Metalloproteinases and Brain Arterial Remodeling Among Individuals With and Those Without HIV Infection

Abstract

Background. This study tests the hypothesis that increased elastolytic activity is associated differentially with dolichoectasia in individuals with and those without human immunodeficiency virus (HIV) infection.

Methods. Large arteries from 84 autopsied brains from HIV-positive individuals and 78 autopsied brains from HIV-negative individuals were stained for metalloproteinase 2 (MMP-2), MMP-3, MMP-9, tissue inhibitor of metalloproteinases 1 (TIMP-1), TIMP-2, CD68, and caspase 3. Average pixel intensity was automatically obtained and categorized as high, moderate, or low. Dolichoectasia was defined as a lumen to wall ratio ≥95th percentile.

Results. High MMP-9 staining alone (P = .001) or coexistent with low TIMP-2 staining was associated with dolichoectasia only in HIV-negative individuals (P = <.001). In HIV-positive individuals, MMP-9 was associated with dolichoectasia only when coexpressed with caspase 3 (P = .01). Thinning of the media was associated with CD68 staining (P = <.001) in HIV-negative individuals, while caspase 3 was associated with a thinner media only in HIV-positive individuals (P = .01). Media thickness modified the association between lumen to wall ratio and MMP expression.

Conclusions. A role for MMP/TIMP balance in dolichoectasia appears more prominent in HIV-negative individuals, while apoptosis, mediated by caspase 3, is the most important determinant of media thinning in HIV-infected individuals. Furthermore, apoptosis and media thickness appear to mediate the effects of MMP in the HIV-infected population.

Stroke is a leading cause of disability in the world, and its prevalence and incidence have risen in individuals with human immunodeficiency virus (HIV) [1]. Although the rise in cardiac disease frequency among HIV-infected individuals has been attributed to increased atherosclerotic burden, stroke mechanisms remain less well understood [2]. Stroke is a heterogeneous disease in which the mechanisms dictate natural history and, in some cases, also dictate the preventive measures to reduce the risk of recurrence [3]. For example, stroke due to atrial fibrillation is better prevented with anticoagulation, while stroke due to varicella zoster virus infection of the brain arteries requires antiviral therapy [3]. Dolichoectasia is an arterial disease characterized by elongation and tortuosity that, in the setting of HIV infection, is known as HIV-associated vasculopathy, a form of secondary dolichoectasia [4]. This vasculopathy is associated with up to a fourth of brain infarcts, as noted in a large autopsy series of patients who died with HIV infection [5]. It is typically seen among patients with severe immunosuppression and large-artery strokes, but radiographically there are no reliable methods to distinguish it from dolichoectasia in uninfected individuals, other than the clinical context [6].

It is with this clinical context in mind that we hypothesized that mechanisms leading to dilatation of the brain large arteries in HIV-positive individuals differ in the absence of HIV. Preliminary work has demonstrated that thinning of the media may be a preclinical stage in the development of brain large-artery dilatations among HIV-infected patients [7]. As thinning of the media seems to precede and enable dilatation of the wall by flow-induced outward remodeling, processes such as matrix degradation and apoptosis that affect the integrity and function of smooth muscle cells may impact the predisposition of a vessel to dilate [8]. In this analysis, we tested the hypothesis that high levels of expression of metalloproteinases and markers of apoptosis are associated with outward remodeling in large arteries of autopsied brains from HIV-positive individuals and that, in HIV-negative individuals, the associations would be different.

METHODS

Brain large arteries were selected from the Brain Arterial Remodeling Study (BARS), a collection of 336 autopsied brains from 4 different brain banks or collections, as previously described [9]. For this analysis, we selected HIV-positive and HIV-negative individuals matched by age (±3 years) and sex. For immunohistochemical staining, a BARS subsample was selected on the basis of the greatest number of arteries available per case and the availability of matched HIV-negative controls. All the HIV-positive individuals in the immunohistochemistry study (n = 84) were obtained from the Manhattan HIV Brain Bank, where they were prospectively followed and voluntarily agreed to donate their brains upon death. The diagnostic pathological procedures of this bank have been previously reported [10]. The HIV-negative controls (n = 78) for this analysis came from the Manhattan HIV Brain Bank (n = 30) and from the Macedonian/New York State Psychiatric Institute brain collection (n = 48). For both studies, demographic characteristics and vascular risk factors such as age, sex, and ethnicity were self-reported, reported by families in a structured interview, or extracted from medical records.

The autopsied brains were fixed in formalin for at least 2 weeks after removal from the body. Brain arteries were removed by one of us (J. Gutierrez), and the most proximal and the most distal parts of each artery (eg, the internal carotid and vertebral arteries) were segmented into 5-mm lengths by cross-sectional cuts. The methods and reliabilities of the measurements obtained have been previously reported [11]. For this study, the lumen to wall ratio was used as the main measure of arterial remodeling, with a higher lumen to wall ratio indicative of outward remodeling and a smaller lumen to wall ratio indicative of inward remodeling. Dolichoectasia was defined as a lumen to wall ratio of at least the 95th percentile, stratified by HIV status and by sex to allow a fair comparison among the groups. Large artery atherosclerosis was defined by the presence of a clearly defined atheroma independent of whether a thin fibrous cap was present or not, as previously described [12].

Immunohistochemical Analysis

Brain large arteries were stained for matrix metalloproteinase 2 (MMP-2), MMP-3, MMP-9, tissue inhibitor of metalloproteinases 1 (TIMP-1), TIMP-2, tumor necrosis factor α (TNF-α), CD68, and cleaved caspase 3. Brain arteries from 22 randomly selected cases (11 with HIV infection) were stained for chemokine (C-C motif) ligand 2 (CCL2; also referred to as monocyte chemoattractant protein 1). The methods used to stain for each of these antibodies are reported in Supplementary Table 1.

Average Pixel Intensity Quantification

Each microphotograph was processed to remove debris, blood residues, and folding artifacts, as described previously [13]. Average pixel intensity for MMP-2, MMP-3, MMP-9, TIMP-1, TIMP-2, TNF-α, and caspase 3 were obtained automatically, using the Visiopharm Integrator System, version 4.6.3.857 (Hoersholm, Denmark), by color thresholding. Batches of 50 slides were stained simultaneously with each antibody. Each batch contained random cases with and without HIV infection. To allow comparison of staining intensity among different batches of slides, a separate measure of average intensity background staining was created to adjust for it. Staining of antibody was considered high or low if the background-adjusted average pixel intensity was in the highest or lowest tercile, respectively. We created a semiquantitative method for CD68 rating. In this method, each artery was rated individually by one of us (J. Gutierrez) blinded to the HIV status of the individual from whom the artery was obtained. The staining was rated in each arterial layer by use of a 5-category scale that ranged from scattered within a focal area (score, 1) to coalescent clusters (score, 5). The total CD68 score was created by adding the value of each individual layer with a total score ranging from 0 to 15. We validated the score against cross-sectional areas derived by automated pixel segmentation of brown staining. This method has good reliability [13].

Statistical Analysis

Differences in demographic and vascular risk factor prevalence among HIV-positive and HIV-negative individuals were assessed with χ² analysis or the Student t test as indicated. The main dependent variables used here to assess remodeling were lumen to wall ratio, analyzed either as a continuous variable or categorically as dolichoectasia (yes or no), and media thickness (in microns: analyzed as a continuous variable). Because large-artery atherosclerosis and dolichoectasia are typically located at the opposite extremes of the lumen to wall ratio spectrum, and because some of the proteins assessed here may also play a role in large-artery atherosclerosis, we adjusted for atherosclerosis in all models. We used generalized linear models with a logit function for a binomial distribution of the categorical variables or mixed models for continuous variables by using each artery as the unit of observation. We used multilevel design of each of these models to account for the lack of independence among arteries extracted from the same individual and adjusted for arterial size [9]. All models were adjusted for ethnicity, age, sex, hypertension, diabetes, dyslipidemia, smoking, cocaine use, and artery type and were stratified by HIV status and tested for statistical interactions to assess whether the reported associations varied by HIV status. Bonferroni correction was used to adjust for multiple comparisons (calculated as 0.05/number of models in each comparison). Bonferroni corrections did not include models investigating costaining. Statistical interactions with a P value of ≤.10 were arbitrarily considered relevant. The statistical analysis was performed using SAS software, version 9.3 (SAS Institute, Cary, North Carolina).

RESULTS

Sample Description

The sample characteristics are reported in Table 1. We analyzed data on 1018 brain arteries, with a mean of 7 arterial segments per individual and HIV-positive cases having a mean of 1 more segment than HIV-negative individuals (P = .001). Based on the 95th percentile cutoff, 50 arteries had dolichoectasia. At the subject level, 21% had ≥1 dolichoectatic artery, and 7% had ≥2 dolichoectatic arteries. All HIV-positive cases died after 1996, during the cART era. z

Relationship Between MMP Staining and Dolichoectasia

Among HIV-negative individuals, dolichoectasia was associated with high MMP-9 expression (β = 1.34 and P = .001) and with low TIMP-2 expression plus high MMP-9 expression (β = 1.33, P < .001, and P_{interaction by HIV status} = .06). Dolichoectasia was associated with low TIMP-2 expression and high MMP-3 expression independent of HIV status (P_{interaction with HIV status} = .22), but it reached significance in HIV-negative individuals only (β = 1.20 and P = .002 vs β = 0.67 and P = .31). The associations between expression of MMPs and TIMPs and dolichoectasia among HIV-positive individuals were attenuated (Table 2).

Inflammation and Apoptosis in Relationship to MMP/TIMP Staining and Dolichoectasia

A higher CD68 score was associated with high MMP-3 expression, whereas high caspase 3 expression was associated with high MMP9 expression, independent of HIV status (Table 3). High MMP-3 expression was differentially associated with caspase 3 expression on the basis of HIV status. High MMP-3 expression was associated with high caspase 3 expression among HIV-negative individuals, while the association was negative among HIV-positive individuals (P_interaction = .017). There were no other differential associations between markers of inflammation and apoptosis with respect to HIV status.

Dolichoectasia was associated with costaining for TNF-α and MMP-9 (β = 1.12 and P = .01) and with costaining of caspase 3 with MMP-2 (β = 1.29 and P = .006), MMP-3 (β = 1.07 and P = .02), and MMP-9 (β = 1.30 and P = .009) in HIV-negative individuals. Dolichoectasia was associated only with costaining of caspase 3 with MMP-9 (β = 1.30 and P = .01) in HIV-positive individuals.

In a BARS random subsample of 22 cases, high CCL-2 expression was associated with HIV infection (β = 1.77 and P = .009). There was a suggestion that high CCL-2 expression was associated with adventitial CD68 score in HIV-infected individuals to a greater extent than that in uninfected individuals (P_interaction = .19). High CCL-2 expression, however, was not associated with high MMP-2, MMP-3, or MMP-9 expression independent of HIV status.

Media Thickness as an Effect Modifier Between Inflammation and Apoptosis With Dolichoectasia

Dolichoectasia was associated with thinning of the media (ie, media thickness in the bottom 5th percentile) in HIV-negative individuals (β = 2.33 and P = .005) and HIV-positive individuals (β = 2.30 and P < .001). However, the strength of the association between the stainings studied here and media thinning varied by HIV status. Among HIV-negative individuals, thinning of the media was associated with a higher CD68 score to a greater extent than in those with HIV infection (P_interaction = .002), while among HIV-positive individuals, thinning of the media was associated with high caspase 3 expression to a greater extent than among HIV-negative controls (P_interaction = .027; Table 4 and Figure 1). A weaker staining of TIMP-2 was associated with a thicker media among HIV-negative controls to a greater extent than in HIV-positive individuals (P_interaction = .06; Table 4). Furthermore, media thickness influenced the direction of effects in associations between inflammation and apoptosis, and elastolytic activity and lumen to wall ratio (Table 5). A progressively thicker media rendered negative the β estimates for the association between high TNF-α expression and high MMP-9 expression among HIV-positive individuals (P_trend = .003) and not among HIV-negative controls (P_trend = .978 and P_{interaction by HIV status} = .03). The media thickness had the same effect modification between caspase 3 and MMP-9 independent of HIV status with lumen to wall ratio, but the strength of association was greater among HIV-positive individuals (P_{interaction by HIV status} = .003).

In a post hoc analysis using the fully adjusted model described in Tables 2 and 3, we found that, in HIV-positive individuals, high caspase 3 intensity was associated with a lower nadir CD4⁺ T-cell count (mean β [±SD], 0.25 ± 0.01 per every 50-cell drop in CD4⁺ T-cell count; P = .02), number of antemortem visits with CD4⁺ T-cell counts of <100 cells/μL (β = 0.25 ± 0.09 and P = .008), years of HIV infection (β = −0.16 ± 0.06 and P = .01), history of opportunistic infections (β = −2.01 ± 0.53 and P < .001), and CD4⁺ T-cell count of <200 cells/µL at the time of death (β = −2.38 ± 0.71 and P < .001).

DISCUSSION

Chronic HIV infection is associated with an increased incidence of vascular events. Among patients with HIV infection and stroke, HIV-associated vasculopathy, a form of secondary dolichoectasia, is a frequent and poorly understood cause of stroke. In this analysis, we report evidence supporting differential mechanisms in dolichoectasia, by HIV status. While high MMP-9 expression combined with a low TIMP-2 expression are associated with dolichoectasia in HIV-negative individuals, high MMP-9 staining was only associated with dolichoectasia in HIV-positive individuals in the context of high expression of caspase 3, a marker of apoptosis. In fact, high caspase 3 expression was associated with thinning of the media in HIV-positive individuals only, further highlighting that apoptosis may play a role in HIV-associated vasculopathy. We also found that the thickness of the media modifies the association of MMPs jointly expressed with TNF-α or caspase 3, such that it is only positively associated with dolichoectasia in arteries with a thinner media, whereas a thick media seems to abrogate the association. Because thinning of the media may weaken the tensile strength of the arterial wall and enable dilatation, the unique role of apoptosis in media thinning and outward remodeling, particularly in HIV-positive individuals, should be reproduced in animal models of HIV. If confirmed, apoptosis may be considered a target for therapeutic interventions.

The association between media thinning and caspase 3 expression in brain large arteries from HIV-positive individuals is novel. This result fits well with animal models of outward arterial remodeling in the absence of HIV. In the initial phases of arterial dilatation in a mouse model of fibrillin-deficient aortic aneurysm, upregulation of transforming growth factor β (TGF-β) was associated with increased caspase 3 activity and increased arterial diameter, but the thickness of the media remained stable [8]. The authors of this work hypothesized and produced data to support a compensatory overproliferation of smooth muscle cells in the initial phases of aneurysm formation. It is also possible that the relationship between TGF-β upregulation and caspase 3 staining is bidirectional, as data suggest that apoptosis can induce TGF-β secretion, which in turn may reinforce the upstream regulation of more caspase 3 [14]. Because upregulation of TGF-β is typically associated with dilatative arteriopathy, TGF-β dysregulation appears as a plausible upstream pathway to be explored in human brain arterial remodeling.

We believe, however, that in HIV-positive individuals, apoptosis may also be related to HIV per se. For example, HIV-related proteins such as tat and gp120 may induce apoptosis in human brain microvascular endothelial cells, arterial-wall derived mesenchymal cells (including myocytes), and blood cells [15, 16]. Evidence exists that HIV may infect smooth muscle cells in coronary arteries, and although evidence of such infection has not been produced in brain arteries, the higher intensity of staining for caspase 3 among those with protracted immunosuppression and lower nadir CD4⁺ T-cell count in this sample suggests that persistent immunosuppression may enable or perpetuate arterial wall infection and trigger apoptosis [17]. Apoptosis of cardiomyocytes has also been described as a key mediator of HIV-related dilated cardiomyopathy [18]. Thinning of the media appears as a key mediator of HIV-induced outward remodeling, either through the reduction of the overall tensile strength of the wall or by altering the homeostatic balance of the extracellular matrix, thus weakening the wall and predisposing these vessels to flow-induced dilatation. The fact that not all individuals with persistent immunosuppression develop outward remodeling suggests, however, that there may be different congenital or acquired predispositions to outward remodeling among HIV-positive individuals that are not fully understood.

The results noted in HIV-negative individuals are also novel. The association between dolichoectasia and either increased MMP staining or imbalances between MMP and TIMP expression have been previously reported in animal models, but little is known about local staining of MMP in human examples of primary dolichoectasia. For example, MMP-2 and MMP-9 have been implicated in flow-induced outward remodeling in models of arteriovenous fistula in rabbits [19]. Pharmacological inhibition of MMP with doxycycline can prevent flow-induced remodeling in rat mesenteric collateral arteries [20]. Upregulation of MMP and downregulation of TIMP seems to be a common phenomenon leading to outward remodeling [19, 21]. Our results also suggest that MMP-9 is more relevant than other MMPs for large-artery remodeling, similar to what has been reported in human aortic aneurysms [22]. Of interest, the intense costaining of TNF-α and MMP-9 was associated with dolichoectasia. Moreover, the monocyte chemotactic protein 1 staining intensity was associated with HIV infection and with adventitial macrophages, which we have demonstrated are far more prevalent among cases with HIV infection [13]. This association may lend support to the role that inflammation and reactive oxygen species have in the activation of the zymogen forms metalloproteinases and inflammatory cell recruitment to the arterial wall [23].

The large sample size, the systematic and reproducible methods to analyze the morphometric and immunohistochemical properties of the brain arteries, and the sex and age matching between HIV-negative and HIV-positive individuals are important strengths of this work. Adjustment for multiple comparisons with Bonferroni correction provides further support to the strength of the main associations noted. There are also limitations to these data that need to be considered. Causality can only be suggested in these cross-sectional results, a major limitation of the majority of autopsy-based studies. By invoking animal model data, we hope to strengthen the argument made here as it refers to causality. Furthermore, immunohistochemical analysis is qualitative, and thus we cannot argue that the staining of a given protein is more or less intense than that for others. By automating the intensity of staining we attempted to reduce the bias inherent to ordinal adjudication of staining intensities, but we did not have absolute values of concentrations of markers. It is also possible that strong background staining may have diluted the associations noted here, and thus there remains a need for a more thorough and quantitative analysis of protein staining to validate and expand these results.

In summary, increased staining intensity for MMP-9 alone or coexistent with weak staining intensity for TIMP-2 was associated with dolichoectasia in HIV-negative individuals, while in HIV-positive individuals only, when MMP-9 was coexpressed with caspase 3, the association with dolichoectasia was noted. Thinning of the media was associated with CD68 in HIV-negative individuals, while caspase 3 expression was associated with media thinning only in HIV-positive individuals. The thickness of the media seemed to be an effect modifier of the association between MMP staining and dolichoectasia, with a thicker media attenuating the dilatative effects of MMP in this sample. If further animal experiments confirm these results, novel therapeutic targets could be tested to modify the natural history of dolichoectasia in the population with and that without HIV infection in a more personalized, pathogenesis-specific manner.

Notes

Financial support. This work was supported by the National Institutes of Health (MH64168 to A. J. D. and U24MH100931 and R25MH080663 to S. M.) and the American Heart Association (13CRP14800040 to J. G.).

Potential conflicts of interest. J. G. reports a research grant related to the submitted work and a research grant and consulting not related with the submitted work. S. M. reports research grants and employment income outside the submitted work. A. J. D. reports research grants outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References