DIAGNOSTIC AND PROGNOSTIC BIOMARKERS FOR HAND

Abstract

In 2007, the nosology for HIV-1 associated neurocognitive disorders (HAND) was updated to a primarily neurocognitive disorder. However, currently available diagnostic tools lack the sensitivity and specificity needed for an accurate diagnosis for HAND. Scientists and clinicians, therefore, have been on a quest for an innovative biomarker to diagnose (i.e., diagnostic biomarker) and/or predict (i.e., prognostic biomarker) the progression of HAND in the post combination antiretroviral therapy (cART) era. The present review examined the utility and limitations of four proposed biomarkers, including neurofilament light (NFL) chain concentration, amyloid (i.e., sAPPα, sAPPβ, amyloid β) and tau proteins (i.e., total tau, phosphorylated tau), resting-state functional magnetic resonance imaging (fMRI), and prepulse inhibition (PPI). Although significant genotypic differences have been observed in NFL chain concentration, sAPPα, sAPPβ, amyloid β, total tau, phosphorylated tau and resting-state fMRI, inconsistencies and/or assessment limitations (e.g., invasive procedures, lack of disease specificity, cost) challenge their utility as a diagnostic and/or prognostic biomarker for milder forms of neurocognitive impairment (NCI) in the post-cART era. However, critical evaluation of the literature supports the utility of PPI as a powerful diagnostic biomarker with high accuracy (i.e., 86.7–97.1%), sensitivity (i.e., 89.3–100%) and specificity (i.e., 79.5–94.1%). Additionally, the inclusion of multiple CSF and/or plasma markers, rather than a single protein, may provide a more sensitive diagnostic biomarker for HAND, however, a critical need for additional research remains. Most notably, PPI may serve as a prognostic biomarker for milder forms of NCI, evidenced by its ability to predict later NCI in higher-order cognitive domains with regression coefficients (i.e., r) greater than 0.8. Thus, PPI heralds an opportunity for the development of a brief, non-invasive diagnostic and prognostic biomarker for milder forms of NCI in the post-cART era.

INTRODUCTION

Human immunodeficiency virus type 1 (HIV-1) remains a human health pandemic, afflicting approximately 36.7 million individuals worldwide (UNAIDS, 2017), including approximately 4.2 million older individuals (>50 years of age; UNAIDS, 2014). By 2030, the prevalence of older individuals living with HIV-1 is expected to reach approximately 73% (Smit et al., 2015); a shift in the epidemiological features of HIV-1 resulting primarily from two phenomena (Justice, 2010). First, the advent of combination antiretroviral therapy (cART) in 1996 dramatically increased the life expectancy for HIV-1 seropositive individuals (e.g., Romley et al., 2014; Teeraananchai et al., 2017). Second, an increased number of older individuals are becoming infected with HIV-1. As a result, older individuals account for approximately 47% of all HIV-1 seropositive individuals in the United States, as well as 17% of new HIV-1 diagnoses (CDC, 2018).

In 2007, the nosology for HIV-1 associated neurocognitive disorders (HAND) was updated to a primarily neurocognitive disorder (Antinori et al., 2007). The gold-standard for the diagnosis of HAND in the post-cART era is a complete neurocognitive battery assessing at least five cognitive domains (i.e., executive function, speed of information processing, motor skills, etc.; Woods et al., 2009; Elbirt et al., 2015). A complete neurocognitive battery provides an opportunity to classify individuals into three categories (i.e., asymptomatic neurocognitive impairment, mild neurocognitive impairment, and HIV-1 associated dementia) based on their level of impairment (Antinori et al., 2007). In cases where detailed assessments were not readily accessible (Joska et al., 2016), standardized mental status examinations are recommended (e.g., Mini-Mental State Examination, Folstein et al., 1975; HIV Dementia Scale, Power et al., 1995; International HIV Dementia Scale, Sacktor et al., 2005). However, standardized mental status examinations currently lack the sensitivity and specificity needed for an accurate diagnosis of HAND (e.g., Haddow et al., 2013; Sakamoto et al., 2013; Zipursky et al., 2013). Thus, the development of an innovative biological marker (or biomarker) able to accurately diagnose milder forms of neurocognitive impairment (NCI), including HAND, and/or predict neurocognitive decline in aging HIV-1 seropositive individuals has the potential for great clinical significance (Chan & Brew, 2014; Zipursky et al., 2013).

At the broadest level, the term biomarker refers to an objective, measurable and reproducible subcategory of medical signs (Strimbu & Tavel, 2010). More precisely, the National Institutes of Health defines a biological marker as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention” (Biomarkers Definitions Working Group, 2001); a definition that overlaps considerably with the one proposed by the World Health Organization (WHO, 2001). Biomarkers have a broad range of clinical utility, including diagnosing (i.e., diagnostic biomarker), monitoring (i.e., monitoring biomarker), and predicting disease progression (i.e., prognostic biomarker; Biomarkers Definitions Working Group, 2001; FDA-NIH Biomarker Working Group, 2016). The development of biomarkers, from identification to clinical validation and clinical use, however, is challenged by the criteria of both high specificity and high sensitivity. Additional stringent criteria must be employed in the development of a novel prognostic biomarker, including high test-retest reliability, a relationship to functional outcome, and practicality (Light & Swerdlow, 2015).

Scientists and clinicians have been on a quest for an innovative biomarker to diagnose and/or predict the progression of HAND in the post-cART era. Cerebrospinal fluid biomarkers (CSF) and neuroimaging biomarkers have been the focus of many clinical studies. However, neurophysiological biomarkers, including prepulse inhibition (PPI) of the auditory startle response (ASR), have also been proposed in preclinical assessments. PPI is a translational experimental paradigm that affords an opportunity for the assessment of temporal processing (McLaurin et al., 2018a); a deficit which may manifest prior to alterations in higher-order cognitive processes. Two stimuli, including a discrete prestimulus and startling stimulus (Hoffman & Ison, 1980), are presented in the PPI experimental paradigm. When the time interval, or interstimulus interval (ISI), between the two stimuli is between 30–500 msec, robust inhibition is observed (Hoffman & Ison, 1980). Given the well-established neural circuitry underlying PPI, the assessment of PPI may not only reveal deficits in temporal processing, but may also elucidate underlying mechanisms of more complex neurological processes (Hoffman & Ison, 1980). Thus, the present review examines the utility and limitations inherent in proposed biomarkers and the need to critically test the proposed biomarkers via statistical assessments, including discriminant function analyses (DFA) and receiver operator characteristic (ROC) curves.

CEREBROSPINAL FLUID (CSF) BIOMARKERS

Neurofilament Light Chain (NFL) Concentration

Within the central nervous system (CNS), neurofilament (NF) proteins, a class of intermediate filaments, are heteropolymers composed of four subunits, including neurofilament heavy (NFH; 190–210 kDa), neurofilament medium (NFM; 150 kDa), neurofilament light (NFL; 68 kDA) and α-internexin (Yuan et al., 2006; Yuan et al., 2012). NF proteins are a key structural component of neurons, commonly observed in perikarya, dendrites, and most abundantly, axons (Yuan et al., 2012). Emerging evidence also supports the structural and functional role of all four CNS NF subunits in synapses (Yuan et al., 2015; Yuan et al., 2016; Yuan et al., 2018). The four NF subunits form distinctive assemblies within synapses, with an increased abundance of synaptic NF proteins in the postsynaptic area (Yuan et al., 2015). Functionally, NF proteins exhibit subunit-specific modulation of synaptic function, including neurotransmission and synaptic plasticity (Yuan et al., 2016). More specifically, NFL modulates the N-methyl-D-aspartate receptor (NMDAR), glutamate-gated cation channels (Lau and Zukin, 2007), via its interaction with the GluN1 subunit (Yuan et al., 2018); a receptor (e.g., Haughey et al., 2001; Krogh et al., 2015) and neurotransmitter system (e.g., Patton et al., 2000; Wang et al., 2003; Gupta et al., 2010; Melendez et al., 2016) commonly altered by HIV-1 viral proteins.

NFL concentration, an indicator of CNS injury, is measurable in both CSF and blood plasma (e.g., Norgren et al., 2003; Rissin et al., 2010). CSF assessments of NFL concentration require a lumbar puncture, after which concentrations are measured using a sensitive enzyme-linked immunosorbent assay (ELISA; Norgren et al., 2003). Development of an ultrasensitive NFL concentration assay for plasma, and its validation in HIV-1 seropositive individuals (Gisslén et al., 2016), addresses the primary limitation of CSF measurements (i.e., lumbar puncture).

Significant increases in NFL concentration have been observed between HIV-1 seropositive individuals and HIV-1 seronegative controls, independent of measurement methodology (e.g., CSF: Abdulle et al., 2007, Peluso et al., 2013, Peterson et al., 2014; Jessen Krut et al., 2014; Gisslén et al., 2016; plasma: Gisslén et al., 2016; Anderson et al., 2018); differences which are further evaluated by conducting post-hoc analyses in individuals categorically classified based on their level of impairment (e.g.,; ‘primary’ HIV-infected, neuroasymptomatic, HAD; Abdulle et al., 2007; Peterson et al., 2014; Jessen Krut et al., 2014; Gisslén et al., 2016). Most notably, NFL concentration is sensitive to the initiation of antiretroviral therapy, leading to decreased concentrations in most HIV-1 seropositive individuals (i.e., 3 months: 10/21; 1 year: 12/16; Mellgren et al., 2007; 63% of patients; Jessen Krut et al., 2014). However, alterations are not disease specific, and have been observed in other neurodegenerative diseases, including Alzheimer’s disease (AD; e.g., Mattsson et al., 2017; Lewczuk et al., 2018), multiple sclerosis (e.g., Haghighi et al., 2004; Cai and Huang, 2018), and progressive supranuclear palsy (e.g., Donker Kaat et al., 2018).

Presence of a significant relationship between NFL concentration and NCI would support its potential utility as a diagnostic biomarker for milder forms of NCI in the post-cART era. Although a significant relationship between higher plasma NFL and increased NCI was observed in HIV-1 infected adults, the association appears to result primarily from individuals off cART (Anderson et al., 2018). In sharp contrast, an investigation of cART-naïve participants with primary HIV infection failed to observe a significant relationship between CSF NFL chain concentration and NCI (Peluso et al., 2013). Despite significant group differences in NFL concentration, further studies are needed to directly correlate NFL concentration and NCI, and thus systematically evaluate its utility as a diagnostic biomarker.

Although not yet systematically evaluated, CSF NFL chain concentration may also serve as a prognostic biomarker for milder forms of NCI observed in the post-cART era. A retrospective study in HIV-1 seropositive individuals in the pre-cART era revealed elevated CSF NFL concentrations during the two years before presentation of acquired immunodeficiency dementia complex (ADC); an increase rarely observed (i.e., 33%) in HIV-1-infected control subjects (Gisslén et al., 2007). Despite the small sample size (i.e., n=9 HIV-1 individuals with ADC; n=27 HIV-1-infected control subjects), results suggest a biomarker with acceptable sensitivity (i.e., 78%) and specificity (i.e., 67%; Gisslén et al., 2007). Relationships between baseline NFL levels and neuropsychological deterioration observed in other neurodegenerative disorders (e.g., Progressive Supranuclear Palsy: Rojas et al., 2016; Primary Progressive Aphasias: Steinacker et al., 2017; Huntington’s Disease: Byrne et al., 2017) support further longitudinal investigations of NFL proteins in HIV-1 seropositive individuals. However, statistical analyses must account for other comorbid conditions observed in HIV-1 seropositive individuals given the lack of disease specificity inherent in NFL chain concentration measurements.

Amyloid and Tau Proteins

Amyloid precursor protein (APP) is a type 1 transmembrane protein preferentially located in the synapses of neurons (De Strooper & Annaert, 2000). Functionally, APP plays an integral role in synaptogenesis (e.g., Moya et al., 1994; Priller et al., 2006), neuronal outgrowth (e.g., Ikin et al., 2007; Billnitzer et al., 2013) and dendritic spine morphology (e.g., Weyer et al., 2014; Zou et al., 2016). The proteolytic cleavage of APP by α-, β-, and/or γ-secretases leads to the generation of a vast number of different cleavage products (e.g., sAPPα, sAPPβ, amyloid β (Aβ)) via two primary pathways. Specifically, α-secretase cleavage via the non-amyloidogenic pathway releases soluble sAPPα from the cell surface (Chen et al., 2017); a proteolytic product that may be neuroprotective (e.g., Mattson et al., 1993; Goodman and Mattson, 1994). Along the amyloidogenic pathway, β-secretase cleavage generates sAPPβ. The cleavage of APP by β- and γ-secretases forms a peptide approximately 40-residues long commonly known as amyloid β (Aβ; Chen et al., 2017). Accumulating evidence supports important physiological roles for Aβ, including the modulation of synaptic activity (e.g., Kamenetz et al., 2003; Cirrito et al., 2008; Bero et al., 2011). However, excessive amounts of Aβ are deleterious, leading to aggregate formation of Aβ plaques in the extracellular space, a hallmark of AD, triggering synaptic deficits (e.g., Varga et al., 2015; Ulrich et al., 2015) and disruption of intracellular calcium (e.g., Kuchibhotla et al., 2008; Lim et al., 2013).

Tau is a microtubule-associated binding protein located preferentially in axons (Binder et al., 1985) and, albeit in small amounts, in dendrites and dendritic spines (e.g., Zempel et al., 2010; Kopeikina et al., 2013). Alternative mRNA splicing of a single gene leads to the formation of six tau isoforms in the adult human brain (Goedert et al. 1989a; Goedert et al., 1989b; Kosik et al., 1989). Historically, tau proteins have been recognized for their key regulatory role in microtubule stability (Weingarten et al., 1975). More recent evidence, however, supports the role of tau proteins in synaptic function (e.g., Ittner et al., 2010; Hunsberger et al., 2015). The phosphorylation of tau is developmentally regulated, decreasing with age and coinciding with the development of phosphatases (Yu et al., 2009). However, the hyperphosphorylation of tau can lead to the formation of neurofibrillary tangles, commonly observed in AD, subsequently leading to synaptic dysfunction (e.g., Merino-Serrais et al., 2013).

Postmortem assessment of Aβ and tau proteins in the brains of HIV-1 seropositive individuals has revealed distinct neuropathological changes. Neither AIDS individuals in the pre-HAART era nor HAART treated HIV-1 seropositive individuals exhibited significant extracellular Aβ plaques (Gelman & Schuenke, 2004; Anthony et al., 2006). However, intraneuronal accumulation of Aβ has been observed in HIV-1 seropositive individuals, with an increased prevalence in individuals with HIV-1 encephalitis (Achim et al., 2009). Additionally, elevated levels of hyperphosphorylated tau were observed in the hippocampus, evidenced by the presence of neuronal threads, pre-tangles, and tangles, in HIV-1 seropositive individuals relative to age-matched controls (Anthony et al., 2006).

The assessment of sAPPα, sAPPβ, Aβ, total tau (t-tau) and phosphorylated tau (p-tau) in CSF has further elucidated a unique pathology in HIV-1 seropositive individuals; a pathology which is largely dependent upon the categorical classification of individuals based on their level of NCI. Specifically, individuals with primary HIV-1 infection display elevated levels of p-tau and Aβ−42 (Peluso et al., 2013). Neuroasymptomatic HIV-1 seropositive individuals and individuals with HAND exhibited decreased levels of t-tau and p-tau relative to HIV-1 seronegative controls (Clifford et al., 2009). The most severe forms of NCI, including ADC and HAD, are characterized by decreased concentration of sAPPα and sAPPβ, as well as elevated levels of t-tau (Gisslén et al., 2009; Peterson et al., 2014). Relative to HIV-1 seropositive individuals, subjects with AD consistently exhibited increased levels of t-tau (Gisslén et al., 2009; Clifford et al., 2009; Krut et al., 2013; de Almeida et al., 2018), p-tau (Gisslén et al., 2009; Clifford et al., 2009; Krut et al., 2013; de Almeida et al., 2018), and sAPPα (Gisslén et al., 2009; Krut et al., 2013; de Almeida et al., 2018). Thus, both the neuropathology and CSF profile of HIV-1 appear distinct from AD.

CSF measurements are currently the only methodology available to systematically evaluate sAPPα, sAPPβ, Aβ, and p-tau. The need for a lumbar puncture, an invasive procedure necessary for CSF measurements, limits the broad applicability of these markers. To date, a plasma biomarker for either Aβ (Toledo et al., 2013) or p-tau remains elusive, constrained by biological and technical aspects. However, an ultrasensitive blood immunoassay for t-tau has been developed, assessed, and proposed as a diagnostic biomarker for AD (e.g., Chiu et al., 2013; Dage et al., 2016) suggesting its potential diagnostic utility for HAND.

Multiple statistical methodologies have been employed to evaluate the utility of sAPPα, sAPPβ, Aβ, t-tau and p-tau as a diagnostic biomarker for milder forms of NCI in the post-cART era. Correlational assessments suggest a selective diagnostic utility for CSF t-tau biomarkers. Deficits in both prospective and retrospective (i.e., episodic) memory have been implicated in HAND (e.g., Hinkin et al., 1996; Carey et al., 2006), however, a significant relationship was only observed between CSF t-tau and prospective memory in HIV-1 seropositive individuals (Woods et al., 2006). Additionally, the HIV-1 Dementia Scale, which has only limited success in diagnosing HAND (e.g., Haddow et al., 2013; Sakamoto et al., 2013), and the Mosaic Test, exclusively assessing visuoconstructive coordination, were significantly related to CSF t-tau (Steinbrink et al., 2013). A principal components analysis (PCA) utilized five CSF markers (i.e., sAPPα, sAPPβ, Aβ_1–42, t-tau and p-tau) to separate individuals into three primary groups, including a group with HIV-1 seropositive individuals with ADC and CNS opportunistic infections, another group including neuroasymptomatic HIV-1 individuals and HIV-1 seronegative controls, and a final group with AD subjects (Gisslén et al., 2009). Failure to distinguish between neuroasymptomatic HIV-1 individuals and HIV-1 seronegative controls, despite the use of five markers, suggests an overall limited utility for amyloid and/or tau proteins as diagnostic biomarkers for milder forms of NCI.

Systematic evaluation of the utility of amyloid and/or tau proteins as a prognostic biomarker for HAND has not yet been conducted. However, studies evaluating the progression from mild cognitive impairment (MCI) to AD support its potential utility. Specifically, p-tau₂₃₁ accurately predicted decline from mild cognitive impairment (MCI) to AD with a specificity of 80% (Brys et al., 2009). Use of two measurements, including CSF Aβ−42 and p-tau, for the calculation of a ratio predicted conversion from MCI to AD with 88% sensitivity and 90% specificity (Buchhave et al., 2012); a study which also suggests the benefit of including multiple measurements for the development of a biomarker.

NEUROIMAGING BIOMARKERS

Resting-State Functional Magnetic Resonance Imaging (fMRI)

Functional magnetic resonance imaging, or fMRI, was developed in 1990 (Ogawa et al., 1990) as an in vivo method to assess brain activity (Casey et al., 2002). Conventionally, fMRI experiments utilize Blood Oxygenation Level Dependent (BOLD) contrast, which assesses changes in oxygenation concentration as a measure of neural activity (e.g., Ogawa & Lee, 1990; Ogawa et al., 1990). The BOLD contrast results from magnetic differences in oxygenated and deoxygenated blood, whereby deoxygenated blood is highly paramagnetic (Thulborn et al., 1982). Thus, the activation of a specific brain region leads to increases in blood oxygenation, paralleling localized increases in blood flow, and decreases in deoxygenated blood; an effect which leads to larger magnetic resonance signals (Casey et al., 2002; Glover, 2011). Changes in magnetic resonance signals reflect neuronal mass activity, a primary limitation of the technology preventing differentiation between excitation and inhibition, as well as function-specific processing and neuromodulation (Logothetis, 2008).

fMRI assessments are a non-invasive procedure performed on a magnetic resonance (MR) imaging scanner. Specifically, for the assessment of brain activity in HIV-1 seropositive individuals, a resting-state analysis is commonly employed (e.g., Thomas et al., 2013; Guha et al., 2016; Ann et al., 2016; Chaganti et al., 2017). Use of resting-state fMRI allows for the examination of spontaneous BOLD fluctuations, an intrinsic property of the brain (Fox and Raichle, 2007). During resting-state fMRI scans, participants are typically instructed where to focus (eyes closed, focus on a point on the screen) and not to think about anything. In sharp contrast, task activation fMRI studies (not reviewed presently) may also be employed, manipulating stimuli (e.g., visual, auditory) to induce different neural states in the brain. Two measures that are commonly examined in fMRI include regional homogeneity, an assessment of functional coherence (Zang et al., 2004) and functional connectivity, providing insight into neuronal connections.

Resting-state fMRI studies have revealed significant differences in regional homogeneity and functional connectivity between HIV-1 seropositive individuals and controls at several ages. Adolescents perinatally infected with HIV-1 display increased regional homogeneity in brain regions associated with the central somatic motor-sensory cortex and decreased regional homogeneity in brain regions associated with the corticostriatal pathway (Wang P. et al., 2018). During primary HIV-1 infection (i.e., on average, less than 1 year), HIV-1 seropositive individuals displayed decreased connectivity strength within the lateral occipital network (Wang et al., 2011), a robust network associated with “cognition space” (Smith et al., 2009). In HIV-1 seropositive adults, decreased corticostriatal functional connectivity (Ortega et al., 2015), abnormalities in cerebro-cerebellar functional connectivity (Wang H. et al., 2018) and decreased synchronicity in salience (i.e., anterior insula and dorsal anterior cingulate cortex) and executive networks (i.e., dorsolateral prefrontal cortex; Changanti et al., 2017) have been observed relative to healthy controls. Studies have also conducted resting-state fMRIs in HIV-1 seropositive individuals with and without HAND, revealing decreased functional connectivity between the precuneus and prefrontal cortex (PFC) in HIV-1 seropositive individuals with HAND relative to individuals without HAND (Ann et al., 2016). Comparisons of HIV-1 seropositive individuals without HAND and healthy controls revealed increased efficiency in the posterior cingulate cortex (Ventura et al., 2018) and decreased functional connectivity in frontal areas, temporal areas and the cerebellum (Ann et al., 2016). In sharp contrast, other studies fail to observe significant genotypic differences in functional connectivity (Guha et al., 2016; Janssen et al., 2017; Abidin et al., 2018; Cole et al., 2018).

fMRI assessments have also been fundamental in assessing the independent and/or synergistic effects of HIV-1 and aging. Across studies, researchers failed to observe a significant HIV-1 x age interaction, after correcting for multiple comparisons, suggesting that HIV-1 and aging are independent processes (Ances et al., 2010; Thomas et al., 2013; Ebgert et al., 2018; Egbert et al., 2019). However, studies were conducted primarily in male HIV-1 seropositive individuals and controls and may not generalize to HIV-1 seropositive females. An extension of existing graph theory approaches to model functional connectivity, however, suggested that HIV and age were associated with divergent measures (i.e., HIV-1: centrality, an assessment of node importance; Age: diversity, an assessment of network disorder; Thomas et al., 2015). The development of new analytic approaches for resting-state fMRI, therefore, may more precisely capture the relationship between HIV-1 and aging.

Statistical techniques have been employed to examine the diagnostic utility of fMRI by examining the relationship between imaging metrics and neurocognitive performance. During primary HIV-1 infection, a relationship between the connectivity strength in the lateral occipital network and visual-motor coordination was observed (Wang et al., 2011). Connectivity alterations in brain regions within the PFC (e.g., posterior cingulate cortex, anterior cingulate cortex, medial prefrontal cortex) were associated with NCI in perinatally infected HIV-1 seropositive youth (Herting et al., 2015) and HIV-1 infected adults (de Plessis et al., 2017; Changanti et al., 2017; Ventura et al., 2018). NCI in HIV-1 seropositive individuals was also associated with functional connectivity in other brain regions and networks, including frontal areas (Ann et al., 2016), the basal ganglia (de Plessis et al., 2017), and somatosensory network (Changanti et al., 2017). Additionally, in HIV-1 seropositive adults, modularity and small-worldness, two measures representing components of the integrative connectivity of the brain, are both significantly associated with executive functioning, as well as overall NCI; a relationship dependent upon the methodology (i.e., non-linear mutual connectivity vs. correlation derived measures) employed to construct brain connectivity profiles (Abidin et al., 2018). Use of a ROC curve suggests the diagnostic utility of fMRI, distinguishing between HIV-1 seropositive individuals with HAND and controls with area under the curve measures greater than 0.8; discrimination which was enhanced with the use of a nonlinear approach (Dsouza et al., 2017). Other studies, however, have failed to find a significant relationship between resting-state fMRI measurements and degree of NCI (Thomas et al., 2013; Ortega et al., 2015; Guha et al., 2016; Wang P et al., 2018). Inconsistencies between studies may be due to a variety of experimental factors (e.g., differences in the time of infection, treatment with cART, neurocognitive assessments, and fMRI analysis methods), however, they likely reduce the utility of resting-state fMRI as a viable diagnostic biomarker for HAND.

To date, no study has systematically evaluated the utility of resting-state fMRI as a prognostic biomarker for milder forms of NCI observed in the post-cART era. However, assessments of HIV-1 seropositive individuals at baseline and after a two year follow-up, revealed no differences in the rates of change in resting-state fMRI relative to controls (Cole et al., 2018). Although further studies are needed to directly correlate neurocognitive progression and changes in resting-state fMRI, at this point, results suggest its utility may be limited.

NEUROPSYCHOLOGICAL BIOMARKERS

Prepulse Inhibition (PPI)

In a series of seminal papers (e.g., Hoffman & Searle, 1965, Ison & Hammond, 1971), Hoffman and Ison introduced and popularized the use of PPI of the ASR for the assessment of temporal processing; a construct analogous to speed of information processing in humans. Cross-modal PPI and gap prepulse inhibition (gap-PPI) are translational experimental paradigms that rely on the presentation of two stimuli: a discrete prestimulus or prepulse and a startle stimulus (Hoffman & Ison, 1980). Notably, cross-modal PPI relies upon the presentation of an added prestimulus (i.e., Visual: light, Auditory: tone, Tactile: air puff), while the salient prestimulus is removed in gap-PPI (i.e., gap in background light, noise, or air). Dramatic decreases in ASR are observed when the time interval, or interstimulus interval (ISI), between the discrete prestimulus and startle stimulus is 30–500 msec; an effect that is independent of sensory modality (Pickney, 1976; Hoffman & Ison, 1980; Campeau & Davis, 1995). Systematically manipulating the ISI (e.g., 0, 30, 50, 100, 200, 4000 msec) establishes the shape of the ISI function (e.g., a sharpening or flattening of the response amplitude curve, a broadening or narrowing of the response amplitude curve, shifts in the point of maximal inhibition), affording a critical opportunity to evaluate temporal processing (McLaurin et al., 2018a).

Prepulse inhibition is mediated through the cortico-striato-pallido-pontine (CSPP) circuit, which includes both sequential and parallel neural connections (Koch & Schnitzler, 1997). Broadly, sensory input is relayed to either the superior colliculus (auditory) or inferior colliculus (visual, tactile) and subsequently transmitted to the pedunculopontine tegmental nucleus (PPTg), triggering a cholinergic projection to the caudal pontine reticular nucleus eliciting a startle response (Fendt et al., 1994; Fendt et al., 2001; Koch et al., 1993; Koch & Schnitzler, 1997). Alterations in neurotransmitter system function, including the dopaminergic system, which is implicated in the pathogenesis of HAND (e.g., Kumar et al., 2009; Kumar et al., 2011), play a prominent role in the regulation of PPI. Specifically, the nucleus accumbens (NAc) receives dopaminergic afferents from the ventral tegmental area (VTA), which subsequently relays information to the PPTg; activity which may be mediated by glutamatergic projections from the medial prefrontal cortex (mPFC) to the VTA (Taber & Fibiger, 1995). Activation of the dopaminergic system via treatment with dopamine (DA) agonists leads to marked decreases in PPI (e.g., Mansbach et al., 1998; Zhang et al., 2000; Fitting et al., 2006;); an effect which is also observed following blockade of DA receptors in the mPFC, either via a large 6-OHDA lesion (Bubser et al., 1994) or local injections of dopamine antagonists (Ellenbroek et al., 1996). Alterations in PPI, therefore, may not only provide critical information on neurocognitive dysfunction, but may also elucidate underlying neural mechanisms in HAND.

The eyeblink reflex is one component of the startle response, providing a clinically relevant experimental paradigm for the assessment of PPI in humans. Although multiple methods are available for the assessment of eyeblink reflexes, the preferred method for clinical startle blink research utilizes surface electromyographic (EMG) recording electrodes (Blumenthal et al., 2005). EMG recording electrodes measure action potentials generated within the orbicularis oculi muscle (Blumenthal et al., 2005). Relative to other methodologies (e.g., magnetic search coils, vertical electrooculographic), EMG recording electrodes require less expensive equipment, have less obtrusive sensors, and exhibit increased sensitivity to blink activity (Flaten, 1993; Blumenthal et al., 2005). Most notably, the eyeblink startle paradigm displays characteristics that are critical for the development of a clinically relevant biomarker (Myers & Brown, 2006), including brevity (i.e., approximately 15–25 min; Fournier & Hebert, 2013; Minassian et al., 2013), test-retest reliability (Braff et al., 1978; Schwarzkopf et al., 1993) and ease of administration.

The assessment of cross-modal PPI and gap-PPI in clinical and preclinical studies has repeatedly revealed significant genotypic differences. HIV-1 seropositive individuals with HAND exhibited impairments in PPI relative to HIV-1 seropositive individuals without NCI; an impairment which was associated with alterations in higher-order cognitive processes, including attention/working memory (Minassian et al., 2013). Significant deficits in temporal processing were also observed across multiple animals systems utilized to model HAND, including the HIV-1 transgenic (Tg) rat (e.g., Moran et al., 2013; McLaurin et al., 2017a; McLaurin et al., 2017b), stereotaxic injections of HIV-1 viral proteins (e.g., Tat: Fitting et al., 2006a; Fitting et al., 2006b; gp120: Fitting et al., 2006c; Fitting et al., 2007), gp120 transgenic mice (Henry et al., 2014; Bachis et al., 2016) and Tat transgenic mice (Paris et al., 2015). Additionally, longitudinal studies in the HIV-1 Tg rat have revealed significant alterations in the progression of temporal processing in both cross-modal PPI (McLaurin et al., 2018b) and gap-PPI (McLaurin et al., 2016); an alteration that was more pronounced in female HIV-1 Tg rats (McLaurin et al., 2016; McLaurin et al., 2018b). Although alterations in PPI are not unique to HAND, having been observed in both schizophrenia (Braff et al., 1978) and attention deficit hyperactivity disorder (Castellanos et al., 1996), temporal processing deficits may demonstrate specificity relative to other neurodegenerative diseases (e.g., AD: Hejl et al., 2004; Ueki et al., 2006)

DFA and ROC curves have been conducted to assess the diagnostic utility of cross-modal PPI and gap-PPI. DFA and ROC curves are complementary statistical techniques that evaluate the diagnostic accuracy, sensitivity and specificity (Zweig & Campbell, 1993). In the HIV-1 Tg rat, cross-modal PPI and/or gap-PPI accurately diagnoses the presence of the HIV-1 transgene with both high sensitivity and high specificity across multiple ages (i.e., pre-weanling: McLaurin et al., 2017c (Figure 1A [DFA: 97.1% accuracy, 100% sensitivity, 93.8% specificity; ROC: Area 1.0]); adolescent/young adult animals: McLaurin et al., 2016 (Figure 1B [DFA: 90.8% accuracy, 100% sensitivity, 79.5% specificity; ROC: Area 0.986]); adult animals: (Figure 1C [DFA: 86.7% accuracy, 89.3% sensitivity, 84.4% specificity; ROC: Area 0.951]); older adult animals: (Figure 1D [DFA: 96.8% accuracy, 100% sensitivity, 94.1% specificity; ROC: Area 0.995])) and sensory modalities (McLaurin et al., 2017b). Thus, the diagnostic utility of temporal processing deficits, assessed using PPI, generalizes across two experimental paradigms, the functional lifespan, sensory modalities, and biological sex in the HIV-1 Tg rat, supporting a promising diagnostic biomarker for milder forms of NCI in the post-cART era.

FIGURE 1.

The diagnostic utility of measures of prepulse inhibition (PPI) and gap prepulse inhibition (gap-PPI) is illustrated as a function of genotype (i.e., HIV-1 transgenic (Tg) vs Control) using both discriminant function analyses (DFA) and receiver-operating curves (ROC; inset). In the HIV-1 Tg rat, cross-modal PPI and/or gap-PPI accurately diagnoses the presence of the HIV-1 transgene with high accuracy, high sensitivity and high specificity across multiple ages. ROC analyses were conducted based on the discriminant scores. ROC area under the curve (AUC) is utilized as a single measure of overall accuracy for the ROC. A: In pre-weanling animals (i.e., Postnatal Day 14–21), a DFA selected 6 PPI variables, exhibiting 97.1% accuracy, 100% sensitivity, and 93.8% specificity. An AUC of 1.0 was observed when an ROC analysis was conducted. Adapted from McLaurin et al., 2017c. B: In adolescent/young adult animals (i.e., PD 30-PD 180), 4 PPI variables best discriminated between HIV-1 Tg and control animals in a DFA, displaying 90.8% accuracy, 100% sensitivity, and 79.5% specificity. An AUC of 0.986 was observed when a ROC analysis was conducted. Adapted from McLaurin et al., 2016. C: In adult animals (i.e., PD 240-PD 480), a DFA selected 3 PPI variables, exhibiting 86.7% accuracy, 89.3% sensitivity, and 84.4% specificity. An AUC of 0.951 was observed when an ROC analysis was conducted. Reanalysis of data presented in McLaurin et al., 2018b. D: In older adult animals (i.e., PD 540), a 5 PPI variables best discriminated between HIV-1 Tg and control animals in a DFA, exhibiting 96.8% accuracy, 100% sensitivity and 94.1% specificity. An AUC of 0.995 was observed when an ROC analysis was conducted. Reanalysis of data presented in McLaurin et al. 2018d.

The prognostic utility of cross-modal PPI to predict disease progression in the HIV-1 Tg rat was examined using a longitudinal experimental design across the functional lifespan (Postnatal Day (PD) 30 to PD 600) revealing sex-dependent expression of NCI (McLaurin et al., 2018c). For HIV-1 Tg males, regressing auditory PPI at PD 30 and PD 60 on the number of errors in 18 month acquisition (i.e., a challenging signal detection task tapping sustained attention) revealed a regression coefficient (r) of 0.81. Therefore, in male HIV-1 Tg animals, auditory PPI explained 65.8% of the variance in the number of errors in 18 months acquisition (Figure 2A). For HIV-1 Tg females, regressing auditory PPI at PD 30 and PD 60 on the number of errors in an operant reversal task at 18 months of age, tapping flexibility and attention, exhibited a regression coefficient (r) of 0.83. Thus, in female HIV-1 Tg animals, auditory PPI explained 68.9% of the variance in the number of errors in a reversal task, tapping flexibility and inhibition, at 18 months of age (Figure 2B). Early alterations in temporal processing, assessed using PPI, therefore, are predictive of later NCI in executive functions.

FIGURE 2.

The utility of early (i.e., Postnatal Day (PD) 30 and PD 60)) prepulse inhibition (PPI) assessments to accurately predict later (i.e., PD 540) neurocognitive impairment in higher order cognitive processes (i.e., sustained attention, flexibility, and inhibition) was assessed in HIV-1 Tg animals. Regression assessments were conducted independently by sex due to sex-dependent expression of the HIV-1 transgene at an advanced age. Specifically, male HIV-1 Tg animals exhibited greater impairments in sustained attention (i.e., 18 Month Acquisition), while female HIV-1 Tg animals displayed greater impairments in flexibility and inhibition (i.e., Reversal; McLaurin et al., 2018c). A trimmed means regression analysis was conducted. A: For HIV-1 Tg males, regressing auditory PPI at PD 30 and PD 60 on the number of errors in 18 month acquisition (i.e., a challenging signal detection task tapping sustained attention) revealed a regression coefficient (r) of 0.81. Therefore, in male HIV-1 Tg animals, auditory PPI explained 65.8% of the variance in the number of errors in 18 months acquisition. B: For HIV-1 Tg females, regressing auditory PPI at PD 30 and PD 60 on the number of errors in an operant reversal task at 18 months of age, tapping flexibility and attention, exhibited a regression coefficient (r) of 0.83. Thus, in female HIV-1 Tg animals, auditory PPI explained 68.9% of the variance in the number of errors in a reversal task, tapping flexibility and inhibition, at 18 months of age. Reanalysis of data presented in McLaurin et al., 2018c. C: Regression Analysis.

CONCLUSIONS

Scientists and clinicians have been on a quest for an innovative biomarker to diagnose and/or predict the progression of HAND in the post-cART era. The present review examined the utility and limitations of four proposed biomarkers, including NFL chain concentration, amyloid and tau proteins, resting-state fMRI, and PPI. Although significant genotypic differences are observed in NFL chain concentration, amyloid and tau proteins, and resting-state fMRI, inconsistencies and/or limited research challenges their utility as a diagnostic and/or prognostic biomarker for milder forms of NCI in the post-cART era (reviewed in Table 1). Evaluation of the literature supports PPI as a powerful diagnostic biomarker for HAND (i.e., 86.7–97.1% accuracy, 89.3–100% sensitivity, 79.5–94.1% specificity), diagnosing presence of the HIV-1 transgene with high sensitivity and high specificity. The inclusion of multiple CSF and/or plasma markers, rather than a single protein, may provide a more accurate diagnostic biomarker for HAND, however, a critical need for additional research remains. Regarding a prognostic biomarker for milder forms of NCI, PPI notably predicts later NCI in higher-order cognitive domains with regression coefficients (i.e., r) greater than 0.8.

TABLE 1.

Comparison of Proposed Biomarkers

Biomarker	Advantage & Utility	Disadvantage & Constraints	GOAL: Diagnostic Biomarker	GOAL: Prognostic Biomarker
NFL Chain Concentration	Plasma and CSF concentrations highly correlated; Sensitive to the initiation of cART	Alterations have been observed in other neurodegenerative diseases (e.g., AD, multiple sclerosis), and thus, it is not disease specific	Inconsistent relationship with NCI that may be driven primarily by cART	In the pre-cART era, increases in NFL concentration were associated with presentation of ADC
Amyloid and Tau Proteins	Distinct pathology from AD	Invasive procedure needed for CSF measurements	Selective relationships between NCI and protein markers; Unable to distinguish between milder forms of NCI and controls, despite the use of multiple markers	CSF Aβ−42 and p-tau have been utilized as a prognostic biomarker for the conversion from MCI to AD
Resting-State fMRI	Brief, non-invasive procedure	Currently unclear what changes in BOLD signaling indicate mechanistically; Inconsistencies between studies	Inconsistent relationship between fMRI measures and NCI	No changes in resting-state fMRI relative to controls were observed in a two-year follow-up
Prepulse Inhibition	Brief, non-invasive procedure; Clinically translational experimental paradigm	Alterations have been observed in other diseases, albeit evidence in other neurodegenerative disease (e.g., AD) is limited	Diagnostic utility generalizes across two experimental paradigms (i.e., cross-modal PPI, gap-PPI), the functional lifespan, sensory modalities, and biological sex; Overall high accuracy, sensitivity, and specificity	Early neurocognitive impairments in PPI predict later alterations in higher order cognitive functions (r≥0.81)

NFL=Neurofilament Light; CSF=Cerebrospinal Fluid; cART=Combination Antiretroviral Therapy; AD=Alzheimer’s disease; NCI=Neurocognitive Impairments; ADC=AIDS Dementia Complex; Aβ−42=Amyloid β 1–42; MCI=Milder Cognitive Impairment; fMRI=Functional Magnetic Resonance Imaging; BOLD=Blood Oxygenation Level Dependent; PPI=Prepulse Inhibition; gap-PPI=Gap Prepulse Inhibition

All four proposed biomarkers examined in the present review are, at some level, related to synaptic dysfunction, functional connectivity, and/or neurotransmitter alterations; neural mechanisms which may underlie HAND. Specifically, synaptic dysfunction in HIV-1 is characterized by structural and chemical alterations in synapses, which have been observed in post-mortem HIV-1 seropositive individuals (Ellis et al., 2007; Gelman & Nguyen, 2010; Gelman et al., 2012). In the HIV-1 Tg rat, profound alterations in dendritic branching connectivity, neuronal morphology, and, to a lesser extent, dendritic spine morphology have been observed in medium spiny neurons of the NAc (Roscoe et al., 2014; McLaurin et al., 2018d), in pyramidal neurons of layers II-III of the mPFC (McLaurin et al., 2018c) and in pyramidal neurons of layers II-III of the mPFC following psychostimulant exposure (McLaurin et al., 2018e). Additionally, dopaminergic system dysfunction has been associated with NCI, evidenced by decreases in DA in the substantia nigra (Kumar et al., 2011), decreased CSF levels of homovanilic acid (Di Rocco et al., 2000) and decreased levels of dopamine transporter (Chang et al., 2008) in HIV-1 seropositive individuals. Thus, not only may these measurements (i.e., NFL chain concentration, amyloid and tau proteins, resting-state fMRI, and PPI) serve as potential biomarkers, but they may also enhance our understanding of the neural mechanisms underlying HAND.

Investigations for an innovative biomarker for HAND may be enhanced by addressing gaps in some of the current studies. First, many studies assessed a primarily, if not exclusively, male sample, limiting the generalizability of assessments. HIV-1 seropositive women are currently inadequately represented in both clinical and preclinical studies (Maki & Martin-Thormeyer, 2009), despite reports of profound sex differences in NCI (e.g., Royal et al., 2016; Maki et al., 2018; McLaurin et al., 2016; McLaurin et al., 2017d; Rowson et al., 2016; McLaurin et al., 2018c). Second, the arbitrary classification of HIV-1 seropositive individuals into three groups based on neurocognitive performance (i.e., ANI, MND, and HAD) naturally decreases variability, decreases power, and results in the loss of individual differences (Rucker et al., 2015). Utilizing the continuous neurocognitive performance scores ought to provide a stronger foundation for assessing the diagnostic and/or prognostic utility of a biomarker.

Use of a multiple biomarker approach may be one of the most critical gaps in the current studies. Despite reporting measurements for multiple CSF proteins or several fMRI metrics, diagnostic and/or prognostic utility was primarily evaluated using a single variable. One notable exception was presented in Gisslén et al. (2009), whereby a PCA was used to assess the ability of five CSF protein markers (i.e., sAPPα, sAPPβ, Aβ_1–42, t-tau, p-tau) to separate individuals into arbitrarily classified groups. Although the PCA failed to distinguish between neuroasymptomatic HIV-1 individuals and HIV-1 seronegative controls, an important separation in the post-cART era, further assessments of multiple biomarkers are necessary. Inclusion of multiple variables may increase sensitivity and specificity, thereby increasing the likelihood of discovering an innovative and accurate diagnostic biomarker. The benefits of a multiple biomarker approach, however, extend beyond the development of an innovative biomarker for HAND in the post-cART era. For example, the pathogenesis of HAND is multidimensional, and may include neurotransmitter alterations (e.g., clinical: Kumar et al., 2011; HIV-1 Tg rat: Sinharay et al., 2017; Javadi-Paydar et al., 2017), synaptic dysfunction (e.g., clinical: Gelman et al., 2012; Desplats et al., 2013; HIV-1 Tg rat: Roscoe et al., 2014; McLaurin et al., 2018e), and neuroinflammation (e.g., clinical: Meier et al., 2009, Royal et al., 2016; HIV-1 Tg rat: Royal et al., 2012). Employing a DFA or multiple regression statistical approach on variables tapping multiple potential neural mechanisms may indicate both their utility as a diagnostic biomarker and their involvement (i.e., via variance accounted for) in the underlying pathogenesis of HAND.

Our review of potential biomarkers for milder forms of NCI in the post-cART era is not without limitation. First, the present review does not examine all biomarkers (e.g., cytokines, Yuan et al., 2013; structural magnetic resonance imaging, Steinbrink et al., 2013) that have been proposed for milder forms of NCI. Second, the scope of the present review could not cover all studies pertaining to the four biomarkers assessed. The integrative assessment of the diagnostic and prognostic utility of four biomarkers, however, may help identify key gaps in our knowledge, providing a foundation for further studies exploring these and other proposed biomarkers.

The development of an innovative biological marker (or biomarker) able to accurately diagnose milder forms of NCI, including HAND, and/or predict neurocognitive decline in aging HIV-1 seropositive individuals has the potential for great clinical significance (Chan & Brew, 2014; Zipursky et al., 2013). Analyses support PPI as a powerful diagnostic biomarker and promising prognostic biomarker. Thus, PPI heralds an opportunity for the development of a brief, non-invasive diagnostic and prognostic biomarker for milder forms of NCI in the post-cART era.