Biomarkers of HIV-1-associated neurocognitive disorders: challenges of proteomic approaches

Abstract

HIV-1 enters the brain shortly after infection, which may lead to neurological complications and in the most severe cases to encephalitis, dementia and death. The introduction of antiretroviral therapy reduced the incidence of the most severe conditions, nevertheless, approximately half of those infected with this virus will suffer to various degrees from HIV-1-associated neurocognitive disorders. Despite many years of research, there are no biomarkers that can objectively measure and, more importantly, predict the onset and the tempo of HIV-1-associated neurocognitive disorders. Here we review biomarker candidates of neurocognitive impairment due to HIV infection of the brain that have been proposed during the last two decades, and discuss perspectives and limitations of proteomic approaches in the search for new, more sensitive and specific biomarkers.

Biomarkers are used to reflect a wide range of measures indicating pathological, physiological, or developmental changes in the state of a biological system. A biomarker does not need to be a biomolecule or substance but can also be a specific characteristic, feature, indicator or a change in any biological structure that can objectively measure changes in an ongoing process, such as disease progression and efficacy of treatment [1–3]. The biomarker must also have a predictive power. Therefore, biomarkers are essential and indispensable in monitoring pharmacologic responses to a therapeutic intervention. Biomarkers are also essential in deciding candidates for and methods of treatment. A lack of good biomarkers may not only lead to ineffective treatment but also to the assembly of improper patient cohorts in clinical trials. In consequence, drug candidates may fail and may not achieve US FDA approval [4].

HIV-1 enters the brain at an early stage of infection (Figure 1), remains persistent throughout the disease and can lead to cognitive, motor and behavioral impairments [5]. Classification of disease resulting from HIV-1 infection of the brain is continually evolving [6] and numerous tools have been developed to measure disease progression [5]. Almost 50% of infected individuals develop some form of cognitive impairment (CI), known as HIV-associated neurocognitive disorders (HAND), ranging from very mild CI to encephalitis and severe dementia. The latter stage of disease, HIV-associated dementia (HAD), currently affects less than 7% of infected individuals [7,8] but quickly leads to death. Introduction of highly active antiretroviral therapy (HAART) reduced the severity of cognitive impairments and had a profound effect on slowing disease progression, increasing survival and an overall decrease in the number of new incidences of dementias [7]. Nevertheless, the rate of HIV-1-infected patients with CI remains the same [7] and the prevalence of HAD has increased due to increased survival of these individuals [8–12].

Figure 1. Neuropathogenesis of HIV-1 infection.

Infected macrophages carry the virus into the CNS. A cascade of events leading to the activation of microglia secrete soluble neutoxic factors, such as quinolinic acid, arachidonic acid and its metabolites, glutamate, proinflammatory cytokines, chemokines and others. This leads to neuronal damage and subsequently HAND with its most severe stage of HAD.

HAD: HIV-associated dementia; HAND: HIV-associated neurocognitive disorder.

Diagnosis and treatment of slowly developing diseases such as HAND is currently based on neuropsychological/psychiatric tests. Although these tests are well developed, they are not as precise as objective measures of validated laboratory tests and can be greatly augmented by diagnostic components at the molecular level. The measure of risk is critical to early intervention and prevention because not every HIV-infected individual will develop HAND, and the degree of cognitive impairment can range from asymptomatic to full dementia. Neuropsychological and psychiatric tests cannot assess risk and are unable to indicate the necessity of such early intervention. When first symptoms become measurable by these tests, irreversible damage to the CNS has already begun. Moreover, a concurrent complication is that in many cases the CNS of HIV-infected patients is exposed to more than one damaging factor, for example, opportunistic infection and/or drug abuse. Thus, it is necessary to evaluate how much damage is caused by each of these components. This, in turn, has important implications on effective treatment [13]. Although many potential biomarkers have been proposed, there is lack of good indicators of HAND progression including its most severe form, HAD [14,15], which would be accurate, sensitive and specific.

Profiling of viral proteins as biomarkers of HAND

The viral envelope protein, along with other HIV proteins, has been investigated as a potential biomarker for CI based on its direct toxicity to neurons and the blood–brain barrier (BBB) [16,17]. Based on computational analysis of the gp120C2-V3 loop, Jurado and coworkers postulated that cognitive motor complex may be linked to ‘hot spots’ in the viral genome. The authors also suggested that it may be used as a marker and considered in designing future treatment strategies [18]. Other studies also showed that viral gp120 contributes to neuronal death via various mechanisms [19–22]. Other proteins have also been shown to be toxic for the CNS, but in a more indirect way. For example, increased levels of sCD40L in cerebrospinal fluid (CSF) and plasma correlates with HIV-1-associated CI; however, this study was performed on a relatively small cohort of patients [23]. Nevertheless, the observation was further supported by in vitro studies, which showed that concurrent exposure to viral protein Tat and CD40L has a synergistic effect and induces a neurotoxic phenotype of uninfected monocytes, which is mediated by increased levels of TNF-α induced by Tat [23]. Despite these studies, a correlation between molecular structure of these viral proteins and risk of developing CI has not been proven.

The level of gp120 in patients treated with antiretroviral therapy is very low and often undetectable, yet some of these patients show symptoms of CI. Identification of some viral proteins in the CSF of HIV-infected individuals by tandem mass spectrometry is possible. Nevertheless, in most instances only one unique peptide per protein has been detected. Gp120 protein was represented by only a few peptides, however, the identified peptides were from conserved regions [24]. Therefore, it is unlikely that proteomic profiling of viral proteins directly in ex vivo specimens will provide useful information unless viral proteins are isolated by affinity chromatography. Gene sequencing of viral isolates, which can be amplified in in vitro cultures, seems to be a more effective approach. Although viral proteins may contribute to brain damage, neither the molecular structure of the polypeptide core, nor their levels in circulation can be used as biomarkers of HAND at this time.

Glycosylation of the HIV gp120 envelope protein, although not required, has a profound effect on folding the extracellular domain of the viral protein, thus facilitating proper conformation for the CD4 binding site [25]. Deletion studies in V1V2 regions of gp120 demonstrated that conformational changes in its structure lead to increase susceptibility of the virus to neutralization by antibodies and/or abolish the ability of viral entry into a target cell [26]. Since glycosylation can significantly alter conformation, it can also alter biological properties of gp120, such as immunogenicity and antigenicity [27,28]. On the other hand, some studies investigating whether amino acid features of the V3 loop region can favor viral transmission from individual to individual did not observe such correlation with the number of N-linked glycosylation sites [29]. This suggests that either oligosaccharide moiety and/or the specific N-glycosylation site have a decisive role in altering biological activity of the viral gp120 protein. A systematic glycoproteomic approach is required to advance our knowledge in this area. Despite rapid technological and methodological advancement, glycoproteomic profiling will face a challenge resulting from difficulties in identification of glycosylation sites and deciphering structures of oligosaccharides.

Since HIV is a global problem, much effort was and still is directed towards understanding the geographical profiling of viral strains as well as HIV-infected populations. Based on viral genotyping studies, a correlation between viral strains with a common ancestor (clades) and risk of developing HAND has not been found; however, additional work needs to be carried out in this area [30]. Profiling of monocyte markers among the HIV-positive population has shown some differences between cohorts in Thailand and North America [31]. It has also been postulated that clade C-Tat is relatively less neurotoxic than respective Tat protein produced by clade B viruses, which may have implications in the degree of neurocognitive impairment [32]. Although these studies give valuable insight into the geographical diversity of HIV, clade classification has not yet been proven as a solid biomarker alone and/or in conjunction with other parameters. Moreover, another independent study showed lack of correlation between phenotypic characteristics of HIV isolates and the degree of neurological disorder [33].

Proteomics of HAND

Proteomic approaches in search of biomarkers for HAND must be considered a part of the broader field of neuroproteomics and as such, presents all of the significant challenges associated with such studies of the brain. A major challenge is that the only clinical material in direct contact with the CNS and can be drawn from humans is CSF and it is limited in quantity. All other samples are obtained postmortem and the quality of such samples and how they reflect brain function is the subject of on-going criticism and discussion. Despite the fact that many proteins and nucleic acids are stable postmortem [34], rapid degradation of other proteins may result in changes of overall protein composition [35–37]. Brain hypoxia can change the protein profile rapidly, within minutes after death, making it very difficult to sort out changes related to disease from those related to death itself [38,39]. Animal models address this issue to some extent [40–45]. Despite these limitations, human brains collected postmortem are and will be used for proteomic studies as two experimental approaches emerge. One is to use material dissected from specific brain regions [46]. However, dissection of specific regions of the brain for proteomic investigations poses a significant challenge. Procedures of brain extraction, dissection, preservation, biochemical and molecular integrity of specific regions are all critical aspects of neuroproteomics. The other experimental approach is to purify subsets of cells and perform experiments in an in vitro setting [47]. Both methods address some inherent weaknesses of clinical material used, yet several problems remain unresolved. A variety of techniques for tissue dissection have been developed [48,49]. Manual dissection methods are still very useful, yielding material for microarray analysis with somewhat limited ‘contamination’ of targeted sample with neighboring region(s) that may affect reproducibility of genomic and proteomic analyses. Ultimately, a laser dissection technique will be the method of choice for retrieving small region(s), including a single cell of interest, from brain tissue to improve anatomical accuracy [50]. On the other hand, there might not be enough material to investigate medium- to low-abundant proteins despite constantly improving sensitivity of instrumentation.

Proteomic profiling experiments utilizing cells of the CNS and in vitro systems have begun to emerge and have the potential to generate new and highly valuable information about how cells communicate and support each other and how they may affect phenotype(s) of infected as well as uninfected cells [47,51–53]. Proteomic profiling of pericytes, which are involved in transport across the BBB and regulation of vascular permeability, may reveal their role in neuropathology [54,55] and has yet to be performed. However, we expect that this avenue will be explored more vigorously in the near future using both cells of human origin as well as from various animal models.

Because less than 10% of patients will end up with full-scale encephalitis and dementia and 40% will experience milder forms of HAND, we also expect that changes representing latter states will be characterized by more subtle changes in protein profiles and thus might be at the borderline of statistical significance, even when measured at intervals of a few months. The consequence of this is not only difficulty in assessing prognosis and aiding clinical evaluation, but also difficulty in assessing effectiveness of therapy. Furthermore, an expected problem with this is how to correlate the significance of such changes with the individual response of a patient considering the wide range of drug treatment protocols and modifications made by physicians over a prolonged period of care. Another complication is the possibility of concurrent opportunistic infections, drug and alcohol abuse, and so forth, which may mask changes in metabolic mechanisms resulting from viral infection. Nair and colleagues performed proteomic profiling of mouse dendritic cells treated with methamphetamine (METH) and proposed that METH-induced proteins may be utilized in the development of new biomarkers [56]. The authors found that METH significantly differentially regulates the expression of several proteins including CXCR3, protein disulfide isomerase, procathepsin B, peroxiredoxin and galectin-1. Although dysregulation of these proteins is not unique for either HIV infection and/or exposure to drug abuse and as such cannot be utilized as a diagnostic biomarker, it may provide clues about the function of mononuclear phagocytes under pathological conditions. A separate study from this group demonstrated that cocaine significantly enhances HIV infection of normal human astrocytes and by itself causes differential regulation of 22 proteins, including regulatory proteins (e.g., HP60 and HP40) and enzymes (e.g., enolase, phosphoglycerate kinase and malate dehydrogenasxe), which may also have regulatory properties. For a full list of proteins the author refers readers to his original publication [57].

Among proteomic technologies currently available, so far only two (2-dimensional gel electrophoresis [2DE] and surface enhanced laser desorption/ionization time-of-f light mass spectrometry [SELDI-TOF]) have been used for systematic profiling of samples from HIV-infected people with CI [24,58–60]. One profiling method cannot reveal all differences, therefore more studies need to be carried out utilizing platforms such as isobaric tag for relative and absolute quantitation (iTRAQ^™) and ProteomeLab PF2D^®. The author has recently performed an investigation of sera from HIV-infected and demented patients using a combination of SELDI-TOF profiling followed by weak cation exchange chromatography and 1-dimensional gel electrophoresis and found differentially expressed proteins that were not identified by other methods [61]. Because the same set of samples was used for two concurrent studies we now have a better view of the limitation of 2DE with differential gel electrophoresis technology (2DE DIGE). On the other hand, the latter method revealed that not all forms of complement C3 are equally differentially expressed. It is worth noting that we have not found particularly interesting cysteine histidine-rich (PINCH) protein among differentially expressed proteins in our 2DE proteomic profiling. Considering its 37,251Da molecular weight and pI 8.43, a spot containing this protein should be located in a good separation region of 2DE. Even PINCH migrating as a dimer of approximately 71 kDa, as shown in western blot analysis by Readen et al. [62], should be detected if differentially expressed. This further reinforces the notion that there is a need for a more systematic study of a single, well-defined cohort of samples using three or four proteomic platforms. One limitation of such a study is the amount of clinical material needed. Realistically, there will not be enough CSF from one patient to provide enough protein. This is particularly true following immunodepletion of 12 or more of the most abundant proteins. With this said, plasma/serum, not CSF, will be the source of choice.

CSF as a source of biomarkers of HAND

Cerebrospinal fluid has direct contact with the CNS and as such it may provide valuable information about brain function. It has therefore been extensively studied in the context of various neurodegenerative diseases [63–65]. Experimental approaches in these studies vary from global proteomic profiling to measurements of targeted modification of proteins that are already known to play a role in disease [24,66–72].

Cerebrospinal fluid circulates behind the BBB, which modulates the flow of all molecules including larger molecules such as proteins. The integrity of the BBB is compromised in many neurodegenerative disorders such as Alzheimer’s disease (AD), Parkinson’s disease (PD), amyotrophic lateral sclertosis, multiple sclerosis [73,74] and also during HIV infection [54,75]. This may lead to the presence of plasma proteins in the CSF as well as CSF proteins in blood circulation. The latter issue is discussed below.

In spite of many proteomic studies of CSF there is a limited number of published reports focused on systematic proteomic profiling of CSF in HIV-infected individuals [24,60,76–78]. This might be related to two major limitations in using CSF for proteomic studies. One is an availability of this material, which results from the fact that CSF collection is an invasive procedure, only limited amounts of fluid is permitted to be drawn from a single lumbar puncture, and it is typically difficult to obtain patients’ consent. Another limitation is that the protein content in CSF is much lower (50–200-times) than in serum, while the dynamic range of proteins remains very similar. These limitations are the reasons why relatively highly abundant proteins have been identified in the majority of proteomic analyses of CSF. Other issues related to CSF collection and storage, supported by experimental protocols, are summarized by Hale and coauthors [79]. These authors discussed the problem of potential CSF contamination with blood, which may occur during lumbar puncture. In the author’s 2DE DIGE analyses, an abundant presence of hemoglobin was found in samples that had no visible signs of being contaminated with blood. It is expected that the presence of hemoglobin will also be detected by other methods such as iTRAQ. Therefore, we postulate the use of detection of hemoglobin as an indicator of CSF sample contamination with blood and consider rejection from the pool of investigated samples. Despite these drawbacks, more studies conducted using CSF in studies of neurodegenerative disorders will emerge in the future.

What may we expect to observe in the CSF proteome during HIV infection of the brain? Can there be any specific group(s) or class(es) of proteins that can be a target of more focused proteomic studies using techniques such as multiple reaction monitoring (MRM)? The answer to these questions is very difficult because these targets are elusive and may change along with the introduction of new drugs and treatment regimens [80]. Interestingly, we have not found overwhelming amounts of viral proteins in the CSF of patients with fully developed dementia. In many instances viral load in CSF is very low or even undetectable in individuals being treated with HAART, yet these patients display various degrees of HAND. One can expect changes associated with persistent low-level inflammatory responses unless acute encephalitis develops despite antiviral treatment.

Changes in levels of complement C3 may serve as an indicator of an ongoing inflammation. However, it is unclear how specifically such changes correlate with the development of HAND because such changes have been reported as significant in fluids from patients with various neurodegenerative diseases [81]. Although it might be a sensitive indicator of an ongoing process in the brain, it also appears to be of low specificity. 2DE DIGE profiling showed significant differences between some but not all protein spots in which complement C3 has been identified. Western blot analysis of these samples has not validated these results and showed fewer bands than we might expect from 2DE analysis. For example, consistent changes in levels of α-40 chain of complement C3 were found in CSF samples of HIV-1-infected individuals with dementia. Interestingly, it was not the case when sera samples were analyzed. Although the identity of this fragment was confirmed using mass spectrometry analysis of trypsin-derived peptides, it was not possible to further validate this observation using western blot analysis due to lack of good antibodies. Whether fragments derived from precursor of complement C3 are more specific for any of the neurodegenerative disorders remains to be addressed and will require further targeted profiling studies.

Another of the experimental approaches in the author’s lab was to correlate changes in the secretome of in vitro HIV-1-infected mononuclear phagocytes with changes in the CSF associated with progressive neurodegeneration [82]. These studies provided new information, which will be useful for a better understanding of molecular mechanisms of disease rather than diagnostically useful biomarkers.

Plasma: can it be a source of biomarkers for neurodegenerative disorders?

The most extensively investigated blood/plasma protein biomarkers reflecting neurodegneration are those for AD and PD [83]. This is not surprising because of the aging population of baby boomers for whom life expectancy has been significantly increased over the last few decades. Is it possible to find protein biomarkers for neurodegenerative diseases in plasma when exchange through BBB is limited and regulated? Or will there only be surrogate biomarkers? Because BBB is postulated as being impaired in AD and PD [84,85], one can expect that mechanisms of retention of proteins within the CNS will be breached and that it should be possible to find these proteins circulating in the blood. The same may apply to HIV-1 infection; however, we have to bear in mind that the degree of damage to the BBB may vary and may not correlate with advancement of HAND [47,54]. There are also several other limitations. First, even if the BBB is impaired to a similar extent within a population of patients, how much of the protein pool in the CSF is leaking into the bloodstream? Second, any protein generated in the brain and filtered out to the plasma will be significantly diluted because of the difference in protein concentration of the CSF, which is lower than the plasma by two orders of magnitude and lower volume. Therefore, we may expect that proteins of interest will be at a very low level, most likely below the level of medium abundant proteins. If they are further processed in plasma (e.g., degraded), in complex with other proteins and/or quickly removed from circulation, the actual level at any time point might be so low that analytical methods available now will not be accurate enough. Third, successful treatment of HIV-1-infected patients with HAART may have a protective effect on BBB integrity, further lowering the amount of leaking proteins [86]. Although the latter effect may indicate improvement of BBB, changes can occur at the borderline level of statistical significance. Keeping these limitations in mind, the author’s lab and very few others used proteomic profiling of plasma in the quest for biomarkers of HAND [58,59,61,78]. Four proteins emerged from our studies as plasma/serum biomarker candidates: complement C3, gelsolin, ceruloplasmin and afamin. Our validation was based on western blot analysis performed on an independent cohort of samples; however, small groups of samples were used. A subsequent study using a larger and much better defined cohort of samples is currently ongoing in the author’s laboratory.

Based on this study we conclude that identified biomarker candidates, as well as those that will emerge in the near future, will be surrogates and will serve more as molecular components of a more complex and multicomponent diagnosis rather than as stand-alone biomarkers with high specificity and sensitivity.

Nonproteomic (nonprofiling) approach to biomarker discovery

Nonprofiling approaches in biomarker discovery will remain a vital part of our quest for better laboratory indicators of disease development and effectiveness of therapy. Such approaches usually target specific protein(s) and/or set of parameters and are based on premises of rational design. Very often these premises originate or result from studies that did not have biomarker discovery as a principal objective. One example is to utilize a combination of three parameters: level of the light subunit of neurofilament protein (NF-L) in CSF, viral RNA (viral load) and level of neopterin [87]. The rationale for choosing NF-L was its correlation with axonal injury of neurons while neopterin was as an indicator of immune activation. NF-L release to CSF starts to be detectable in patients who are already diagnosed with neurocognitive disorder at the AIDS dementia complex 0.5 level, therefore it can serve as a confirmatory factor rather than an early indicator. On the other hand, the level of NF-L may serve as a measure of effectiveness of therapy; however, such a hypothesis has not been tested in a specific cohort of patients.

The correlation of viral RNA in CSF with the degree of neuropathogenicity has been a subject of focus in many studies. Although the level of viral RNA indicates lack of control of HIV replication by the immune system, it is not specific enough to be used as a biomarker directly correlated with progression of CI [88–91]. PINCH protein is another example. The function of this 37.25-kDa protein has been investigated in connection with its aberrant expression in post-injury of Schwann cells, neuronal damage and myelin loss [92]. Based on these observations, Rearden and colleagues investigated expression of PINCH in postmortem-collected brain tissue from three groups: controls, HIV-infected patients without CNS alteration and HIV-infected patients with HIV encephalitis (HIVE) [62]. In addition to histopathological studies, the authors also investigated a relatively small number of CSF samples and found PINCH to be upregulated in HIV-positive individuals, but to a significantly lesser extent in those with signs of HIVE [62]. This is a very recent study performed on CSF samples obtained postmortem and there are no data in the literature conferming this observation in a separate cohort of samples obtained from lumbar puncture.

Calpains are a family of calcium-regulated, nonlysosomal, thiol-proteases catalyzing limited proteolysis of a broad range of substrates; they are involved in cytoskeletal remodeling, signal transduction and membrane trafficking. Calpain 15, which is widely and highly expressed in the brain, or its fragments, has been postulated as a useful diagnostic marker [93,94].

Characterization of viral proteins at the molecular level has also been explored as a potential source of biomarkers. Elimination of N-linked glycosylation at position Asp386 of gp120 envelope protein has been shown to enhance viral replication and has been postulated as a factor associated with HAD [95–97]. Nevertheless, correlation of the presence of specific viral isolates present in the brain with neurovirulence remains controversial [33].

Proteomics & experimental models of HIV infection

The human brain is excluded as clinical material until it is recovered postmortem, therefore other experimental models are and will be used despite their inherent limitations. Simian immunodeficiency virus infection of primates most closely reflects the course of retroviral infection in humans including infection of the brain [45]. Although this model has been extensively utilized in many studies there are very few published reports of systematic proteomic investigations of infected monkeys [59,98,99]. An alternative rodent in vivo animal model has been developed and has been used to study pathological processes in the brain for more than a decade [100–104]. Nevertheless, there no systematic proteomic studies have been performed using tissue or cells extracted from HIV-infected rodents. A limitation here is the ratio between infected human cells implanted into mouse brain compared to the total mass of mouse brain. Regardless of the pro-filing platform used, resulting changes in protein expression might be too subtle to be measured with confidence. One approach would be to use MRM-based profiling using a predetermined set of potential markers. Although these experimental systems are to some extent artificial, they can provide valuable information about interactions between cells of the CNS.

Cell-based in vitro experiments remain a very attractive model to study the effects of viral infection owing to relatively low costs, the short time necessary to remove brain after euthanasia and the wide range of reagents available for validation studies. The central point of interest in these studies is how one cell can modulate the phenotype of another cell and how such modulation contributes to overall neurotrophic or neurotoxic environment, which may result in neuronal death thus leading to CI. Proteomics provides an attractive experimental approach to investigate global changes at the cellular level as well at the level of subcellular components. Indeed proteomic investigations of cellular responses in various experimental settings have been conducted in recent years by both the author’s group and others. As a result of our proteomic analysis of the secretome of HIV-infected monocyte-derived macrophages (MDM) [82], we found differentially expressed proteins belonging to various classes such as structural, regulatory, enzymes and redox. The latter class of proteins is of particular interest because it has been shown that disruption of redox balance in the CNS can lead to neuronal death [105,106]. Similar in vitro experiments [57,107] have been performed using dendritic cells differentiated from monocytes, exposed to methamphetamine and infected with HIV-1 IIIB (CXCR4-tropic, X4 strain) [56]. The authors investigated cellular proteins and found filamin 1, tallin1 and coronin, all structural proteins, to be upregulated among other differentially expressed proteins belonging to classes of enzymes, regulatory and redox proteins.

In macrophages (MP), HIV buds into intracellular vesicles and uses exosomes for spreading virus into surrounding environment. Therefore, extracellular vesicles that circulate in the blood and CSF have been of particular interest in the context of HIV infection. Proteomic analysis of highly purified virions produced in vitro in MDM showed the presence of cellular proteins belonging to cytoskeleton, adhesion, signaling, intracellular trafficking, chaperone, metabolic, ubiquitin/proteasomal and immune response systems [108].

Experimental systems are far more complex than those involving cells in co-cultures. Such systems could consist of various types of cells, such as mononuclear phagocytes (MP or microglia) and astrocytes, in which one type of cell is infected [51,109], or one cell type may stably express the antigen of interest. In the latter system uninfected umbilical cord blood mononuclear cells were co-cultured with HEK.293 cells expressing X4Env protein [110].

Although the in vitro proteomic studies described above provide new and valuable information regarding the impact of local environment on the function of CNS cells, the challenge remaining is to make a direct connection between generated results and the development of disease in an in vivo situation; in particular, in the context of diagnostic biomarkers.

Protein microarrays & tissue MALDI profiling

Protein microarray is a very fast-developing branch of proteomics [111–114]. A recent proteomic study conducted by Sartain and colleagues utilized this method of proteomic profiling to investigate sera for discovering biomarkers of Mycobacterium tuberculosis infection in HIV-infected patients [115]. The authors generated 960 unique fractions of mycobacterial cytosolic and secreted proteins that were used to generate microarrays and subsequently used to screen sera. Sera from tuberculosis (TB)-positive and HIV-positive patients recognized altogether 20 fractions out of which 15 were unique compared with samples from TB-positive-only patients [115]. Although this study was not directly aimed to address the issue of HAND, it might be connected based on the hypothesis that even successfully treated opportunistic infections of the brain may still lead to a high risk of encephalitis and dementia [116]. Another study performed in Brazil aiming to characterize neurological disorders in HIV/AIDS patients and their relationship to HAART showed a relatively high percentage of patients with toxoplasmosis (42.3%) and to a lesser extent cryptococcosis meningitis (12.9%) and TB (10.8%) among those who already had or were developing neurological complications [117]. Therefore, efforts focused on biomarker discovery and diagnosis of opportunistic infections will be beneficial for better understanding diseases that result from viral infection [118]. Another array approach was to use immunophenotyping of cluster of differentiation (CD) antigens to more precisely define a specific signature of leukocytes, which can reflect a particular state of disease [119]. Oxidative stress pathways have also been studied in causes of neurodegeneration associated with HIV infection of the brain. Using spectrophotometric techniques, Gil and coworkers tested blood samples from 85 HIV-infected and 40 healthy individuals to measure the array of redox status indices of proteins such as glutathione, malondialdehyde, peroxidation potential, total anti oxidant status, glutathione peroxidase, superoxide dismutase, total hydroperoxide, DNA fragmentation and relative CD4, CD95, CD38/CD8 T lymphocyte counts [120]. It is worth noting that cytokine arrays in HAND have been extensively studied [121–125]; however, it is not a subject of this review.

MALDI-MS tissue profiling is a relatively new technology platform [126–129] that represents an effort to utilize MS in a novel way to enhance clinical diagnosis. One big promise of this technology is that signals for proteins and peptides can be obtained directly from tissue sections [130]. Although much needs to be done, imaging MS (IMS) seems to be a very powerful tool for advancing our capabilities of profiling thin tissue sections with high throughput [131,132]. Another very attractive part of this technique is that paraffin-embedded sections can also be investigated [133]. This means that archived clinical material can be examined retrospectively. So far there are no published reports of IMS investigations of brain sections obtained from HIV-1-infected individuals with neurocognitive impairments.

Surrogate biomarkers & other measures

Surrogate biomarkers are used in therapeutic trials as a substitute for a clinically meaningful end point. They indicate the effect of treatment and/or disease progression and can be used to evaluate (measure) health status of patients enrolled in specific treatment regimens. Although surrogate markers are indirect measures, their importance cannot be underestimated and they are quite broadly used. Surrogate markers in HAND are also well aligned with the hypothesis that neuronal damage is a secondary effect resulting from a cascade of virus–host cell interactions redirecting cell-coded signals and producing toxins.

Surrogate biomarkers include proteins circulating in body fluids such as CSF and/or blood. Nevertheless, there are limitations to how and under which circumstances such markers can be used:

• Their use is justified by the seriousness of disease, very often leading to death;

• Measurements of the effect of clinical end point cannot be detected with more direct markers;

• There is a reasonable clinical benefit.

The US FDA approves use of surrogate biomarkers as a temporary solution to accelerate the process of new drug approval. It needs to be pointed out here that surrogate biomarkers can also be misleading, thus we need to be very careful in interpretation of such correlates. A number of biomarkers that can be classified as surrogates have also been used to evaluate progression and treatment of HIV infection. Table 1 summarizes examples of surrogate markers [134–143].

Table 1.

Surrogate biomarkers of HIV-associated neurocognitive disorders.

Marker	Role in HIV infection of the CNS	Ref.
Neopterin	Interferon-γ stimulates human monocytes and macrophages to produce neopterin. Elevated levels of neopterin in CSF have been found during viral infections including HIV and also during infections by intracellular bacteria and parasites, autoimmune diseases and malignant tumor diseases	[134,135]
Quinolonic acid	HIV-infected macrophage and microglia secrete quinolonic acid and its level in CSF increases with development of HAD, however not specific for diagnostic purposes	[136,137]
3-2 microglobulin	Immune activation marker, elevated in CSF and plasma of HIV-infected patients with neurological disease suggesting that it correlates with neurological complications due to HIV infection	[138]
Prostaglandins	Postulated to play a role in inflammation. Levels of prostaglandins are increased in the CNS of patients with HAD, suggesting their role in the development of neurological dysfunction	[139,140]
MCP-1	MCP-1 is a chemokine that facilitates the recruitment of infected and/or activated monocytes into the brain contributing to CNS inflammation and degenerative changes. HIV Tat protein activates production of MCP-1 by astrocytes	[141,142]
sICAM	sICAM-1 is a circulating form of ICAM-1, a surface glycoprotein, expressed on vascular endothelium, macrophages and activated lymphocytes. ICAM-1 functions as macrophage differentiation antigen. Plays a key role in the process of leukocyte circulation and extravasation from the blood into the areas of inflammation. Elevated in serum and CSF of HIV-infected patients with encephalopathy	[143]

CSF: Cerebral spinal fluid; HAD: HIV-associated dementia; MCP-1: Monocyte chemoattractant protein 1; sICAM: Soluble intercellular adhesion molecule-1.

Certainly, complexity of clinical manifestation and complications induced by HIV infection of the brain makes it much more difficult to sort out how any given biomarker reflects disease progress or just viral infection. It is even harder using indirect (surrogate) biomarkers and their utility lies more in monitoring effectiveness of therapy rather than in risk assessment. CD4 cell counts and viral load have also been used early on to evaluate efficacy of azidothymidine monotherpy [144–149]. Despite the utility of these two biomarkers in the treatment of viral infection, they do not correlate with neurocognitive impairment [150].

Other approaches include neuroimaging methods. Although these methods are not aimed directly at the discovery of proteinaceous biomarkers, in combination with molecular components offered by proteomics, in the future they will provide new information and may lead to the discovery of more direct, sensitive and specific markers of HAND. However, before we will see the first results of such combined studies, bioinformatic methods of comprehensive data analysis across platforms need to be developed. Such work has already been initiated [130,151].

Conclusions & future directions

Biomarkers of HAND cannot be fully separated from biomarkers of other neurodegenerative disorders because many molecular mechanisms leading to neuronal death are very likely shared among these diseases. Increasing incidence of neurodegenerative disorders is already on the rise due to an aging population of baby boomers and an overall increased life expectancy, thus we can observe an increased effort placed on both biomarker discovery and in the number of newly initiated studies. Although HAND is not age related and the incidence of the most severe dementia was substantially reduced after the introduction of HAART, some expect that it will come back in the years ahead due to growing anti-HIV drug resistance. Comparisons of most extreme conditions – such as low-level or asymptomatic neurocognitive impairment with fully developed dementia and encephalitis – increases the chance of discovering meaningful differences in protein profiles. On the other hand, proteomic analyses of CSF and/or plasma from patients with advanced stages of brain inflammation may provide biomarkers that are secondary and are not accurate in reflecting conditions that initiated this disease. Therefore, we must be very careful in data interpretation and drawing conclusions. Another major challenge is that differences between proteome profiles in samples taken from individuals with asymptomatic neurocognition and those whom are mildly impaired are very subtle and will require larger cohorts, more technical replicates and multiple proteomic platforms to support statistical significance. Prospective proteomic study is needed to assess the value of markers discovered in retrospective studies. Thus, the overall cost of these studies will be significantly increased.

Despite the continuous development of proteomic methods and protocols, significant improvement of sensitivity and specificity will be needed to overcome the limited sample quantity available from the CSF. Finally, the importance of proteomic investigations using animal models including primates and nonproteomic approaches will continue to be an indispensable part of the search for biomarkers of HAND as well as other neurodegenerative disorders. The development of bioinformatics to combine data from proteomic studies with those from other fields, such as metabolomics, genomics and imaging will be critical in the advancement of our ability to enhance clinical diagnosis.

Executive summary.

Profiling of viral proteins as biomarkers of HIV-associated neurocognitive disorders

• Although viral proteins such as gp120 and Tat exhibit neurotoxic properties, levels of these proteins in circulation are not correlated with HIV-associated neurocognitive disorders (HAND).

• Genome sequencing of HIV isolates is, at this point, a more effective method of profiling of viral proteins and their variants.

Proteomics of HAND

• Limited availability of cerebrospinal fluid (CSF) is a major obstacle in proteomic profiling for CNS diseases.

• Rapid postmortem degradation of proteins make the quality of such samples questionable.

• The pathological complexity of HIV infection, concurrent opportunistic infections, drug abuse and malnutrition, among other factors, requires a careful approach to proteomic data interpretation.

CSF as a source of biomarkers of HAND

• CSF has been explored as a source of potential biomarkers for neurodegenerative disorders including HAND.

• Decreased levels of complement C3 in CSF and sera/plasma after HIV infection correlates with rebound levels after antiretroviral therapy treatment.

Plasma: can it be a source of biomarkers for neurodegenerative disorders?

• Plasma can be obtained much more easily than CSF and therefore will be preferred as a source of biomarkers.

• There are two limiting factors in using of plasma/serum as a source of biomarkers for neurodegenerative disorders: limited exchange due to the blood–brain barrier and the dilution factor of proteins that are passed from CSF to blood.

Nonproteomic (nonprofiling) approach to biomarker discovery

• The nonproteomic approach to biomarker discovery has advantages and will be considered as an alternative approach.

Proteomics & experimental models of HIV infection

• Animal models of HIV infection will be indispensible in biomarker discovery.

• Simian immunodeficiency virus-infected primates most closely reflects HIV infection in humans, however, such a model is expensive.

• There are some rodent models of HIV infection available, however, their utility to biomarker discovery has not yet been tested.

Protein microarrays & tissue matrix-assisted laser desorption/ionization profiling

• Protein arrays and tissue-matrix-assisted laser desorption/ionization technologies need to be further developed to be more widely utilized.

• Protein arrays are excellent tools for fast screening of large numbers of samples; however, their utility in discovering new biomarkers is limited.

Surrogate biomarkers & other measures

• Surrogate biomarkers will remain an important element for drug discovery and monitoring therapeutic efficacy even if new biomarkers are discovered.