig. 1).

Fig. 1
figure 1

Insight from biomarkers of HAND into disease pathogenesis in the presence of ART. HIV-infected monocytes (and T cells) transmigrate the blood-brain barrier (BBB) to seed perivascular macrophages, astrocytes and microglia. Viral persistence may drive activation of microglia and astrocytes inducing pro-inflammatory cytokine production, altered glutamate uptake and altered metabolism which drives neuronal injury and degradation contributing to BBB damage, brain atrophy and ultimately HIV-associated neurocognitive disorders. Both CNS-derived and peripheral biomarkers of HAND are shown

Full size image

2 HAND Prevalence, Diagnosis, Prognosis and Treatment

HAND is diagnosed by clinical neurological assessment with neuropsychological examination followed by radiology such as magnetic resonance imaging of the brain and CSF analysis. Since 2007 the ‘Frascati’ criteria – a consensus proposed by the National Institute of Mental Health and the National Institute of Neurological Diseases and Stroke (Antinori et al. 2007) – have been used to assess individuals’ neurocognitive capabilities, by scoring their (1) motor skills, (2) sensory-perceptual capacity, (3) speed of information processing, (4) memory (learning and recall), (5) abstraction and executive function, (6) attention and working memory, (7) verbal and language skills and (8) whether individuals need assistance with daily living (Antinori et al. 2007), with impairments in two or more cognitive areas indicative of disease. Scores are compared to the mean scores of age- and education-matched neurologically unimpaired individuals, and diagnosis is made as below:

HAD

Similar to dementia in HIV-uninfected individuals, HAD causes significant deficits in cognitive function and is the most severe form of HAND (Antinori et al. 2007; Vivithanaporn et al. 2010; Tozzi et al. 2005). As such, affected individuals have neurological impairments reflected of at least two standard deviations below those of neurologically unimpaired individuals in at least two cognitive domains, and they show mark decline in everyday function. Although the incidence of HAD has reduced globally (Heaton et al. 2010), HAD remains a significant burden in regions where access to efficacious ART is limited or non-existent with a 31–38% incidence rate (Wong et al. 2007; Lawler et al. 2010). Notably, HAD diagnosis is an independent predicator of mortality in individuals with advanced HIV/AIDS (Sevigny et al. 2007).

MND

MND affects ~11.7% of virally suppressed PLWH receiving ART (Heaton et al. 2010) and interferes with daily functioning mildly to moderately. It requires acquired impairment of at least one standard deviation below that of matched neurologically unimpaired individuals in two cognitive domains (Antinori et al. 2007). Typically, cognitive impairments associated with MND are detected by either self-reporting or observation by close relations. Impairments may present as mild interference in mental acuity; absenteeism; issues regarding relationships and social functioning; and/or poor adherence to ART, for which some individuals may require professional assistance (Hinkin et al. 2002).

ANI

ANI is the most recent subcategory of HAND (Antinori et al. 2007) and has been estimated to affect ~32.7% of PLWH despite viral suppression, representing ~70% of HAND cases (Heaton et al. 2010). Similar to the diagnosis of MND, individuals with ANI exhibit acquired impairment of at least one standard deviation below those of neurologically unimpaired individuals in two cognitive domains (Antinori et al. 2007). However, unlike individuals with MND, those with ANI do not exhibit interference with everyday living (Antinori et al. 2007). As such, ANI remains difficult to diagnose, and the clinical implications of ANI remain unclear and controversial (Torti et al. 2011; Nightingale et al. 2014). For example, some suggest that due to the asymptomatic nature of ANI, the ‘diagnosis’ of ANI may be a result of suboptimal performance in neurocognitive tests due to subjects being ‘tired’ or ‘distracted’ (Levine et al. 2017) and impart an unnecessary categorisation that could incur a degree of stigmatism (Chiao et al. 2013).

Conversely, growing evidence suggests that individuals with ANI are prone to disease progression with a twofold to sixfold increased risk of earlier development of symptomatic HAND (Grant et al. 2014), highlighting the importance of early identification affected individuals. These findings are supported by large longitudinal clinical studies describing disease progression including in patients with ANI (Sacktor et al. 2016; Gott et al. 2017; Heaton et al. 2015). Results from the CNS HIV Antiretroviral Therapy Effects Research (CHARTER) cohort found that 22.7% of participants (436 individuals in total) declined whilst 60.8% remained stable (Heaton et al. 2015). Similarly, the Multicentre AIDS Cohort Study (MACS) showed that whilst the majority of individuals (77%) remained neurocognitively stable, 13% deteriorated to a more severe form of HAND and 10% improved (Sacktor et al. 2016). Further, our recent data have shown that ANI has a biological underpinning of focal frontal brain atrophy, thereby validating its importance (Nichols et al. 2019). Despite these findings, the direct pathogenic mechanisms driving disease progression remain unclear.

2.1 Treatment

One of the early issues with HIV antiretroviral therapies was the variable and often poor ability of drugs to cross the blood-brain barrier (BBB) and target virus in the CNS. Treatment with antiretrovirals with the capacity to penetrate the CNS well (including abacavir, lamivudine, maraviroc, nevirapine, darunavir, lopinavir, dolutegravir and possibly other integrase inhibitors) is associated with improved neurocognitive functions (Cysique et al. 2009) and increased suppression of CSF HIV RNA in comparison to treatment with other drugs with poorer CNS penetrance scores (Carvalhal et al. 2016). However, some evidence suggests that standard ART regimens in some patients are sufficient (Lanoy et al. 2011; Ellis et al. 2014), possibly because there is already BBB impairment which allows drug access to the brain. Definitive results from randomised controlled trials have been hampered by under-recruitment, unexpected drug resistance, confounds and underappreciated delayed time to maximum efficacy of pre-trial entry therapies (sometimes up to 6–12 months) (Ellis et al. 2014). Current guidelines recommend the immediate initiation of therapy following the diagnosis of HIV (Panel on Antiretroviral Guidelines for Adults and Adolescents). Nonetheless, recent results from the START neurology sub-study have shown no neurocognitive advantage of immediate vs delayed initiation of therapy in individuals with CD4+ T cell counts >500 cells/μL (Wright et al. 2018). Novel therapeutic strategies, regimens and limitations are further discussed later in this chapter.

3 Risk Factors of HAND

HAND is influenced by many risk factors including current and nadir CD4+ T cell count, duration of HIV disease, age, co-infection, other comorbidities, notable diabetes and substance use (Saylor et al. 2016). Although viral suppression with ART has improved the life expectancy of PLWH to almost that of uninfected individuals (Trickey et al. 2017), chronic HIV-infection presents its own issues as age is an independent risk factor for HAND and chronic HIV infection is hypothesised to induce ‘premature ageing’ as people living with HIV exhibit higher risk and possibly earlier onset of age-related disease than uninfected individuals. Whether the mechanisms underlying the ‘premature ageing’ phenomenon reflect true ‘accelerated ageing’ (i.e. earlier onset and/or faster progression of disease) or ‘accentuated ageing’ (i.e. heightened severity of disease with same time of onset) is unclear, but work assessing onset/severity of neurocognitive disorders in people living with HIV supports a combination of both scenarios (Sheppard et al. 2017). Older PLWH (>50 years) are 3.26 times more likely to develop HAD than younger PLWH (<40 years) (Valcour et al. 2004) and show more rapid rates of progressive atrophy despite therapy, particularly in subcortical regions, than age-matched individuals (Clifford et al. 2017; Pfefferbaum et al. 2014; Nir et al. 2019). Older PLWH also show deficits in episodic memory and motor function relative to age-matched HIV-uninfected individuals (Goodkin et al. 2017), and functional brain changes in blood flow and brain age appear similar to those observed in ~15–20 year older HIV-uninfected individuals as measured by functional MRI (Ances et al. 2010). Furthermore, recent findings from a longitudinal study assessing 549 people living with HIV for up to 14 years have shown that these individuals exhibit accelerated age-related decline in brain volume, especially in the frontal cortex as measured by MRI (Pfefferbaum et al. 2014, 2018). Importantly, whilst these age-related changes were accentuated in individuals with a history of substance abuse and/or HCV co-infection, they persist even in individuals without these cofounders (Pfefferbaum et al. 2014, 2018).

The discrete mechanisms driving accelerated/accentuated age-related changes are unknown, but persistent inflammation and immune activation are at higher levels than the ageing process alone can explain (Angelovich et al. 2015). Specifically, measures of glial activation, neurotransmitters and ketone bodies in the CSF of PLWH mimic levels observed in older HIV-uninfected individuals and also correlated with age and adverse neurocognitive scores (Cassol et al. 2014). Recently, CSF levels of galectin-9, an immune modulatory protein linked to HIV persistence, was also found to be higher in virally suppressed individuals (Premeaux et al. 2019). However, levels were only indicative of neurocognitive impairment in older individuals, but not younger individuals. Furthermore, structural neuroimaging analyses by MRI suggests that virally suppressed PLWH experience accentuated brain atrophy, indicative of age-related changes, in comparison to HIV-uninfected controls (Cole et al. 2017). Epigenetic changes have also been implicated as the ‘brain age’ of PLWH with HAND, estimated by the epigenetic clock using DNA methylation analysis, appears 3.5 years older than individuals without HAND (Levine et al. 2016). Circulating HIV DNA in PBMCs also correlates with neurocognitive impairment, namely, in executive function, in older PLWH despite viral suppression (de Oliveira et al. 2015), also suggesting the role of HIV reservoirs in disease pathogenesis.

It is of note that whilst the extended duration of HIV infection is known to influence the pathogenesis of HAND, the median age of HIV infection may also influence neurocognitive disease prognosis. In 2016 over 17% of new HIV notifications in the USA were from individuals older than 50 years of age (Centers for Disease Control and Prevention 2017). This is problematic as older individuals tend to present with more advanced clinical disease as they are not routinely tested for HIV despite exhibiting some similar risk factors as younger persons such as multiple sexual partners (Asher et al. 2016). Furthermore, the dynamics of disease pathogenesis, and the use of biomarkers used to identify disease, may also change with age-related risk factors. Specifically, in a study by Fogel and colleagues, lipid dysfunction was found to be more highly associated with HAND in older PLWH, whilst methamphetamine use was more strongly associated with HAND in younger people (Fogel et al. 2015). Thus, chronological age, duration of infection and age of infection are significant confounders in HAND pathogenesis and may influence the efficacy of prognostic biomarkers of disease.

HAND is often associated with several comorbidities as HIV disease progresses (Guaraldi et al. 2011). As such, cardiovascular disease with risk factors such as hypercholesterolemia and hypertension is associated with neurocognitive impairment in virally suppressed PLWH (Wright et al. 2010). HAND has also been associated with dyslipidaemia in the CSF (Bandaru et al. 2013) as well as glomerular filtration rate, an indicator of both kidney function and CVD (Yuen et al. 2017). Non-HAND neurocognitive diseases such as Alzheimer’s disease (AD) may also confound HAND diagnosis. Until recently it was controversial whether PLWH have a propensity to develop AD. Beta-amyloid, characteristic of AD, has been found in brain tissue of PLWH (Green et al. 2005), and phosphorylated tau has been found in the CSF and plasma of PLWH and was associated with HAND (Brew et al. 2005), however, not in all cases (Krut et al. 2017; Clifford et al. 2009). PET scans have provided the first conclusive evidence of beta-amyloid accumulation in the CNS of a HIV-infected individual (Turner et al. 2016), suggesting the possible co-presence of HAND and/or AD. Thus, delineating HAND from AD, and more importantly treating appropriately, is of the utmost importance. Whilst PLWH appear at increased risk of Parkinson’s disease (Tisch and Brew 2009, 2010), the issue remains controversial and requires further study (Moulignier et al. 2015). It seems likely that the increased risk only pertains to those who already have HIV brain involvement as the mechanism is probably a ‘double hit’ to the basal ganglia and dopaminergic pathways both of which are also targets for HIV.

Host gene factors pose a risk to developing HAND with several immune-related genotypes associated with higher risk including CCR5-wt/wt (Singh et al. 2003), 195ApoE ε4, MBL2-O/O (Spector et al. 2010) and CCL2-2578G (Thames et al. 2015), although these genotypes do not encompass all individuals with HAND. Co-infection also contributes to HAND pathology. Cytomegalovirus (CMV) is a common co-infection in people living with HIV, and anti-CMV IgG levels have recently been linked with neurocognitive impairment in PLWH (Brunt et al. 2016; Letendre et al. 2018). Hepatitis C virus co-infection also confers ~twofold increased risk of cognitive impairment than HIV-infection alone (Ciccarelli et al. 2013).

4 HAND Pathophysiology

4.1 Natural HIV Infection of the CNS

HIV probably infects the CNS via a ‘Trojan horse’-like mechanism, whereby HIV-infected peripheral blood monocytes, particularly the CD14+CD16+ subset (Veenstra et al. 2017; Ellery et al. 2007), and T cells cross the BBB (Fischer-Smith et al. 2001; Honeycutt et al. 2018) (Fig. 1). Current thinking considers that BBB damage is a result of CNS infection though the reverse may be true to an extent – that is, systemic infection may also directly damage the BBB. During acute infection, HIV-1 RNA is detected in human CSF as early as 8 days postinfection (Valcour et al. 2012) and probably continues throughout natural infection in response to chemokines produced by microglia and astrocytes, as seen in individuals diagnosed with HAD (Persidsky et al. 1999). Furthermore, monocyte DNA content in untreated individuals is associated with progressive stages of HAND (Valcour et al. 2013). As such, recruitment of HIV-infected monocytes from sites such as the bone marrow to the CNS is thought to be a major contributor to CNS infection and neurological impairment (Burdo et al. 2010; Strickland et al. 2014).

Following transmigration to the CNS, monocytes mature into perivascular macrophages, which may (1) sustain HIV replication and mediate infection of surrounding cells, such as CD4-expressing macrophage and microglia (Fischer-Smith et al. 2001; Thompson et al. 2011; Burdo et al. 2013a; Kim et al. 2006), and (2) harbour latent HIV reservoirs (Thompson et al. 2011). We and others have shown that up to 19% of astrocytes, which express little to no CD4, are infected with HIV in individuals with HAD (Churchill et al. 2009), can latently harbour HIV and may mediate trans-infection HIV to CD4+ T cells (Gray et al. 2014). A recent report also suggests that pericytes forming the BBB may also be a site of HIV infection (Cho et al. 2019). Although HIV-infected CD4+ T cells and cell-free virus also enter the CNS, likely due to breakdown of the tight junctions forming the BBB (Strazza et al. 2011), macrophage-driven viral replication is thought to be the primary source of HIV replication in the CNS of patients at least those with severe HAND (Schnell et al. 2009).

Following CNS infection, HIV viral proteins, such as nef, vpr, gp120 and, in particular, extracellular tat (Bagashev and Sawaya 2013), activate surrounding macrophages, microglia and astrocytes to release cytokines and chemokines that drive localised neuroinflammation (Shah et al. 2011a, b; Shah and Kumar 2010; Sami Saribas et al. 2017); the effects of which target multiple sites in the CNS. Firstly, cytokines and chemokines can activate and actively damage the BBB, resulting in enhanced permeability and monocyte recruitment (driven by MCP-1 and CCL2 (Conant et al. 1998)), particularly of CD14+CD16+ monocytes (Williams et al. 2013). Secondly, cellular activation can act to enhance HIV replication (osteopontin, TNF and CXCL-10 (Brown et al. 2011; Williams et al. 2009)). Finally, components of the inflammatory milieu such as CXCL-10 (Mehla et al. 2012) and quinolinic acid (Kandanearatchi and Brew 2012) ultimately have neurotoxic effects on neurons via the generation of reactive oxygen species and lipid peroxidation; disruption of calcium homeostasis of neurons (Haughey et al. 1999) leading to neuronal apoptosis (IL-1β, IL-8, TNF (Guha et al. 2012)) and dysfunction of astrocytes ability to buffer glutamate via the N-methyl-D-aspartate receptor; all of these contribute to neuronal dysfunction, apoptosis and neurodegeneration. Neuronal injury occurs early in infection with PLWH with primary, untreated infection displaying heightened measures of neuronal injury such as CSF levels of neurofilament light chain protein (NFL) and neurometabolite dysfunction as determined by proton-magnetic resonance spectroscopy (MRS) (Peluso et al. 2013). Thus, HIV-induced neuroinflammation acts as a positive feedback loop driving HAND and other pathological injuries.

One major complication of uncontrolled HIV infection is HIV-associated encephalitis (HIVE), a neuropathological correlate of HAD. HIVE is strongly associated with reduced brain volume (as measured by MRI) and chronic immune activation and inflammation. Specifically individuals with HIVE express a higher percentage of CD163+CD14+CD16+ perivascular macrophages in the brain (Fischer-Smith et al. 2001; Kim et al. 2006); an expansion of ‘inflammatory’ CD16+-expressing monocytes in the periphery (Ancuta et al. 2008); and multinucleated giant cells, comprising aggregated microglia and macrophage cells (Brew et al. 1995; Takahashi et al. 1996). Furthermore, individuals with HIVE have widespread dysregulation of genes involved in synapto-dendritic functioning and integrity, toll-like receptors, interferon responses, mitochondrial genes, synaptic transmission and cell-to-cell signalling (Masliah et al. 2004; Gelman et al. 2012). Recent evidence also suggests that microglia in individuals with HIVE are also dysfunctional, with lower gene expression levels of functional markers (Ginsberg et al. 2018). Thus, HIVE is strongly associated with advanced neurocognitive impairment such as HAD.

Finally, neurotropic HIV variants, geno- and phenotypically distinct from peripheral viruses, exist in PLWH and may drive disease (Gorry et al. 2001, 2002; Haggerty and Stevenson 1991). These variants are detectable at different stages of infection and induce inflammatory responses such as increased neopterin and pleocytosis in CSF (Sturdevant et al. 2015; Hagberg et al. 2010), suggesting that a compartment of neurotropic/neurovirulent viral strains may play a more critical role in HAND than absolute viral load in the CNS.

4.2 HAND Pathophysiology Following Viral Suppression with ART

The persistence of HAND despite viral suppression with ART indicates an underlying pathology in the CNS. However, the exact pathophysiology of disease is unclear mainly due to the inability to perform biopsy analysis of the CNS from individuals who are virally suppressed with ART and limited numbers of autopsy samples from such patients. Therefore, most evidence of ongoing pathology in virally suppressed individuals is based on peripheral indicators of CNS disease or MRI/PET scans. Individuals with HAND have heightened indicators of persistent neuronal injury (as measured by cerebral metabolites) (Young et al. 2014), degradation of cortical grey matter, BBB dysfunction (Chaganti et al. 2019), microglial activation (Rubin et al. 2018; Garvey et al. 2014) and a continued reduction of brain volume despite therapy (Cardenas et al. 2009; Küper et al. 2011). Indicators of axonal damage such as CSF neurofilament light chain are also elevated in virologically suppressed PLWH (Jessen Krut et al. 2014) and are associated with cognitive decline (Sun et al. 2017; Sailasuta et al. 2016). Notably, osteopontin, a pro-inflammatory bone-matrix protein, is also found in high levels in virally suppressed individuals with HAND (Brown et al. 2011; Yu et al. 2017; Vera et al. 2016). These biomarkers of disease, discussed further below, are indicative of chronic neuroinflammation and neurodegeneration despite ART (Figure 1).

Virally suppressed individuals also show peripheral indicators of chronic cellular activation and dysfunction that may indicate/contribute to neuropathology. Individuals with HAND have a higher number of activated macrophages/microglia in the CNS relative to neurologically unimpaired individuals (Tavazzi et al. 2014). Plasma levels of MCP-1 and CXCL-10, produced by activated macrophages and monocytes, are associated with neuronal injury and higher inflammation in virally suppressed individuals with HAND (Letendre et al. 2011). Levels of quinolinic acid, a product of tryptophan metabolism, are also elevated in virally suppressed PLWH and are directly associated with levels of pTau that are predictive of neurocognitive impairment in this population (Anderson et al. 2018). Some evidence suggests that the perivascular macrophage phenotype may be predisposed by the precursor monocyte phenotype as monocyte activation (as measured by CSF sCD14) remains elevated in virologically suppressed individuals with HAND (Eden et al. 2007; Spudich et al. 2011; Yilmaz et al. 2013) and is associated with significantly worse neurological outcomes for those on non-suppressive ART (Kamat et al. 2012). Furthermore, PBMCs from individuals with HAD or MND express higher levels of pro-inflammatory cytokine genes (such as TNF, IL-6, IL-27) and lower levels of associated miRNAs (miR-124-3p, miR-210) than PLWH who were neurologically unimpaired (Venkatachari et al. 2017).

HIV-associated astrocyte dysfunction – CNS cells that normally help to maintain cerebral immunity and homeostasis – has been associated with HAND. Specifically, astrocyte dysfunction results in improper buffering of glutamate, dysregulation of BBB permeability, and release of pro-inflammatory cytokines (such as TNFα). Moreover, because astrocytes are susceptible to HIV infection, they may serve as a viral reservoir. We have further shown that astrocytes are less responsive to some antiretroviral drugs (Gray et al. 2013), thus potentially contributing to continued viral presence in the CNS.

5 Mechanisms Driving HAND in Virally Suppressed Patients

It is well accepted that HIV viremia is not the sole driver of neurocognitive impairment as incidence of HAND continues to affect between 18 and 69% of people living with HIV despite viral suppression with ART. Therefore, other factors such as persistent viral transcription in CNS reservoirs and/or chronic peripheral inflammation outside of the CNS play pertinent roles in determining the HAND status of virally suppressed individuals. These mechanisms will be discussed below.

5.1 HIV Viral Presence/Replication in the CNS

HIV is considered to persist in latent reservoirs in the CNS such as macrophages and astrocytes (Thompson et al. 2011) and is controversially thought to directly contribute to HAND pathogenesis (Churchill et al. 2015). HIV RNA (Dahl et al. 2014) and anti-HIV antibodies (Burbelo et al. 2018) are detectable in the CSF after 10 years of ART, potentially indicating viral persistence and ongoing replication in the CNS. Notably, increasing ratios of ‘HIV RNA in CSF’-to-‘HIV RNA in plasma’ are associated with worsening neurological outcomes in virologically suppressed PLWH (Anderson et al. 2017; Canestri et al. 2010) and elevated neopterin levels (Dahl et al. 2014), indicating that localised replication in the CNS may contribute to neuroinflammation and disease progression. A recent longitudinal study identified CNS escape in 6% of cases (6/101 individuals) in the absence of peripheral viremia following 3 years of ART (Joseph et al. 2019). Further genotypic analysis suggests that the virus was derived from persistent CNS replication, likely from macrophage/microglia, linked to emtricitabine resistance and low nadir CD4 count (10 copies/mm3). Interestingly, we have shown that some CNS-derived viruses have impaired transcriptional activity due to polymorphisms in the transcription factor Sp1 (Gray et al. 2016). Thus, whether virus in the CSF truly represents active replication in the CNS of virologically suppressed individuals remains unclear, possibly confounded by poor suppression in the periphery associated with a blip in plasma HIV levels (Eden et al. 2016). HIV resistance to lamivudine and emtricitabine (Mukerji et al. 2017) and also treatment with protease inhibitor + NRTI regimens (Mukerji et al. 2018) are also associated with CSF viral escape. Finally in vivo mouse studies have shown that chronic low levels of neurotoxic tat, a step in the HIV life cycle not affected by current ART, can cause astrocyte activation (Dickens et al. 2017) which may contribute to HAND pathogenesis as tat has been found in both CNS tissue and the CSF of virally suppressed individuals (Johnson et al. 2013). Thus, further studies are required to establish the role of active replication and/or viral persistence in the CNS in HAND associated pathology.

Whilst active replication is an important focus of current research in defining the CNS as a reservoir of HIV, persistence of non-replication competent virus in the CNS is also detrimental to the patient. Studies in the periphery have identified that the bulk (~93%) of the HIV reservoir in T cells consists of non-replication competent viruses harbouring hyper-deletions that may generate short abortive transcripts or viral proteins that are immunogenic (Bruner et al. 2016). As such, generation of highly neurotoxic HIV nef and tat in the CNS by persistent non-productive infection in cells such as astrocytes may drive neuroinflammation, cellular dysfunction and neuronal degradation (as discussed above). Thus, persistent non-replication competent virus in the CNS may play as important a role in HAND pathogenesis as active viral replication and must be considered in appropriate therapeutic strategies.

5.2 Host Cell and Epigenetic Factors

Host cell and environmental factors induced by microRNAs (miRNA) – short non-coding repressors of transcription – and DNA methylation have been identified to be associated with HAND pathogenesis (Tatro et al. 2010), suggesting a role for epigenetic changes to components of the immune system in driving HAND. Large genomic studies have found strong associations between dysregulated miRNAs and mRNA in the brains of individuals with HAD in comparison to neurocognitively unimpaired individuals (Zhou et al. 2012). Specifically, miR-21, known to repress transcription factors involved in neuronal functioning, is present in high levels in the CNS of PLWH with HAD (Yelamanchili et al. 2010). Furthermore, both miR-128a (Eletto et al. 2008) and miR-34a (Mukerjee et al. 2011) are upregulated by HIV Tat or Vpr, respectively, resulting in adverse effects on neuronal function in vitro. HIV Tat also influences histone acetylation/deacetylation via chromatin structure remodelling by histone deacetylases in neuronal cells in vitro, resulting in synaptic plasticity and neuronal dysfunction (Saiyed et al. 2011). Epigenetic changes derived by alterations in DNA methylation sites in monocytes are also strongly associated with worse neurocognitive scores (Corley et al. 2016), implying that epigenetic changes in peripheral cells transmigrating to the CNS may also influence HAND.

5.3 Chronic Peripheral Inflammation

Chronic immune activation and inflammation is a hallmark of HIV infection despite viral suppression, and measures of inflammation/immune activation are reliable indicators of HIV disease progression and have been linked with HAND pathogenesis (Hazenberg et al. 2003; Somsouk et al. 2015; Kuller et al. 2008; Boulware et al. 2011; Hunt et al. 2011; Pandrea et al. 2012). The causes of persistent immune activation in chronic HIV infection are multifactorial, but the host response to translocated microbial products from the gastrointestinal (GI) tract, resulting from the depletion of gut-associated lymphoid tissue CD4+ T cells during acute infection, is thought to be a major contributor (Brenchley et al. 2004, 2006; Estes et al. 2010). This is exemplified in elegant in vivo studies where experimental damage of the GI tract of African green monkeys (AGMs), a natural host of SIV that does not develop AIDS, induces local and systemic inflammation similar to that of HIV infection in humans (Hao et al. 2015). Specifically, bacterial lipopolysaccharide (LPS) – a cell wall component of gram-negative bacteria and potent immune activator – remains elevated in HIV/SIV infection despite ART and is thought to drive chronic inflammation (Deeks et al. 2004). Progression to AIDS in PLWH may be driven (in part) in response to LPS as sooty mangabeys, natural hosts of SIV that do not develop AIDS, have recently been shown to have a C-terminal frameshift in the LPS receptor (TLR-4) which renders it non-functional (Palesch et al. 2018).

Measures of microbial translocation, especially measures of monocyte activation following exposure to LPS such as sCD14 and sCD163, have been linked with neurocognitive impairment, indicating a central role for monocyte activation in neuropathology (Vassallo et al. 2013). Notably, elevated levels of circulating CD16+ monocytes, CD14+ monocyte HIV DNA content, sCD14 and sCD163 are associated with microbial translocation and correlate with an increased risk of neurocognitive impairment (Valcour et al. 2013). sCD14 levels in CSF are also associated with NFL, indicative of axonal damage, in untreated PLWH (Jespersen et al. 2016). Moreover, plasma LPS levels are higher in HIV-HCV co-infected individuals with HAND, relative to HIV-HCV co-infected neurologically unimpaired individuals (Vassallo et al. 2013). GI tract damage in PLWH is also associated with more pathogenic microbial populations, compared to uninfected individuals (Mutlu et al. 2014). Notably, similar alterations in the gut microbiome have been linked to other CNS disorders, namely, multiple sclerosis (Glenn and Mowry 2016) and Parkinson’s disease (Houser and Tansey 2017). Whilst it follows that, within PLWH, changes in the gut microbiome and the resultant inflammatory response may contribute to HAND, studies are required to test this hypothesis.

5.4 ART Neurotoxicity

Whilst current antiretrovirals are less neurotoxic than earlier iterations, ART neurotoxicity continues to be linked to HAND pathogenesis, and many ARVs may have lasting adverse ‘legacy’ effects that may contribute to disease decades after use (Jonathan et al. 2015). Efavirenz, a commonly prescribed non-nucleoside reverse-transcriptase inhibitor, is still in use despite known neurological side effects such as psychiatric episodes and neurocognitive effects on information processing and executive functioning (Ma et al. 2016). The underlying pathology of efavirenz is most likely due to either the induction of pro-inflammatory cytokines or mitochondrial toxicity (Funes et al. 2015) and endoplasmic reticulum stress (Funes et al. 2014; Bertrand and Toborek 2015). Treatment of neuronal cell lines with 15 different ARVs at physiological concentrations identified modest neuronal toxicity in vitro (Robertson et al. 2012). In vitro raltegravir has also been linked with enhanced production of macrophage-derived IL-8 (Tatro et al. 2014).

Protease inhibitors also have a higher risk of drug-drug interactions by metabolism via CYP450 enzymes and have also been linked to neurotoxicity. Treatment of human astrocytes with the protease inhibitors amprenavir and lopinavir induced dysregulation of glutamate transport in vitro and was associated with worse neurocognitive performance in mouse models (Vivithanaporn et al. 2016). In vitro ritonavir or lopinavir treatment also impaired oligodendrocyte maturation, and similar treatment in mice reduced myelin protein levels in the frontal cortex (Jensen et al. 2015). Importantly Jensen and colleagues further showed that levels of myelin basic protein in the frontal cortex of PLWH were lower than those present from untreated PLWH or HIV-uninfected patients. Finally, recent evidence has found that individuals treated with integrase stand transfer inhibitors (INSTIs) such as dolutegravir are associated with poorer neurocognitive performance and reduced brain volume than those on non-INSTI containing regimens (O’Halloran et al. 2019). As many patients transition to regimens including INSTIs, the long-term neurocognitive effects of these drugs need to be closely monitored. Thus, there remains a clinical demand for highly effective and non-neurotoxic antiretrovirals as well as improved treatment strategies to treat HAND.

6 Biomarkers of CNS Infection and HAND

Identifying biomarkers of cognitive impairment, and more specifically HAND, is a major research focus due to the clinical demand for reliable means to (1) predict HAND before it occurs, (2) diagnose HAND, (3) distinguish the early stages of HAND (ANI and MND) and (4) predict and monitor changes in cognition for individuals with HAND. Biomarkers must be readily accessible, cheap and reliable, and as such most emphasis has been placed on indicators of neuroinflammation or neuronal damage in the blood and CSF. Identifying biomarkers of HAND has proven to be a difficult task as well-established plasma and CSF biomarkers of inflammation (such as sCD14, osteopontin (Brown et al. 2011), TNFα, IFNγ, sCD14, IL-1β, IL-6, S100β, TGF-β levels) or HIV viremia (such as the surrogate marker nadir CD4+ T cell count) that are predictive of HAND in viremic or ART-naïve PLWH (Abassi et al. 2017; Lyons et al. 2011) are not predictive of HAND in virologically suppressed individuals receiving ART (Kamat et al. 2012; Lyons et al. 2011). Biomarkers of HAND may be indicative of neuronal damage, altered metabolism, generally neuroinflammation or even HIV itself. Specific biomarkers of HAND and their related pathological implications are discussed in detail below (Fig. 1).

6.1 Measures of HIV Viremia and Viral Latency

Although viral suppression with ART has reduced the predictive value of many traditional HIV-related biomarkers of HAND such as CSF HIV RNA (Ellis et al. 2002; Brew et al. 1997), some measures of HIV remain indicative of HAND despite therapy with ART. Low CD4+ T cell counts continue to be associated with HAND, albeit mainly in individuals failing therapy or with moderate immune suppression (CD4+ T cell count 200–349 cells/mm3) (Bhaskaran et al. 2008). CD4+ T cell nadir may offer some prognostic value (Ellis et al. 2011). However, a recent study found that immediate therapy initiation with a CD4 nadir >500 cells/mm3 had no neurocognitive benefit over delayed treatment initiation (Wright et al. 2018), questioning the predictive value of CD4 nadir at high CD4 T cell counts. Similarly, CSF and/or plasma levels of HIV RNA are also indicative of neurocognitive decline in individuals who exhibit therapy failure or drug resistance (Canestri et al. 2010; Peluso et al. 2012; Bingham et al. 2011).

Additionally, biomarkers of persistent viral replication/presence in the CNS are gaining attention as possible prognostic markers of disease. HIV DNA persists in the CNS despite viral suppression and studies using research tools including PCR and the in situ hybridisation technique DNAscope have offered insight into HIV persistence in the CNS (Churchill et al. 2009; Ko et al. 2019; Lamers et al. 2016; Estes et al. 2017). However, its application as a true diagnostic/prognostic biomarker is limited as these assays generally investigate autopsy brain tissue. Therefore, HIV DNA content in peripheral cells such as T cells and monocytes may be a more practical approach to assessing the CNS reservoir. HIV DNA in peripheral blood monocytes (Valcour et al. 2013; Cysique et al. 2015a) and CSF cells (Oliveira et al. 2017; Shaunak et al. 1990) is associated with risk of HAND and brain atrophy in non-suppressed individuals (Kallianpur et al. 2014) and recent evidence supports that this may also be true in suppressed patients. In a recent study of 69 people living with HIV who were virally suppressed for a median of 8.6 years with ART, Spudich and colleagues found that ~50% of individuals harboured cell-associated HIV DNA in the CSF (Spudich et al. 2019). Furthermore, cell-associated HIV DNA was associated with adverse neurocognitive outcomes, suggesting that persistent HIV in the CSF despite ART may contribute to HAND pathogenesis. BCL11B, an inhibitor of viral transcription, is upregulated in latently infected cells of the CNS (Desplats et al. 2013), and higher CSF levels are associated with lower creatine levels in frontal white matter (Cysique et al. 2019), potentially validating BCL11B as an indicator of latent HIV-related pathology. However, the associations of CSF BCL11B and HAND remain to be defined. CSF escape is also associated with higher neopterin levels (Eden et al. 2016), although as this is a rare phenomenon and neopterin is produced by multiple cells types during cellular activation, this may not be the most robust method of confirming latency. A recent study also found that HIV gp120 protein sequences are predictive of HAD and/or HIVE in silico (Ogishi and Yotsuyanagi 2018), highlighting the possibility of HIV quasi-species as a predictive tool for HAND. However, the methodology required to characterise sequences from isolated immune cells is more challenging than standard biochemical assays which may limit its practical application as a biomarker.

6.2 Neuronal Damage

NFL, a component of myelinated axons, is readily detected in the CSF of neurodegenerative diseases and is indicative of neuronal axon damage. Multiple studies have found that CSF levels of NFL are present at higher levels in untreated individuals with or without neurocognitive impairment and CSF NFL levels in individuals with HAD are significantly higher than those in more mild disease (McGuire et al. 2015; Gisslén et al. 2007, 2015; Peterson et al. 2014). Moreover, NFL is associated with worse CD4 T cell counts and is positively associated with plasma viral load, suggesting that it is indicative of HIV-directed neuropathology. Correlative analysis in individuals with primary infection further shows associations between higher NFL levels in CSF and neuroinflammatory markers (CXCL-10, neopterin) and adverse levels of CNS metabolites (Peluso et al. 2013). NFL has also been shown to be a more sensitive neuronal marker than total and phosphorylated tau and amyloid precursor proteins (Peterson et al. 2014). Furthermore, CSF NFL levels correlate with plasma NFL levels, a significant advantage due to ease of blood draw vs lumbar puncture required for CSF analyses. Therapy reduces NFL levels; however, they remain elevated in comparison to HIV-uninfected controls (Jessen Krut et al. 2014). However, abnormal NFL levels have only been detected in ~16% of virally suppressed subjects, questioning whether this is sensitive enough to detect ANI or MND (Peterson et al. 2014).

Exosomes are gaining attention as biomarkers of neuroinflammation and neurocognitive disorders including AD and HAND (Pulliam et al. 2019). Exosomes are 30–150 nm vesicles containing host cell proteins and DNA that are released following cellular activation into plasma. As plasma exosomes harbour protein and RNA from their host cell such as neurons, they potentially offer a method of interrogating cell-specific pathology from different, difficult-to-access sites such as the brain. Studies in AD have identified the presence of neuronally derived exosomes (NDE) from patients with preclinical AD that predicted AD 10 years prior to onset (Fiandaca et al. 2015). More recently, NDEs have been detected in the plasma of PLWH with cognitive impairment (Sun et al. 2017) and are indicative of neuronal health. Sun and colleagues identified that neurocognitively impaired individuals had fewer NDEs than non-neurocognitively impaired individuals, possibly due to neuronal stress or death. Furthermore, NDEs from these individuals contained higher levels of proteins associated with neuronal damage such as high motility group box 1 (HMGB1), NFL and amyloid β than exosomes from HIV-infected neurocognitively unimpaired individuals (Sun et al. 2017). Thus, exosomes, particularly NDEs, may offer some prognostic benefit over other inflammatory biomarkers that are not truly specific to the CNS.

Mitochondrial DNA (mtDNA) is indicative of the energy levels/demands of a cell. Each cell contains mitochondria with a number of copies of mtDNA; thus mtDNA copy number may be indicative of cellular dysfunction and apoptosis. mtDNA damage is present in approximately 45% of total cells in the frontal cortex of patients with HAND without viral suppression (Zhang et al. 2012). Furthermore, individuals with HAND in this cohort had less mtDNA than those who were not cognitively impaired. Similarly, mtDNA damage can be found in the periphery. Following viral suppression, mtDNA copy number from peripheral blood is associated with worse cognitive performance in CHARTER patients (Hulgan et al. 2018). Furthermore, CSF levels are higher in MND patients in comparison to neurocognitively normal individuals (Mehta et al. 2017); however, no difference was observed in ANI patients – this may just be a reflection of the insensitivity of the CSF as a marker of mild events in a discreet part of the frontal lobe.

6.3 Cellular Activation

Monocytes, and markers of monocyte of activation, are readily measurable indicators of HAND due to their key role in pathogenesis. CCR2 expression on CD14+CD16+ monocytes is negatively associated with neurometabolite levels, and by extension pathology, as measured by MRS imaging (Veenstra et al. 2019). CCR2 expression on CD14+CD16+ monocytes is also associated with HIV DNA copies per 106 PBMCs (Veenstra et al. 2019). CCR2’s ligand, CCL2, is a chemokine that acts to recruit monocytes into the brain. In individuals with the CCL2-2578G, CSF levels of CCL2 are higher than in non-carriers, and higher levels are associated with poor cognitive impairment, albeit mainly in individuals with high plasma viral (Thames et al. 2015). Plasma CCL2 levels are also elevated in individuals with either HAD or minor cognitive motor disorder (precursor to MND) in viremic individuals (Ancuta et al. 2008). However, it is of note that levels are similarly elevated in individuals with non-HIV-related neurocognitive impairments and no difference in ANI patients, and neurocognitively normal individuals were observed in this study.

Soluble indicators of monocyte activation have also been linked to HAND. LPS activates monocytes via TLR-4 as described above, and plasma levels are associated with HAD in individuals with AIDS (Ancuta et al. 2008). LPS levels are known to remain elevated following viral suppression; however, to date no evidence links plasma LPS levels with HAND in the absence of co-infection such as HCV. Nonetheless, following monocyte activation by immunogens such as LPS, monocytes shed CD14, the co-receptor for LPS, which can be measured and has been associated with worse global deficit scores, lower CD4 T cell counts and worse learning, attention and motor T scores in a cohort of generally unsuppressed individuals (Lyons et al. 2011). The reliability and sensitivity of sCD14 to identify HAND in virally suppressed individuals is less clear. Similarly, soluble CD163 – a scavenger receptor – is shed from activated macrophage/microglia and detectable in both plasma and CSF. sCD163 levels are associated with adverse neurocognitive scores (Burdo et al. 2013b; Royal et al. 2016).

Neopterin, a pteridine produced by activated macrophages and monocytes, is commonly found in higher levels in the plasma and CSF of PLWH, despite viral suppression. Neopterin levels have been shown to predict HAND and are associated with HAND severity in therapy-naïve individuals (Brew et al. 1990, 1996); however, there is limited evidence supporting the prognostic value of neopterin in virally suppressed individuals. Plasma levels, but not cellular gene expression in monocytes (Quach et al. 2018), of CCL2, HMGB1, IL-8, CXCL-10 and neopterin are predictive of HAND in virally suppressed individuals (Sun et al. 2017; Eden et al. 2016; Yuan et al. 2013). CSF fractalkine levels are higher in individuals with MND/HAD than ANI (Letendre et al. 2011). CSF levels of HMGB1, produced by activated macrophages and astrocytes, may also be predictive of HAND, even very minor disease, in virally suppressed PLWH (Gougeon et al. 2017).

Whilst many peripheral biomarkers of monocyte/macrophage activation are associated with adverse neurocognitive outcomes (as discussed above), individual biomarkers (such as sCD14, MCP-1, CXCL-10, etc.) are associated with unique neuropathology and inflammation at different locations in the brain (such as frontal white matter, basal ganglia, grey matter). For example, MRS studies assessing levels of neurometabolites in the CNS of virally suppressed individuals found that CSF levels of sCD14 were negatively associated with Cho/Cr ratios in frontal white matter (Anderson et al. 2015), indicative of monocyte-induced neuronal damage. However, MCP-1 levels, indicative of monocyte recruitment, were positively associated with Cho/Cr ratios in the basal ganglia (Anderson et al. 2015), supporting findings from similar studies (Harezlak et al. 2011). Thus, it is important to acknowledge that each biomarker of myeloid activation may be indicative of different stages and/or sites of neurocognitive diseases which must be considered when interpreting results.

S100β is a calcium binding protein mainly expressed on glial cells such as astrocytes and thus is useful as a biomarker of astrocyte activation. S100β levels in the CSF have been associated with cognitive impairment in PLWH and were higher in those with more severe AIDS dementia complex (precursor terminology to HAND) or progressed to more advanced disease more quickly (Pemberton and Brew 2001). S100β levels are also inversely associated with executive function including the rapid generation of verbs testing ‘action fluency’ (Woods et al. 2010).

Finally, measures of T cell activation are also altered with symptomatic HAND (HAD and MND) exhibiting different T-cell immune activation profiles to those with ANI (Vassallo et al. 2015) (diagnosis of MND was associated with a CD4:CD8 ratio <1), suggesting the pathophysiological mechanism underlying symptomatic- and asymptomatic-HAND is different or the difference could reflect a longer time with symptomatic disease given that ANI leads to MND. T-cell phenotype/function and HIV itself are also indicative of neurocognitive impairment. The percentages of IFN-γ producing CD8+ T-cells in the CSF are directly associated with the severity of neurocognitive decline as measured by global deficit scores in virally supressed PLWH (Schrier et al. 2015). However, whether these observations are indicative or causative in disease pathogenesis is unknown.

6.4 Metabolic Dysfunction

PLWH with neurocognitive impairment experience dysregulated kynurenine pathways of tryptophan catabolism, skewing to the production of neurotoxic metabolites. As such CSF levels of quinolinic acid/tryptophan ratios are indicative of CNS neurocognitive impairment in SIV-infected macaques, particularly early in infection (Drewes et al. 2015; Heyes et al. 1990). Furthermore, a recent study showed that virally suppressed individuals had higher CSF levels of quinolinic acid and kynurenine/tryptophan ratios than HIV-uninfected controls (Anderson et al. 2018). Similarly, HAND has been associated with lipid dysfunction and cholesterol accumulation in the brain (Mielke et al. 2010). In a large multi-centre study, HIV-infected virally suppressed individuals had altered levels of cholesterol species in the CSF; namely, an increase in the ceramide species C24:1 that was associated was higher likelihood of neurocognitive decline (Bandaru et al. 2013). Furthermore, a shift in lipid species was associated with progressive disease.

6.5 Novel Non-traditional Biomarkers

Progranulin is produced by microglia and neurons and is thought to have neuroprotective effects by modulating neuroinflammation whilst also acting as a neuronal growth factor. Progranulin also has known anti-viral effects that inhibit HIV replication, and its production by microglia is upregulated during active HIV infection (Suh et al. 2014a). However, during viral suppression with ART, lower levels of progranulin in the CSF are present and are associated with worse cognitive impairment and elevated levels of other pro-inflammatory biomarkers (Suh et al. 2014b). This may be because downregulation of the production of neuroprotective progranulin by microglia may leave neurons more susceptible to damage (Suh et al. 2014a). Thus, low progranulin levels in the CSF may offer insight into underlying CNS neuronal damage and HAND pathogenesis.

Total tau (t-tau) and phosphorylated tau (p-tau) levels in the CSF are commonly used as an indicator of Alzheimer’s disease; however, levels have also been assessed in the context of HAND. We and others have found that CSF levels of p-tau (as well as total-tau levels) are higher in individuals with AIDS dementia complex and HAND than those who are uninfected (Brew et al. 2005; Cysique et al. 2015b), which is in line with more recent reports where p-tau levels in CSF were inversely associated with scores of prospective memory in PLWH (Anderson et al. 2018). However, others have found a difference in only CSF levels of t-tau, but not p-tau (Krut et al. 2017; Steinbrink et al. 2013) in comparison to the HIV-uninfected population, possibly suggesting that patient viral load and/or ongoing neuronal damage at the time of sampling may influence these associations.

MicroRNAs are non-coding molecules that regulate both viral and host gene expression by binding and repressing RNA and can be detected in PBMCs, plasma and CSF. A small cross-sectional study found that HIVE patients express altered CSF levels of 11 different miRNAs in comparison to subjects without HIVE (Pacifici et al. 2013), with miR-937 being highly upregulated in the CSF. Furthermore, levels of miRNAs in plasma are different between individuals with or with HAND (Kadri et al. 2016; Asahchop et al. 2016) making miRNAs a potentially useful biomarker of disease. The Veterans Aging Cohort Study (VACS) index, consisting of traditional HIV biomarkers of disease and non-traditional biomarkers of comorbidity, is also predictive of neurological decline (Marquine et al. 2016).

6.6 Limitations of Biomarkers and Current Challenges

Whilst the aforementioned biomarkers are indicative of cognitive decline in PLWH, no clinically validated biomarker can reliably differentiate the stages of HAND, especially ANI. This may be because many are broad inflammatory components that are not uniquely specific to the brain, or HIV infection, and thus they may be influenced by inflammation from other sites or other diseases/disorders or even age. Furthermore, it is likely that ‘waves’ of neuroinflammation may take place during disease pathogenesis, confounding the accuracy and specificity of biomarkers with HAND. This is evidenced by diffusion tensor imaging assessing white matter damage where reductions in white matter in chronic infection were associated with markers of inflammation (Wright et al. 2015), whilst changes in white matter in primary infection were associated with breakdown of the BBB, thus, representing a change in neuropathogenesis and neuroinflammation with disease progression. Therefore, different biomarkers of disease may be representative of stages of neuropathogenesis that are not strictly related to Frascati Criteria stages of disease. In summary, whilst much work has been conducted evaluating biomarkers of HAND, a need remains for cheap, accessible, and highly accurate biomarkers that accurately diagnoses, distinguishes and compares distinct forms of HAND. Combinations of biomarkers are more likely to be more predictive or diagnostic of disease.

7 Biomarkers of Neurocognitive Decline as Indicators of ART Success

Due to effective ART regimens suppressing HIV viremia to undetectable levels in CSF/plasma, biomarkers of HIV disease such as CSF HIV RNA offer little insight into underlying pathology and neurocognitive disease in virally suppressed individuals. Therefore, biomarkers of underlying neuropathology or cellular activation such as those described above provide critical information regarding the success of novel treatment strategies that complement traditional indicators of HIV viremia.

Intensification of ART to treat CNS infection has been evaluated with some success. Intensive therapy with maraviroc, a CCR5 inhibitor, for 24–52 weeks in combination with standard ART reduced the proportion of CD16+ monocytes harbouring HIV DNA and was associated with better neuropsychological performance in the majority of patients (Ndhlovu et al. 2014; Gates et al. 2016). Early initiation of ART (i.e. <30 days postinfection) also has elicited some neuroprotective effects in both small cross-sectional (Evering et al. 2016) and short-term longitudinal studies (Hellmuth et al. 2016). Furthermore, early initiation also reduced the diversity of HIV DNA in PBMCs, cells isolated from CSF and levels of pro-inflammatory cytokine (IL-6 and TNF) in the CSF, compared to delayed ART initiatives (>14 months post-exposure) (Oliveira et al. 2017). However, the lack of large long-term longitudinal studies limits our understanding of the efficacy of such treatment regimens.

Targeting neuroinflammation using adjunctive therapy in concert with ART also may have beneficial effects. However, to date all attempts to reduce the risk of HAND by targeting neuroinflammation in HIV infection have been unsuccessful (Meulendyke et al. 2014; Sacktor et al. 2011, 2014, 2018; Schifitto et al. 2007). Treatment with the anti-oxidant minocycline reduced CSF levels of some oxidative stress markers in comparison to placebo (Sacktor et al. 2014), but did not improve neurocognitive impairment over 24 weeks (Sacktor et al. 2011). Similarly treatment with selegiline, a MAO-B inhibitor and anti-oxidant, had no neurocognitive or functional benefit in PLWH with mild to moderate neurological impairment (Schifitto et al. 2007). Although treatment with a combination of paroxetine and fluconazole reduced CSF neurofilament light chain and amyloid precursor protein levels in SIV-infected RMs (Meulendyke et al. 2014), a recent double-blind placebo-controlled study using this regimen showed neurocognitive improvement in some, but not all, domains (Sacktor et al. 2018), highlighting that further validation is required. Interestingly, these drugs target HIV gp120 and tat-related neurotoxicity and did not show any improvement in neuroinflammatory markers in plasma or CNS (Meulendyke et al. 2014), suggesting that peripheral inflammation may be more important in human HAND. However, in other studies, drugs specifically targeting extracellular Tat have been shown to counteract its production of IL-1β and TNF in glial cell lines in vitro (Mediouni et al. 2015), and several novel Tat/TAR (and Tat/cyclin T1/CDk9) small molecule inhibitors have been described to date (Serena et al. 2013). These initiatives are important as current therapies do not stop tat secretion from CNS cells (Sonia et al. 2012).

Therapies directed against metabolic dysfunction in the CNS such as intranasal insulin have also been shown to improve neurobehavioural performance in animal models and are currently guiding clinical studies in humans. Specifically, cats who were infected with feline immunodeficiency virus and treated with 20 IU of intranasal insulin for 6 weeks had reduced markers of glial activation and neuroinflammation and improved neurobehavioural performance than placebo controls (Mamik et al. 2016). Furthermore, intranasal insulin treatment of chimeric EcoHIV-infected mice, who exhibit neurocognitive impairment similar to HAND, was found to transiently reverse hippocampal dendritic injury and improve learning and retention ability in some animals; however, this only occurred in the presence of the drug (Kim et al. 2019). These findings support other studies testing intranasal insulin in neurocognitive disorders such as Alzheimer’s disease and have led to Phase I/II clinical trials (NCT03277222, NCT03081117) that are underway testing the effects of intranasal insulin treatment in combination with ART with findings expected in 2020–2021.

Finally, targeting peripheral inflammation is also a possible adjunctive therapy to ART in both HAND and other HIV-associated comorbidities. Targeting monocyte activation using atorvastatin and simvastatin treatment, generally used for treating hyperlipidaemia, reduced the number of circulating CD16+ monocytes (Yadav et al. 2016) and treatment with the dual CCR2 and CCR5 antagonist cenicriviroc; thus migration of CD14+CD16+ monocytes improved neurocognitive performance following 24 weeks of treatment (D’Antoni et al. 2018). Recent in vitro evidence also suggests that treatment with buprenorphine, an opioid derivate used as a replacement for heroin in addicts, may have anti-inflammatory properties by binding monocytes and reducing monocyte transmigration across a BBB model (Jaureguiberry-Bravo et al. 2018). Whilst this treatment may be beneficial in injecting drug using PLWH, its efficacy and safety needs to be confirmed in vivo in non-heroin addicts. Thus, targeting cellular inflammation and immune activation in the periphery may be a viable treatment strategy with beneficial effects across multiple HIV-associated co-morbidities.

8 Where to from Here?

HAND remains a significant burden affecting PLWH globally; however, due to the relatively specialised nature of neurocognitive testing required to diagnose HAND, especially ANI, people living with HIV and neurocognitive impairment are either being misdiagnosed or missed completely. The limited ability to accurately diagnose mild HAND coupled with a change in the HIV epidemic to a chronic inflammatory disease whereby individuals are advancing to older age with an elevated risk of age-related comorbidities represents a difficult problem. Thus, early and accurate diagnoses of HAND are essential in improving health outcomes.

Significant steps forward in understanding the neuropathology of HAND have provided some clues into possible biomarkers for diagnosis and prognosis. Continuous viral presence in the CNS, chronic peripheral immune activation and inflammation as well as underlying drug toxicity and cellular dysfunction all pose viable mechanisms driving neurological diseases in PLWH. However, the discrete effects of each remain unclear. It is likely that a fine balance exists, whereby chronic systemic inflammation, which persists despite viral suppression, may amplify virally induced disease to contribute to severe diseases outcomes such as HAD. This, however, complicates the clinical specificity of many biomarkers that are in effect broad indicators of inflammation such as sCD14 and sCD163. Thus, an added focus on biomarker research delineating underlying CNS pathology such as neuronally derived exosomes, NFL or even MRI of neuroinflammation may be more useful clinically. Furthermore, combining a number of biomarkers to form a ‘risk score’ or ‘index’ using an algorithmic approach may be more predictive than any one biomarker. Whilst significant advances have been made in identifying biomarkers of HAND, the need remains for truly specific, sensitive and reliable biomarkers to reduce the risk of misdiagnosis and delayed treatment and ultimately improve health outcomes in people living with HIV.