Cerebrospinal fluid (CSF) biomarkers of iron status are associated with CSF viral load, antiretroviral therapy, and demographic factors in HIV-infected adults

HIV affects an estimated 40 million people worldwide [1], and owing to the success of combination antiretroviral therapy (ART) in suppressing viral replication, this population is aging [2‐4]. An increasing number of chronic disease complications, such as HIV-associated neurocognitive disorders (HAND), are therefore the focus of current efforts to improve quality of life for HIV-infected (HIV+) individuals. HAND remains disturbingly common despite ART, for reasons that are only partially understood, occurring in 30–60% of unselected HIV+ persons, even when the virus is undetectable [5, 6]. Intensive research has suggested many possible mechanisms and etiologies for HAND, including persistent low-level immune activation and inflammation within the central nervous system (CNS) [4, 7, 8], comorbid conditions [9], and accelerated aging [10], but few of these mechanisms are directly actionable, and effective interventions remain elusive. Risk factors for HAND include the nadir CD4+ T-cell count, indicating the depth of immunosuppression achieved during the course of HIV disease, older age at seroconversion, anemia, delay in initiating ART, and the presence of HIV RNA in the CSF and plasma, as well as comorbidity [11‐14]. Anemia is prevalent in HIV infection, even when the virus is undetectable during successful ART, and it has been consistently associated with increased morbidity and mortality in HIV infection in retrospective and prospective studies [15]. In addition, the possibility that CNS toxicity due to ART contributes to HAND remains a matter of debate; certain antiretroviral drugs such as zidovudine (ZDV) have well-described effects on iron transport and frequently cause anemia, and others may exhibit neurotoxicity [14, 16, 17]. Female sex appears also to be associated with higher susceptibility to HAND, but women have been underrepresented in prior HIV studies, and factors unrelated to HIV infection may also explain this finding [18‐20]. Finally, HIV+ persons of Hispanic ethnicity experience an increased risk of neurocognitive decline, possibly due in part to reduced or delayed access to health care [13, 18, 21‐23].

Iron homeostasis is essential for normal brain function: iron is required for the synthesis of monoamine neurotransmitters, for mitochondrial function, myelin synthesis, and the catalytic function of a large number of metabolic enzymes [4, 24]. Alterations in brain iron content and distribution (e.g., excess iron deposition) have long been associated with complex neurodegenerative disorders such as Parkinson’s disease, Alzheimer’s disease, and amyotrophic lateral sclerosis (ALS) [25]. Neurocognitive impairment has also been reported in the restless leg syndrome (RLS), in which brain iron deficiency contributes to dopaminergic dysfunction within the brain [26]. It has been challenging to determine the extent to which iron plays a pathogenic as opposed to a bystander role in these conditions, but evidence is accumulating in favor of the former [27‐30]. An extensive literature has established that peripheral iron deficiency also leads to reduced cognitive performance and behavioral disturbances, which in children and adolescents may be long-lasting and not fully reversible upon iron repletion [31‐33]. Despite the known importance of iron to brain health, iron and iron transport proteins in the CSF have been measured only anecdotally in small numbers of HIV-negative persons with specific neurodegenerative conditions, or in healthy controls [34‐38]. CNS iron homeostasis has not been adequately investigated in HIV+ individuals with or without HAND, either by measurement of iron or iron-related biomarkers in CSF, or indirectly, by iron-sensitive neuroimaging [4, 39, 40].

Transferrin, the chief iron-transporter in the circulation, binds and transports ferric iron in a soluble, nonreactive state to metabolically active cells throughout the body, whereupon transferrin-bound iron is internalized by receptor-mediated endocytosis [41]. Transferrin is synthesized and secreted by the choroid plexus and is also produced by oligodendrocytes; it is the chief mode of delivery of iron to neurons [42]. Ferritin, the principal intracellular iron-storage molecule in tissues, is frequently elevated in HIV+ persons, possibly reflecting a prolonged acute-phase inflammatory response [43, 44]. Ferritin comprises varying proportions of heavy (iron-avid) and light (relatively iron-poor) chains (H- and L-ferritin, respectively), depending on the tissue of origin. H-ferritin is the predominant form in the CNS, and it delivers iron to oligodendrocytes via the Tim-2 receptor [45, 46]. Iron delivery by H-ferritin to other cell types has also been suggested [47]. Serum ferritin levels generally rise, and transferrin levels fall, with increasing systemic iron stores; the reverse changes occur in functionally iron-deficient states [41, 43]. Both transferrin and H-ferritin can be influenced by inflammatory stimuli, such as infection or comorbid conditions [48, 49]. The synthesis, transport, and regulation of iron, transferrin, and ferritin within the CNS are incompletely understood and the subject of ongoing investigation [24] In contrast to many other serum proteins, levels of these iron-related proteins in CSF have often been reported not to correlate with those in the peripheral circulation, and the brain is well buffered against systemic iron deficiency or excess [50]. Although a number of studies have examined peripheral (serum or plasma) iron biomarkers in HIV infection as a factor in disease outcomes [51‐54], CSF levels of iron, transferrin, and H-ferritin remain inadequately characterized from an epidemiological perspective [55, 56].

Serum is the most convenient source for routine, noninvasive testing of iron parameters, but the CSF is more likely to provide an accurate reflection of iron transport in the CNS. There is also considerable precedent for the use of CSF biomarkers as surrogate markers of iron metabolism and/or neuronal injury in the brain [57‐59]. Recent studies have documented selective prioritization and partitioning of iron across different tissues in the anemic state, for example, and thus, blood indices may not reflect CNS iron depletion [60, 61]. Another advantage of CSF as the biological medium for estimating CNS iron status is that CSF is sequestered behind both the blood–brain barrier (BBB) and blood–CSF barrier. This isolation from the periphery permits measurement of iron and iron-management proteins at least in part derived from or in equilibrium with the brain parenchyma and specific to the CNS, potentially reflecting unique aspects of iron transport within that compartment.

It is of interest to evaluate associations of known predictors of neurocognitive decline in HIV infection with levels of CSF iron and iron-transport proteins, because such associations would suggest a possible novel role for dysregulation of iron metabolism or iron transport pathways in the pathogenesis of HAND that might be targeted therapeutically. Analogous approaches have proven instructive in achieving a better understanding of other chronic disease phenotypes linked to inflammation, such as cancer [62]. The purpose of this study was therefore to quantify, for the first time, iron and biomarkers of iron transport in the CSF of HIV+ persons and systematically evaluate their relationships to host factors such as age, sex, ethnicity or ancestry, and key HIV disease-related factors like ART (including ZDV use) and the presence and quantity of CSF virus.

This represents the first large study to systematically quantify levels of iron and the two major iron-transport proteins, transferrin and (H)-ferritin, in the CSF of HIV+ persons and to evaluate their associations with known predictors of neurocognitive impairment in this population. Due to the invasive nature of lumbar puncture, published data for CSF iron biomarkers have been scarce even in HIV-negative individuals, in whom data are available only in small numbers of individuals with specific disorders and at different stages of neurodegenerative disease [36, 37, 67‐69]. Hence, normal variations of these biomarkers in healthy persons are also not well documented in the literature [38, 67, 70]. Our findings in over 400 clinically well-characterized HIV+ individuals demonstrate that CSF iron and transferrin are independently associated with demographic factors (race/ethnicity and/or sex) and HIV-disease-specific factors (plasma virus detectability, ART, and/or CSF VL), while CSF H-ferritin is related to comorbidity and virus detectability in plasma. Interestingly, observed differences in CSF iron by self-reported race/ethnicity parallel known differences in the risk of neurocognitive decline in HIV+ adults, with the highest levels occurring in persons of Hispanic ethnicity who also appear to be at increased risk of HAND [13]. This study provides the only available estimates of the range and degree of variability of iron biomarkers in HIV + CSF currently available, and the observed relationships to known risk factors for HAND are hypothesis-generating with regard to the potential role of iron dyshomeostasis in HIV-associated neurocognitive decline.

Both nonhuman primate and human studies have shown that systemic iron deficiency and anemia have lasting impact on brain iron homeostasis and neuronal health, influencing memory and cognition [32, 60, 71‐73]. In Alzheimer’s disease and ALS, both peripheral and CSF iron biomarker profiles suggest disruption of iron homeostasis with concomitant inflammation within the brain [28, 30, 34, 35]. The situation in HIV infection may be similar, with neuroinflammation due to HIV-activated macrophage–monocytes migrating across a damaged BBB triggering a reduction in iron bioavailability in the CNS that may promote neurocognitive decline. Accumulating evidence for systemic changes in iron transport by HIV [54, 74, 75], mediated at least in part by the iron-regulatory hormone hepcidin, and appearance of the virus in the CNS within days of acute infection, suggest that HIV-mediated iron dysregulation may also occur in the CNS. HIV does not infect neurons, but the virus has been shown to infect perivascular macrophages, microglia, and other supporting cells within the CNS, some of which play central roles in regulating brain iron content and distribution [4, 24, 50]. HIV infection of macrophages has also been shown to downregulate expression of the MHC-class I-like iron-regulatory protein Hfe on the macrophage cell surface via the actions of viral proteins, thereby increasing macrophage iron content to benefit viral replication [76].

In the present study, higher CSF iron content tended to be associated with detectable virus in the plasma but with lower CSF VL, when adjusting for the Q_Alb, a measure reflecting variability in integrity of the blood–CSF barrier. This apparent discordance suggests that either (1) more iron fails to cross the CSF–brain barrier when HIV levels in the brain or CSF are low, leading to higher levels in the CSF, or (2) higher amounts of virus in the brain or CSF are associated with greater extraction of iron from the CSF, because HIV requires iron for its replication [77]. The significantly lower CSF transferrin levels among persons of Hispanic ethnicity and non-Hispanic Whites compared to Blacks after adjustment for Q_Alb were also in accord with the corresponding observed CSF iron differences. Even when accounting for the effects of ZDV, an antiretroviral drug that can affect iron transport, ART was independently associated with higher CSF transferrin, raising the possibility that other commonly used antiretroviral drugs may impact transferrin levels in the CNS, possibly by inducing a functionally iron-deficient state. Initiation of ART has been temporally linked to an increase in soluble transferrin receptors, a sensitive indicator of tissue iron deficiency [78, 79]. Other mechanisms by which functional iron deficiency might be induced in the CNS during HIV infection are the depletion of cellular iron due to active viral replication and/or chronic inflammation in the brain that results in iron entrapment within supporting cells that normally deliver iron, or iron deficiency of these cells themselves [24, 76, 80]. Indeed, higher CSF VL was associated in our analyses with higher CSF transferrin levels, with and without Q_Alb adjustment. Interestingly, HIV+ individuals are at increased risk of RLS, a condition associated with iron deficiency in the brain [81, 82].

Although men are known to have significantly higher systemic iron stores than pre-menopausal women, differences in CSF levels of iron, transferrin, or H-ferritin levels have not been documented, and they may have clinical relevance [83]. Women were observed to have significantly lower levels of both transferrin and H-ferritin before Q_Alb adjustment; this association persisted following adjustment only for transferrin, and the direction of the estimate remained the same for H-ferritin although significance was lost. That higher H-ferritin levels were associated with undetectable plasma virus but a higher quantity of CSF virus suggests that induction of H-ferritin in the CNS may be a physiological (cytoprotective) response to viral compartmentalization and replication in the CNS. HIV-induced oxidative stress is manifested by upregulation of heme-oxygenase-1 in the brain [84], and the ferroxidase activity of H-ferritin has recently been shown to mediate the antioxidant activity of heme oxygenase-1 [85]. Due to its high metabolic demands, the brain also requires a steady supply of iron in non-reactive form, for energy production as well as myelin and neurotransmitter synthesis [61, 86‐88]. Todorich et al. [46] previously identified H-ferritin as an important source of iron for oligodendrocytes, and although the origin of CSF iron, H-ferritin, and transferrin cannot be established in this type of study, it is possible that both proteins are upregulated in persons with suppressed viremia in an attempt to preserve iron delivery to oligodendrocytes, other supporting cells of the CNS, and neurons. Indeed, HIV infection is associated with decreased amounts of myelin or loss of myelin integrity [89‐91]. One neuroimaging study that used modern, iron-sensitive techniques reported possible decreases in brain iron in HIV-infected individuals [40], and studies using diffusion tensor imaging or other modalities have consistently revealed white matter damage [40, 83]. It is not surprising that CSF H-ferritin levels were also associated with more clinically significant comorbidity, as comorbidity has emerged as a powerful predictor of neurocognitive decline and is a common correlate of chronic inflammation in HIV+ persons [92]. Mechanisms other than inflammation may also explain increases in H-ferritin in CSF, including HIV-induced changes in autophagy or even ferritinophagy [93, 94]. Finally, associations with age, a strong risk factor for HAND, with both CSF transferrin and H-ferritin were pronounced before adjusting for Q_Alb but lost significance after this adjustment; this begs the question whether the loss of iron regulation (or the loss of compartmentalization of iron) in the CSF due to age- or HIV-related damage to the blood–CSF barrier may in part mediate age-related neurocognitive decline.

Levels of CSF iron biomarkers in this HIV+ population were not dramatically different from levels previously reported in small numbers of healthy individuals, but the different assay methods used complicate any comparisons: CSF ferritin levels in this study were lower than levels in individuals either with or without RLS that were reported by Mizuno et al. and they were also lower than levels observed in chronic demyelinating disorders like chronic progressive multiple sclerosis [37, 38, 67]. Assays for total ferritin, rather than H-ferritin, may also have been used in previous studies. As with other intracellular microorganisms, the manipulation of cellular iron homeostasis by HIV to benefit its own replication may eventually result in altered brain iron balance (either functional iron deficiency, or iron excess in specific brain regions, or a combination of these). Significant changes in total CSF iron biomarker levels may or may not accurately reflect these regional changes. CSF iron, transferrin and ferritin did not appear to be correlated with their respective serum levels in 11 individuals with visit-matched CSF and serum samples, but the small size of the sample with both sets of measurements precludes firm conclusions. While possibly due to insufficient power, the absence of significant correlations supports the concept that levels of iron and iron transporters in the CNS are actively regulated and do not merely reflect passive influx from the circulation; this view is entirely consistent with prior studies [50]. Differences in serum and CSF assays used for iron-biomarker quantification and the fact that H-ferritin was measured in CSF, while the L subunit (iron-poor form) of ferritin predominates in serum, could also account for these differences. Relatively little is known about mechanisms controlling total or H-ferritin levels in either compartment [95].

We acknowledge the limitations of this study, including its cross-sectional design, which limits interpretation, the lack of an HIV-negative comparison group, and absence of matched serum iron biomarker data for most participants. Moderately correlated, within-individual values of iron, transferrin and H-ferritin obtained at baseline and 6 months in 100 of the study participants, however, support the use of single CSF biomarker measurements for our analyses and favor their stability in HIV+ persons over short periods of time. Regarding the lack of HIV-negative controls, we would stress that our objective was not to determine the impact of HIV on CSF iron biomarker levels, but to ascertain their associations with key demographic, HIV disease- and treatment-related predictors of neurocognitive decline, since measures of HAND are often imprecise and fluctuate in individuals over time [4]. Better knowledge of iron biomarker variability and associations with these other host factors will facilitate the design of future studies and inform the precision of predictive models. Longitudinal follow-up studies with measurements of biomarkers at more than one time-point in individual disease trajectories will be needed to clarify the temporal sequence of iron dysregulation after HIV infection, although such studies may be challenging due to the invasive nature of CSF sampling. This study provides a systematic assessment of the demographic and disease-related factors associated with CSF iron biomarker levels in HIV+ persons, laying the foundation for studies that specifically evaluate whether changes in these iron-related measures promote the development of HAND.

In HIV+ adults, CSF levels of iron and major biomarkers of iron-transport are independently associated with sex, self-reported race/ethnicity, ART, plasma virus detectability and/or CSF VL. Associations of CSF iron biomarkers with age, which disappeared following adjustment for a measure of blood-CSF barrier function, suggests the possibility that age- or HIV-related declines in the integrity of the blood-CSF barrier may contribute to changes in CNS iron balance as HIV+ individuals age. Furthermore, this study raises the intriguing possibility that consistent correlates of HAND such as race/ethnicity, ART, and HIV RNA in the CSF, may be linked to HAND in part via changes in CNS iron homeostasis. Additional studies, preferably including longitudinal samples and a higher number of participants over 50 years of age, as well as matched HIV-negative controls, are needed to further investigate these questions.