Undercarboxylated osteocalcin is associated with cognition in women with and without HIV

Abstract

Introduction:

Bone loss and cognitive impairment are common in women living with HIV (WLWH) and are exacerbated by menopause. Bone-derived undercarboxylated osteocalcin (ucOCN) and sclerostin appear to influence cognition. The current study investigated whether the circulating levels of these two proteins are associated with cognition in midlife WLWH and demographically similar HIV seronegative women.

Methods:

Plasma samples from women enrolled in a musculoskeletal (MSK) substudy within the Women’s Interagency HIV Study (WIHS) were used to measure ucOCN and sclerostin. A neuropsychological (NP) test battery assessing executive function, processing speed, attention/working memory, learning, memory, verbal fluency, and motor function was administered within 6 months of MSK enrollment and every two years after (1 to 4 follow-up visits per participant). A series of generalized estimating equations were conducted to examine the association between biomarkers and NP performance at the initial assessment and over time in the total sample and in WLWH only. Primary predictors included biomarkers, time, and biomarker by time interactions. If the interaction terms were not significant, models were re-run without interactions.

Results:

Neither biomarker predicted changes in NP performance over time in the total sample or in WLWH. ucOCN was positively associated with executive function in the total sample and in WLWH and with motor skills in WLWH. ucOCN was negatively associated with attention/working memory in the total sample. There were no significant associations between sclerostin and NP performance.

Conclusion:

The current study suggests an association between bone-derived ucOCN and cognition in women with and without HIV infection.

Introduction

About one third of people living with HIV (PLWH) in the US were over 55 years of age in 2018 and 17% of all new diagnoses were in people over 50 years old^1,2. As PLWH age, more women living with HIV (WLWH) reach menopause and do so earlier compared to their HIV-seronegative (HIV−) counterparts^3,4. Two chronic conditions, bone loss and cognitive decline are worsened by menopause transition^5–7 and appear to affect WLWH more than HIV− women⁸.

A growing body of literature suggests connections between low bone mineral density (BMD) and poor cognitive performance^9–12. An endocrine role has recently been ascribed to the skeleton with several bone-derived proteins proposed as potential regulators of cognitive function, including undercarboxylated osteocalcin (ucOCN) and sclerostin^13–15. Importantly, both ucOCN and sclerostin levels are affected by menopause transition^16,17, HIV-infection, and antiretroviral therapies^18,19. Therefore, the current study aimed to determine the relationship between these bone-derived proteins and cognitive function in WLHW.

Methods

Study Participants

We utilized stored plasma samples collected during a musculoskeletal (MSK) substudy nested within the Women’s Interagency HIV Study (WIHS), an ongoing multisite longitudinal cohort study of women with HIV infection and demographically similar HIV− women. A total of 252 women, 164 WLWH and 88 HIV− women, were enrolled in the MSK substudy, at one of three WIHS sites: Bronx, Chicago, or San Francisco. MSK participants were between 40 and 60 years of age at the MSK substudy entry (index visit - 2011 to 2014) and were either peri-menopausal or postmenopausal according to Study of Women’s Health Across the Nation (SWAN) study definitions⁵. A detailed description of the MSK substudy has been previously published^20–22. At MSK substudy index visit, WLWH had a CD4 count greater than 100 cells/μL. For the current study, we limited the biomarker testing and analyses to participants who completed a comprehensive neuropsychological (NP) test battery within 12 months after the MSK index visit, leading to a total sample size of 208 women (132 WLWH and 76 HIV−).

Blood Biomarkers

Blood samples were collected at the MSK index visit in BD Vacutainer® CPTs™ containing buffered sodium citrate. Plasma aliquots were stored at −80°C. Samples were thawed and assayed in batch (duplicate) for circulating sclerostin (TECOmedical, Sissach, Switzerland, <5% intra-assay and <5% inter-assay CV) at Rush University Medical Center and ucOCN (Takara Shuzo Co, Otsu, Singa, Japan, <5% intra-assay and <10% inter-assay CV) at the Hektoen Institute of Medicine Laboratory using enzyme-linked immunosorbent assays. ucOCN was log transformed in analyses due to right skewness.

Cognitive Function

A comprehensive NP test battery was administered every two years to English speaking participants as part of routine WIHS visits²³. Participants completed between 1 and 4 NP tests in subsequent six-monthly follow-up visits, with most (55% of the cohort) completing 3 NP tests (see Supplemental Table 1). The domains assessed included learning, memory, attention/working memory, executive function, processing speed, verbal fluency, and motor skills. Learning and memory were assessed with the Hopkins Verbal Learning Test-Revised (HVLT-R) test (outcome for learning=total words recalled across 3 trials; outcome for memory=delay free recall), attention/working memory with letter-number sequencing test (outcomes=total correct from the experimental and control conditions), executive function using Trail Making Test (outcome=time to completion on Part B) and Stroop (outcome=time to completion on Trial 3), processing speed with Symbol Digit Modalities Test (outcome=total correct) and Stroop (outcome=time to completion on Trial 2), verbal fluency with the Controlled Oral Word Association Test (outcome=total correct words generated on “F”, “A”, “S” trials) and Animal Fluency (outcome=total correct animals generated) and motor skills with Grooved Pegboard (outcomes=time to completion with dominant and nondominant hand). Timed outcomes were log transformed to normalize distributions and reverse scored, so higher equated to better performance. Each test outcome was converted to a T-score using demographically corrected norms (accounted for age, years of education, race/ethnicity, Wide Range Abilities Test-3 reading subtest, and prior number of test exposures) that were developed using the HIV− WIHS women²³. Primary outcome measures were the domain-specific continuous T-scores.

Covariates

Covariates included the time in years (including fractions) between index visit and the date of NP assessment, number of current/prior post blood draw NP follow-up visits, site (Chicago, Bronx, vs. San Francisco); age (in years) at the MSK index visit , self-reported race/ethnicity (white, black or Hispanic/other), education (less than high school, high school or greater) and income (less or greater than $12,000 per year), body mass index (BMI), estimated glomerular filtration rate (eGFR, less than or greater than 60 mL/min), elevated depressive symptoms according to the Center for Epidemiological Studies Depression questionnaire (≥16), and in the prior 6 months, tobacco use (current, former, never), alcohol consumption (more or less than 7 drinks per week), and cocaine, crack, or heroin use (current users [past 6 months] or former/never). Additional clinical variables measured at MSK index visit for WLWH included HIV viral load (detectable vs. undetectable [<20 copies/mL]) and CD4 count (cells/μL).

Statistical Design

A series of linear regression models using generalized estimating equations (GEE) with compound symmetry working correlation (independence working correlation gave similar results) were conducted to examine the association between biomarkers (ucOCN, sclerostin) and NP performance at the initial assessment (within 6 months of the MSK visit) and over a median 4.3 years of follow-up in the total sample and in WLWH only. Primary predictor variables included biomarkers, time, and each biomarker by time interaction. If the interaction terms were not statistically significant, we re-ran the model without the interactions. All models in the total sample adjusted for HIV-serostatus, enrollment site, depression symptoms, smoking status, drug use, alcohol use, kidney function (eGFR), income, BMI, ancestry and education level. For models in WLWH only, the same variables were included (except for HIV-serostatus) as well as CD4 count and detectable viral load. A statistical significance threshold of p<0.05 was set for hypothesis testing.

Results

Participant Characteristics

A total of 208 (132 WLWH) were included in the analysis (Table 1). Median age for the total sample was 49 years, 68% were Black, and 36% had less than a high school education. Forty percent of WLWH had detectable HIV RNA (>20 cp/ml). Sclerostin and ucOCN levels in WLWH were similar to the levels in the total sample. With respect to NP performance, the mean T-scores for each cognitive domain were close to 50 in the total sample and in WLWH (Supplemental Table 1). Attention/working memory and learning significantly decreased over the follow-up period in both the total sample and in WLWH (Supplemental Table 2).

Table 1:

Participant Characteristics in the total sample and in WLWH.

Variable	Total sample (N=208) N (%)	WLWH (N=132) n (%)
Age [years, median (IQR)]	49.0 (45, 53)	49.0 (46, 54)
Less than high school education	75 (36)	47 (36)
Annual household income ≤ $12,000	127 (61)	75 (57)
WIHS Enrollment Site
BRONX/MANHATTAN	66 (32)	38 (29)
SAN FRANCISCO	92 (44)	61 (46)
CHICAGO	50 (24)	33 (25)
Race
White	31 (15)	23 (17)
Black	141 (68)	86 (65)
Hispanic/Other	36 (17)	23 (17)
Smoking Status
Never	28 (13)	17 (13)
Current	116 (56)	69 (52)
Past	64 (31)	46 (35)
Former/Recent cocaine, crack, &/or heroin use	33 (16)	17 (13)
Recent heavy alcohol use (> 7 drinks/wk)	37 (18)	19 (14)
Early post-menopause/Post-menopause	131 (63)	84 (64)
Renal dysfunction (eGFR <60 mL/min)	10 (5)	7 (5)
Depressive symptoms (CES-D ≥ 16)	76 (37)	53 (40)
BMI [kg/m2, median (Q1, Q3)]	29.1 (25.0, 33.9)	28.6 (24.7, 32.4)
CD4+ cell count (cells/μl) [median (Q1, Q3)]	-	564 (389.5, 760.5)
HIV RNA viral load undetectable (<20cp/mL)	-	78 (60)
Sclerostin, pmol/L [median (Q1, Q3)]	0.25 (0.16, 0.36)	0.24 (0.16, 0.36)
ucOCN, log ng/mL [median (Q1, Q3)]	0.05 (−0.25, 0.35)	0.03 (−0.26, 0.33)

Associations between bone and cognition

Sclerostin and ucOCN levels in the total sample and in WLWH were not associated with any changes in NP performance over time (all p values>0.10). After removing the biomarker by time interactions, in the total sample, higher levels of ucOCN were associated with better executive function and poorer attention/working memory at the initial NP assessment (Table 2). In the sample of WLWH, similar associations of ucOCN with executive function and attention/working memory were observed, but the latter did not reach statistical significance. Higher ucOCN levels were also significantly associated with better motor function. Sclerostin was not significantly associated with NP performance in either the total sample or WLWH only. Higher levels of sclerostin were associated with lower executive function, although this results was marginally signficiant at the 0.05 level (B=−6.85, p=0.054). Unadjusted univariate associations between biomarkers and cognitive performance are presented in the supplemental materials (Supplemental Tables 3 & 4).

Table 2:

Association of biomarkers and time with cognitive domain T-scores in the total sample and in WLWH

	Entire Cohort (N=208)*		WLWH (N=132)‡
	ucOCN, log ng/mL Biomarker B Estimate (SE, P-value)	Sclerostin, pmol/L Biomarker B Estimate (SE, P-value)	ucOCN, log ng/mL Biomarker B Estimate (SE, P-value)	Sclerostin, pmol/L Biomarker B Estimate (SE, P-value)
Executive Function	2.75 (1.32), 0.037	−6.85 (3.56), 0.054	3.22 (1.63), 0.049	−7.81 (4.13), 0.059
Processing Speed	0.21 (1.53), 0.890	−4.71 (3.39), 0.165	2.26 (1.56), 0.148	−6.07 (3.86), 0.116
Attention/Working Memory	−3.30 (1.18), 0.005	1.80 (3.81), 0.637	−2.93 (1.53), 0.056	−0.23 (4.10), 0.956
Learning	−0.57 (1.48), 0.701	−4.29 (4.11), 0.297	0.72 (1.75), 0.680	−5.14 (4.42), 0.245
Memory	−0.32 (1.45), 0.824	−3.31 (4.39), 0.450	0.87 (1.62), 0.592	−3.12 (4.97), 0.530
Motor Function	1.86 (1.65), 0.259	−6.56 (4.28), 0.125	5.58 (1.77), 0.002	−7.35 (5.03), 0.144

Data are presented as [estimate (SE, p-value)], with p<0.05 bolded for clarity. B=unstandardized coefficient; SE=standard error

NP domains are presented as T-scores adjusted for demographic variables including age, years of education, race/ethnicity, Wide Range Abilities Test-3 reading subtest, and prior number of test exposures

^*models using the entire cohort were adjusted for the following variables as described in the Methods; HIV-serostatus, enrollment site, depression symptoms, smoking status, drug use, alcohol use, kidney function (eGFR), income, BMI, race and education level

^‡models using WLWH only were adjusted for enrollment site, depression symptoms, smoking status, drug use, alcohol use, kidney function (eGFR), income, BMI, race and education level, index CD4 count, and viral load

Discussion

Bone mass loss and cognitive impairment are two common, age-related clinical conditions that can be exacerbated by HIV-infection. Indeed, PLWH are reported to be at three times higher risk for osteoporosis²⁴ and have a 50% greater risk of dementia²⁵ compared to seronegative persons. As menopause increases the risk of both bone loss and cognitive decline^5,6, aging WLWH are a particularly high-risk population. Due to recent discoveries of a connection between the skeleton and brain^9–12, we investigated the relationship between bone-derived proteins and cognitive function in a cohort of WLWH and demographically similar HIV− women. We found associations between undercarboxylated osteocalcin (ucOCN) and both executive function and attention/working memory. Interestingly, the relationship between ucOCN and executive function was positive while the association with attention/working memory was negative, which is contrary to published data that has suggested that these two NP domains are generally positively correlated with one another²⁶. We also found a significant association between undercarboxylated osteocalcin and motor function in analyses restricted to WLWH, suggesting that HIV itself or HIV-related factors may influence bone-brain crosstalk.

Undercarboxylated osteocalcin (ucOCN) has emerged as a bone-derived protein with potential hormonal actions in a variety of extra-skeletal targets, including the brain²⁷. However, most of these data were obtained using mouse models, where studies have demonstrated a positive effect of ucOCN on mouse cognitive performance. The first evidence that linked osteocalcin to cognitive performance was from osteocalcin null mice, wherein loss of osteocalcin led to increased anxiety-like behaviors and reduced spatial learning and memory ²⁸. Further investigations have demonstrated that osteocalcin injection into both osteocalcin null²⁸ and aged mice²⁹ improves cognition. To our knowledge, only three published studies have investigated the association between osteocalcin and cognitive performance in human subjects^30–32 with conflicting results, perhaps as a result of differences in methodology; some studies measured total osteocalcin^30,31, while others used undercarboxylated osteocalcin³². The current study measured the undercarboxylated isoform of osteocalcin, which has been identified as the hormonally active isoform^29,33, and found that higher ucOCN levels were associated with better executive function, but worse working memory in both HIV− women and WLWH.

These findings are somewhat contrary to our previous study that found no associations between undercarboxylated osteocalcin and NP domains that included episodic and working memory measures in a cohort that was older (mean age of 80 years), all HIV−, and inclusive of both men and women³². The reason for this discrepancy is currently unknown due to the relatively few studies that have evaluated the link between osteocalcin and cognition in human cohorts. However, the link between osteocalcin and glucose metabolism has been more extensively evaluated and the relationship has been reported to be affected by both age and sex. Specifically, Jung et al ³⁴ reports that circulating osteocalcin is negatively associated with glucose metabolism in men both younger and older than 50 years, but only present in women below the age of 50, which the authors suggest could be due to elevated remodeling rates around the menopause transition that increase the amount of osteocalcin in circulation. Therefore, it is possible that increasing circulating osteocalcin levels around the menopause transition^35,36 affect the association between osteocalcin and cognition and future studies are necessary in cohorts including more diversity in participant age and sex. This is also one of only two studies in human subjects that has evaluated whether ucOCN levels predict subsequent cognitive change over time. In both this study and our previously published report³², ucOCN was not associated with cognitive change, suggesting that ucOCN is unlikely to be a driver of age-related cognitive performance change.

Sclerostin is a circulating antagonist to the Wnt signaling pathway. In the skeleton, inhibited Wnt signaling has been implicated in the development of age-related osteoporosis³⁷, while in the brain inhibited Wnt signaling is seen with Alzheimer’s Disease progression^38,39. Therefore, suppression of Wnt signaling in both tissues is one possible explanation for shared bone and brain pathologies. Indeed, inhibited bone and brain Wnt signaling has been reported in a mouse model of Alzheimer’s disease⁴⁰. A direct effect of sclerostin in the brain has been reported in mice where intracerebroentricularly administered sclerostin protein induced anxiety-like behaviors and reduced neuronal dendritic branching and dendritic length in the hippocampus⁴¹. While this study demonstrates a direct effect of sclerostin within the brain, the treatment strategy bypasses the blood brain barrier (BBB). Recent findings have identified osteocyte derived sclerostin crosses the BBB, likely transported as part of osteocyte-derived extracellular vesicles⁴². The current study did not find any significant associations between sclerostin and cognitive function, which is consistent with similar studies in HIV− cohorts^30,32,43. However, the association between sclerostin and executive functioning was nearly significant in both the entire cohort (p=0.054) and in WLWH (p=0.059) in this relatively small cohort that included women who were younger than those in other published studies (mean age of 49 years vs. 80³² and 58⁴³). Importantly, sclerostin levels are known to be influenced by age⁴⁴ and menopause¹⁶, therefore future work in a larger cohort better able to adjust for these variables is necessary to evaluate whether a relationship between bone loss and decline in executive function exists during menopause.

As PLWH age, an increasing number of WLWH are reaching menopause⁴⁵. WLWH may begin the transition to menopause at younger ages when compared to their HIV-seronegative (HIV−) counterparts due to a variety of risk factors, including HIV infection^3,4,46,47. HIV also increases the severity of menopause symptoms^4,48,49, including bone loss²². Prior work in the MWCCS found that cognitive decline during the menopause transition did not appear to be worsened by HIV infection⁵⁰, and may explain the similar relationships between ucOCN and cognition in both WLWH and the entire cohort. Improving the accuracy of measuring the menopausal transition through objective biomarkers, self-report and statistical modeling is an area of active investigation in the MWCCS.

An interesting finding in the current study is the positive association between ucOCN and motor skills that was present only in WLWH. In exploratory models using only HIV− women, there was no association between ucOCN and motor skills and the estimate indicated a negative relationship (B=−3.32, p=0.254, data not shown). Several studies have suggested that ucOCN influences muscle performance. Mechanistic studies in osteocalcin knockout mice demonstrate that ucOCN promotes glucose metabolism in myocytes⁵¹, and correlative studies in humans show a positive association between circulating ucOCN and muscle strength⁵². Fewer studies have assessed the relationship between ucOCN and physical function. In a rat model of Parkinson’s Disease, osteocalcin treatment diminished locomotor defects present in affected rats⁵³. However, somewhat conversely, in a study of women with a mean age of 75 years, an increased ratio of ucOCN to total osteocalcin was associated with worse physical function and increased fall frequency⁵⁴. The fact that the ucOCN and motor skills relationship was only significant in WLWH suggests that factors related to HIV serostatus may be contributing to the associations. The current study is underpowered to tease out the various contributions of HIV-related factors, such as CD4 count, or current antiretroviral regimen, that have been reported to contribute to cognitive decline in PLWH^55,56. Therefore, future work in larger cohorts of PLWH would be necessary to tease out what is driving these associations.

Study limitations include a relatively small sample size, as described above. It is also currently unknown what is causing the directionality differences in the relationship between ucOCN and executive function and attention/working memory. Further, because we only measured bone biomarkers at a single time point, we were unable to determine whether ucOCN or sclerostin changes over time are associated with cognitive performance. Additionally, we did not have information on attention-deficit/hyperactivity disorder (ADHD) in WIHS participants, which has been associated with NC dysfunction⁵⁷. We also did not have information about previous head trauma in this cohort. However, there were also several strengths, including the well-matched cohort of HIV-seronegative and HIV-seropositive women with detailed characterization of their demographics and lifestyle factors. Further, the focus on the menopause transition is also considered a strength as this period is associated with both rapid bone mass loss⁴ and cognitive decline⁶.

In summary, this study suggests that plasma ucOCN levels are associated with cognitive performance in women living with and without HIV during the menopause transition. While the relationships between ucOCN and executive functioning and attention/working memory appear to be independent of HIV-serostatus, the association between increased ucOCN and better motor skills appears to be dependent on currently unknown HIV-related factors. The clinical implications of these findings are currently unknown and further studies are necessary to determine whether osteocalcin holds diagnostic or therapeutic potential for cognitive impairment.