Urinary Eicosanoid Metabolites in HIV-Infected Women with Central Obesity Switching to Raltegravir: An Analysis from the Women, Integrase, and Fat Accumulation Trial

Chronic inflammation is a hallmark of HIV infection. Eicosanoids reflect inflammation, oxidant stress, and vascular health and vary by sex and metabolic parameters. Raltegravir (RAL) is an HIV-1 integrase inhibitor that may have limited metabolic effects. We assessed urinary F2-isoprostanes (F2-IsoPs), prostaglandin E2 (PGE-M), prostacyclin (PGI-M), and thromboxane B2 (TxB2) in HIV-infected women switching to RAL-containing antiretroviral therapy (ART). Thirty-seven women (RAL = 17; PI/NNRTI = 20) with a median age of 43 years and BMI 32 kg/m2 completed week 24. TxB2 increased in the RAL versus PI/NNRTI arm (+0.09 versus −0.02; P = 0.06). Baseline PGI-M was lower in the RAL arm (P = 0.005); no other between-arm cross-sectional differences were observed. In the PI/NNRTI arm, 24-week visceral adipose tissue change correlated with PGI-M (rho = 0.45; P = 0.04) and TxB2 (rho = 0.44; P = 0.005) changes, with a trend seen for PGE-M (rho = 0.41; P = 0.07). In an adjusted model, age ≥ 50 years (N = 8) was associated with increased PGE-M (P = 0.04). In this randomized trial, a switch to RAL did not significantly affect urinary eicosanoids over 24 weeks. In women continuing PI/NNRTI, increased visceral adipose tissue correlated with increased PGI-M and PGE-M. Older age (≥50) was associated with increased PGE-M. Relationships between aging, adiposity, ART, and eicosanoids during HIV-infection require further study.


Introduction
Fat redistribution in HIV-infected patients is associated with antiretroviral therapy (ART), including protease inhibitors (PI) and nonnucleoside reverse transcriptase inhibitors (NNRTI) [1,2]. Central fat accumulation or lipohypertrophy may be more common in women [3,4] and has been associated with multiple metabolic abnormalities and inflammation in HIV-infected persons on ART [5][6][7][8][9]. Chronic HIV infection is also associated with persistent inflammation [10], and treating HIV infection improved endothelial function in treatment-naïve subjects with low cardiovascular disease (CVD) risk [11]. Given the complex interactions between chronic HIV infection and ART and the likelihood that traditional Framingham prediction may underestimate cardiovascular risk in HIV-infected persons on ART [12,13], novel biomarkers are needed to assess metabolic risk in HIV infection and response to interventions.
Eicosanoids are endogenous products of arachidonic acid metabolism involved in oxidant stress, inflammation, and endothelial function, all of which are important in atherosclerosis and cardiovascular disease pathogenesis [14]. 2
In HIV-negative populations, urinary F 2 -IsoP correlates with traditional CVD risk factors [23] and surrogate measures of CVD including brachial artery flow-mediated dilation and carotid intima media thickness [24,25]. In obese children and adolescents, plasma F 2 -IsoP was positively correlated with visceral adipose tissue (VAT), a marker of CVD risk [26]. Cigarette smoking is also associated with higher eicosanoids [27][28][29][30]. In HIV-infected persons, crosssectional studies to date have identified associations between higher F 2 -IsoP and lipoatrophy, lactic acidosis, virologic suppression on ART, heavy smoking, higher body mass index (BMI) and waist circumference, elevated liver transaminases and hepatitis C virus (HCV) RNA, and female sex [31][32][33][34][35]. Early analyses suggested higher F 2 -IsoP in persons receiving efavirenz or zidovudine [32], and lower levels in those on nevirapine compared with other ART [34] but consistent associations with specific ART drugs or classes have not been seen. It is not yet clear why F 2 -IsoP levels are consistently higher in women (HIV-infected and uninfected) than in men. A recent study of women in Haiti found higher levels of cervical COX-2 and urinary PGE-M in HIV-infected women than in uninfected women and a positive correlation between systemic PGE-M and both plasma HIV RNA and cervical COX-2 levels [36].
Raltegravir (RAL) is an HIV-1 integrase inhibitor that has not been associated with metabolic perturbations or fat redistribution during short-or long-term therapy [37][38][39].
A randomized, open label study was designed to assess the effects of switching from PI-or NNRTI-based ART to a RAL-based regimen in women with lipohypertrophy and suppressed HIV-1 RNA on stable therapy [40]. Adipose tissue volumes by computerized tomography (CT), anthropometrics, and fasting metabolic parameters were performed. The objectives of these secondary analyses were to determine (a) the 24-week change in F 2 -IsoP and other urinary eicosanoid metabolites and (b) correlations between 24-week changes in F 2 -IsoP, other urinary eicosanoid metabolites, and changes in VAT, the primary outcome of the parent study. We hypothesized that urinary F 2 -IsoP and other eicosanoid metabolites would decrease after 24 weeks in women switching to RAL compared to those continuing a PI/NNRTI and that these decreases would correlate with decreased VAT.

Materials and Methods
Complete methods for the parent study have been published previously [40]. Briefly, women with HIV-1 RNA <50 copies/mL, stable ART including two NRTIs (tenofovir or abacavir and emtricitabine or lamivudine) plus a PI or NNRTI, and central adiposity (waist circumference >94 cm or waist: hip > 0.88) were enrolled at five centers in North America from September 2008 to July 2010 and randomized 1 : 1 to switch their PI/NNRTI to open label RAL 400 mg twice daily (RAL arm) or continue to present ART for 24 weeks (PI/NNRTI arm). Relevant exclusion criteria included current use of metformin, thiazolidinediones, or androgen therapy, use of growth hormone or growth hormonereleasing factor in the six months prior to screening, change or initiation of lipid-lowering therapy in the three months prior to screening, and intent to significantly modify diet or exercise habits during the study. The primary endpoint of the parent trial was between-group change in percent of VAT volume 24 weeks following a switch to RAL versus continued PI or NNRTI. All study procedures were approved by the institutional review boards of the participating institutions, and all subjects provided informed consent prior to initiation of study procedures. Procedures were performed in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of the World Medical Association.

Anthropometric Measurements.
Visceral and subcutaneous adipose tissue (VAT and SAT, resp.) volume was measured via single slice L4-L5 CT scan at weeks 0 and 24. Scans were performed locally but standardized and read centrally by a blinded reader at the Tufts University Body Composition Center. Waist, hip, and neck circumferences were performed according to AIDS Clinical Trials Group standards [41] at weeks 0, 12, and 24.

Laboratory Assessments.
Fasting (>8 hours) glucose, lipoprotein profile, high-sensitivity C reactive protein (hsCRP), and CD4+ T cell counts were assessed at weeks 0, 12, and 24. HIV-1 RNA (50 copies/mL assay sensitivity) was measured at screening and weeks 4, 8, 12, and 24. Labs were performed at the individual sites in real-time and according to local standards.

Urinary Eicosanoids.
Clean-catch urine samples were collected, and three-milliliter (mL) aliquots of urine were stored at −80 ∘ C until analysis. Samples were shipped overnight on dry ice to the Vanderbilt University Eicosanoid Core Laboratory where analyses were performed. Urinary F 2 -IsoP, TxB 2 , and PGI-M were measured using gas chromatography-negative ion chemical ionization mass spectrometry employing stable isotope dilution methodology, as described elsewhere [15,18,20,42]. PGE-M was measured by liquid chromatography-mass spectroscopy (LC-MS), as previously described [19]. Results for all urinary metabolites are presented as ng/mg urinary creatinine (cr). At one study site, a freezer malfunction led to transient thawing of urine samples from nine subjects. Urinary eicosanoid results from these samples were not statistically different from the other sites, and analyses with and without data from these samples were performed (data not shown). As there were no substantive changes in the results, we report results including data from all sites.

Statistical Analysis.
Baseline characteristics of the two randomization groups were compared using the Mann-Whitney test for continuous variables and Fisher's exact test for categorical variables. Analyses also included Spearman correlations between continuous baseline and 24-week change variables. Median values and interquartile ranges (IQR) are reported for continuous variables, and percentages are reported for categorical data. Comparison of median between-group 24-week change scores for eicosanoids was performed using the Wilcoxon signed-rank test. The primary analysis was as-treated, excluding subjects who did not remain on the study regimen and/or did not have an observed primary endpoint. Generalized linear models assessed associations between eicosanoid changes and study arm, adjusting for baseline PI use, BMI, smoking status, study site, and age (≥50 versus <50 years). All statistical tests were two-sided with a nominal level of 0.05. Given the exploratory nature of these analyses, we did not adjust results for multiple testing. Data analysis and management was performed using SAS 9.2 (SAS Institute, Inc., Cary, NC, USA).

Baseline Demographics.
Sixty-one women were screened and 39 enrolled in the trial. Eighteen subjects were randomized to the RAL arm and 21 to continue PI/NNRTI. One subject from each arm withdrew for reasons unrelated to the study intervention [40], leaving 37 subjects who completed the week 24 primary endpoint. Complete demographic and baseline clinical characteristics of the 37 participants included in the as-treated analysis are provided in Table 1. At baseline, the study groups were well balanced, with the exception of the PI/NNRTI arm having a higher rate of current smoking (60% versus 24%; = 0.045). The median age was 43 years, BMI 32 kg/m 2 , and 75% of subjects self-identified as Black or Hispanic. Sixty-two percent of subjects were on a PI at entry (versus 38% NNRTI), and the most commonly prescribed NRTI was tenofovir (78%).  (Table 1). When comparing study arms, baselines PGI-M, PGE-M, and TxB 2 were all lower, and F 2 -IsoP was higher in the RAL arm (Table 1), but only PGI-M was statistically different ( = 0.005). Baseline PGE-M tended to be higher in current smokers ( = 0.1; data not shown). Eicosanoid levels also tended to differ by baseline NRTI, with women receiving abacavir having consistently lower levels (Figure 1), including statistically significantly lower F 2 -IsoP ( = 0.05; Figure 1(a)) and TxB 2 ( = 0.04; Figure 1(d)) levels. None of the urinary eicosanoids were correlated with fasting lipids, glucose, insulin resistance, or hsCRP in this study population at baseline (data not shown).

Changes in Urinary Eicosanoids by Study Arm.
Median 24-week urinary eicosanoid levels and changes from baseline are shown in Figure 2 and Table 2 No other statistically significant correlations between lipids and urinary eicosanoids were seen. Changes in eicosanoid levels over 24 weeks were not statistically different by baseline NRTI (abacavir versus tenofovir; data not shown).   16 (43)  4 (24)   = 0.04), independent of the covariates above (Table 3). No baseline factors were significantly associated with changes in other urinary eicosanoids (Table 3).

Discussion
In these HIV-infected women with central adiposity and suppressed HIV RNA on ART, switching PI-or NNRTIbased ART to RAL did not have significant effects on urinary eicosanoids over 24 weeks. Overall, F 2 -IsoP, PGE-M, and TxB 2 levels were higher and PGI-M levels were lower than published levels reported in healthy adults [15, 18-20, 43, 44]. Although formal comparisons were not performed, urinary F 2 -IsoP (lower), PGE-M (higher), and TxB 2 (higher) levels also differed from those observed in previously studied HIVinfected women who were younger and had lower BMI [33]. Of note, several of the markers differed-though not with statistical significance-at baseline between the two study arms. Given known effects of smoking on these biomarkers [27][28][29], this difference may have been driven in part by the significantly greater number of smokers randomized to the PI/NNRTI arm, and this may therefore have limited our capacity to identify differences in changes over time or due to RAL switch. Although baseline PGE-M tended to be higher in current smokers ( = 16; median [IQR] 9.9 [4.9-14.4]) than nonsmokers ( = 21; median [IQR] Tenofovir DF F 2 -IsoP (ng/mg cr) * P = 0.05 Tenofovir DF PGE-M (ng/mg cr)

Abacavir
Tenofovir DF , in a multivariate model, neither current smoking nor study arm was significantly associated with 24-week change in PGE-M and did not attenuate the relationship between age and change in PGE-M. Additionally, F 2 -IsoP, which is increased in HIV-infected and uninfected smokers, tended to be higher in smokers at baseline ( = 0.06) but was not significantly higher in the PI/NNRTI arm. Age and BMI were lower and higher in the RAL than the PI/NNRTI arm, respectively (Table 1), and though they were not statistically different, we did include these as covariates in adjusted models.
In primary analyses, women in the switch arm had significant improvements in total and LDL cholesterol but did not have a statistically significant improvement in VAT compared to women continuing an NNRTI or PI [40]. They also had a significant decrease in soluble CD14, a marker of monocyte activation [45]. A recent analysis of extensively treatment experienced, predominantly male, subjects switching enfuvirtide to RAL in France reported significant 24week decreases in interleukin-6, D-dimer, and hsCRP [46] that were not observed in these less treatment-experienced women with central adiposity [45]. In subjects remaining on PI/NNRTI, increasing VAT was marginally correlated with increasing PGI-M and PGE-M over 24 weeks of follow-up, suggesting a relationship between these markers and central adiposity that was altered by a switch to RAL. Older age (≥50 years) at enrollment was associated with an increase in PGE-M independent of study arm, smoking status, PI use at baseline, or other factors. Urinary eicosanoids other than TxB 2 were not significantly associated with BMI at baseline. This was unexpected given associations with F 2 -IsoP and TxB 2 in prior cross-sectional studies of HIV-infected persons [33,34], and this may be due to the inclusion of both males and females in prior analyses and/or the high prevalence of obesity and relative lack of normal BMI ranges in this study population.
This is the first study to prospectively assess the effects of an ART switch on urinary eicosanoid metabolites. The small sample size of our study and the imbalance of smokers in the PI/NNRTI arm likely limited our ability to detect differences between study arms. Nonetheless, intriguing trends and preliminary associations were noted. Although not routinely

Baseline
Week 24 measured in clinical practice, eicosanoids are known markers of cardiovascular and metabolic disease risk in HIV-negative populations. In particular, F 2 -IsoP has been associated with CVD disease risk factors, hsCRP, carotid intima medial thickness, coronary artery calcium, and angiographic coronary artery obstruction [23,24,47]. Urinary TxB 2 was associated with a composite clinical cardiovascular outcome [48]. PGE 2 is a complex mediator of inflammation and vasodilation with variable effects on vascular tone depending on the tissue and prostanoid receptor [49]. Increased PGE-M has also been associated with malignancies in HIV-negative populations [50][51][52][53], and urinary PGE-M correlated with plasma HIV RNA and cervical COX-2 levels in a recent study of Haitian women [36].
It was notable that baseline F 2 -IsoP and TxB 2 were lower in subjects on abacavir. Abacavir exposure has been associated with increased cardiovascular risk in HIV-infected persons [54], but the association has not been consistent [55]. Although potential mechanism(s) are not clearly defined, recent studies have reported abnormal platelet reactivity with abacavir [56,57]. To our knowledge, TxB 2 has not been assessed in these or other studies. Lower F 2 -IsoP in these abacavir-treated subjects and in a prior cross-sectional analysis of men and women [33] suggests that if there is excess cardiovascular risk due to abacavir, it is independent of lipid peroxidation-related pathways; prospective studies would be needed to determine this. Due to providers' knowledge of potential cardiovascular risk with abacavir and risk of renal  -- toxicity with tenofovir DF, it is also possible that abacavir use is simply a marker of some other unmeasured factors associated with risk for these conditions and lower urinary eicosanoid levels.
In addition to the qualifications above, this analysis has other limitations that should be noted. Use of aspirin or nonsteroidal anti-inflammatory drugs (NSAID) was not an exclusion criterion for the parent study, and detailed information on dosing was not collected. Given the even distribution of NSAID use across study arms (Table 1), we did not adjust for this variable in multivariate models and do not believe it would explain differential associations within or between groups. Neither menopausal status nor sex hormone levels were ascertained as part of this study. Based on data from HIV-infected women [58], the age distribution of our population suggests the majority of subjects were pre-or perimenopausal. The association between age and PGE-M may have been due to postmenopausal changes in the older women (≥50 years). The number of older women included in this study was small. Platelet reactivity assays were not performed, so relationships between eicosanoids and platelet function cannot be determined. CT scan was used to assess SAT, so we were unable to fully assess peripheral (limb) lipoatrophy. Finally, this analysis may have been too small and/or of insufficient duration of follow-up to detect meaningful changes in eicosanoids. Additional studies and longerterm follow-up in HIV-infected persons are needed to further elucidate the role of eicosanoids in metabolic and other agingrelated comorbidities and determine their role as useful clinical biomarkers.   The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflict of Interests
T.H. received a research grant for the conduct of this study through the Merck and Co. Investigator-Initiated Studies Program. J.S.C. received a research grant for the conduct of the parent study through the Merck and Co. Investigator-Initiated Studies Program. J.E.L. has provided consulting services to Merck and Co. and to GlaxoSmithKline. G.A.M. has served as a scientific advisor or speaker for Bristol Myers Squibb, Tibotec, and Merck and has received research grants from Bristol Myers Squibb and Gilead Sciences and is currently serving as the DSMB Chair for a Pfizer-sponsored study. C.A.W. has received grant funding from GlaxoSmithKline and Theratechnologies, and served as an event adjudicator for a Pfizer study. A.M. has served as the Medical Director for HIV/Endocrinology at EMD Serono, Inc. but performed this work independently of this position through her affiliation with Tufts University. S.L.W. has provided consulting services to Merck and Co. and received a research grant from Merck Frosst Canada Ltd. to help support this work. She has also served as an advisor and speaker to Abbvie