Assessment of macrovascular and microvascular function in aging males

Abstract

Aging is associated with reductions in endothelial function, observations primarily reported using brachial artery ultrasound. There is growing interest in the use of peripheral artery tonometry (PAT) of microvessels in the fingertip to assess endothelial function because it is less technically demanding and has a high sensitivity and specificity for assessing coronary endothelial function. Moreover, similar to brachial artery flow-mediated dilation (FMD), PAT predicts cardiovascular disease outcomes. However, the relationship between PAT and FMD have yet to be examined in the context of aging. To address this question, reactive hyperemic index (RHI) using EndoPAT and FMD using brachial artery ultrasound were assessed after 5 min of forearm ischemia in 20 younger (18–40 yr old; 29 ± 4 yr) and 20 older (60–75 yr old; 65 ± 4 yr) healthy adult men. Higher values of both FMD and RHI indicate better endothelial function. Endothelial function assessed via brachial artery FMD was lower in older (4.8 ± 2.1%), compared with younger (7.5 ± 1.6%) men (P < 0.001). In contrast, the RHI assessed via PAT was greater in older (2.2 ± 0.6), compared with younger (1.8 ± 0.5) men (P = 0.014). FMD and RHI were not correlated (r = –0.15; P = 0.35). We conclude that PAT may not be an appropriate measure to evaluate age-associated changes in endothelial function.

NEW & NOTEWORTHY Microvessel endothelial function assessed via finger plethysmography may not reflect age-associated reductions in large artery endothelial function assessed via brachial artery flow-mediated dilation.

INTRODUCTION

Since its initial description in 1989 (1), reactive hyperemia in response to limb ischemia has been developed as a popular research tool for investigating vascular endothelial function (2). Over the intervening three decades, myriad research has identified that a reduction in the vasodilator response signifies endothelial dysfunction (3–5), one of the earliest detectable stages in the progression of atherosclerotic cardiovascular disease (6, 7). Importantly, it has also been demonstrated repeatedly that brachial artery flow-mediated dilation (FMD) is a clinically relevant, noninvasive measure of coronary artery function (8–11). Subsequently, brachial artery endothelium-dependent FMD is considered a biomarker of endothelial function (7), and its use is widespread in research. However, despite the utility of brachial artery FMD in research laboratories, it has yet to be adopted in the clinic because of the technical requirements that can lead to poor reproducibility (12–14).

Recently, however, growing interest around the study of microvascular (rather than macrovascular) function, along with a desire for less technically demanding procedures, has led to the development of alternative methodologies for studying vascular function. Included among these alternative measures is peripheral artery tonometry (PAT), which uses finger plethysmography to measure the change in pulse wave amplitude in response to reactive hyperemia (reactive hyperemia index; RHI) (15). In a cohort of patients with abnormal coronary artery microvascular endothelial function, RHI was reduced compared with individuals with normal coronary artery microvascular function (16), indicating that RHI is a potential noninvasive test of coronary artery function. Previous reports have demonstrated a significant positive relationship between brachial artery FMD and RHI in relatively young and healthy men and women (15, 17).

However, vascular aging is an important component of cardiovascular disease risk in middle-aged and older adults (18), and a number of studies have demonstrated age-related reductions in vascular endothelial function assessed via brachial artery FMD (18–23). Similar to findings using brachial artery FMD, invasive studies of microvascular function have demonstrated age-related reductions in microvascular function (24–28). Despite the previous reports of age-related reductions in vascular endothelial function, data utilizing PAT to examine vascular function in aging is lacking, and to our knowledge, brachial artery FMD and PAT have yet to be compared in primary aging.

The purpose of this study was to compare brachial artery FMD and RHI to study macrovascular and microvascular function in aging men. It is widely accepted that vascular endothelial function is reduced with age (18–23); therefore, we hypothesized that older male adults would have reduced brachial artery FMD and RHI. Because previous studies have described a relationship between FMD and RHI (15, 17), we also hypothesized that these measures would be correlated in aging men.

METHODS

The data reported herein were baseline data collected as a part of an ongoing registered clinical trial (ClinicalTrials.gov identifier: NCT02758431). All participants provided written and verbal consent before participating. The Colorado Multiple Institution Review Board approved all study protocols and procedures, and they conformed to the provisions of the Declaration of Helsinki.

Study Participants

Forty healthy male participants aged 18–40 or 60–75 yr were recruited from the Denver, CO, metropolitan area. Participants in the present cohort are part of a larger study designed to examine the influence of testosterone on cardiovascular health and, therefore, only males are included in the present analysis. All participants were nonobese, nonsmoking, and free from cardiovascular, renal, liver, or respiratory disease, as assessed by medical history, physical exam, standard blood chemistries, and EKG at rest and during a graded exercise treadmill test to volitional fatigue. Total body composition (total fat mass, fat mass, and fat-free mass) was determined via dual X-ray absorptiometry (Delphi-W, Hologic). Hip and waist circumferences were measured by a trained technician in triplicate, and the average of those measures are reported here. Seated blood pressure was measured via oscillometric assessment following ≥10 min of seated rest (Carescape V100, GE Medical Systems). Blood pressure was assessed in triplicate on both arms, and the higher of the two averages is reported here. No participants were taking medications with actions on the cardiovascular system, and the participants abstained from taking antioxidant supplements, aspirin, or other NSAIDs for at least 4 wk. All study visits and measurements were conducted at the University of Colorado, Denver Colorado Clinical and Translational Sciences Institute Clinical and Translation Research Center.

Study Visit

Participants were required to fast and abstain from caffeine and alcohol for ≥4 h and exercise for ≥20 h before the start of the study. All experiments were conducted in the morning between 0800 and 1200. Upon arrival, participant height and weight were measured, and a venous blood sample was collected. Next, an infusion of isotonic saline was started. Following an initial bolus of 100 mL given at 5 mL/min for 20 min, the infusion rate was maintained at 1.7 mL/min until testing was complete. During the infusion, participants underwent measures of PAT followed by brachial artery FMD (described below). Participants rested in supine for ≥30 min before measurements, and at least 20 min separated the PAT and brachial artery FMD measures. Room temperature was maintained between 70 and 74°F.

Measurements

PAT.

Finger plethysmography (EndoPAT 2000, Itamar Medical) was used to measure changes in pulse wave amplitude in response to reactive hyperemia induced via forearm ischemia. Ischemia was produced by inflating a blood pressure cuff around the forearm to 250 mmHg and was maintained for 5 min. Pneumatic cuffs were placed on the index finger of each hand and recorded changes of finger volume with each pulse wave continuously before, during, and after cuff deflation. The data were analyzed using the automated algorithm built into the EndoPAT software. RHI is the ratio of the average pulse wave amplitude during reactive hyperemia to the average pulse wave amplitude during the baseline period in the occluded arm relative to the same ratio in the nonoccluded arm. The coefficient of variation measured in 10 individuals for RHI is 6.8%.

Brachial artery ultrasound.

Brachial artery diameter and blood flow velocity measurements were acquired using Doppler ultrasound (Vivid I, GE) using a multifrequency linear-array transducer, as previously described (29, 30). Briefly, a blood pressure cuff was placed around the forearm and brachial artery images were acquired ∼3–6 cm above the antecubital fossa. The ultrasound probe was held in place with a stereotactic clamp to ensure the location of the arterial segment remained constant with involuntary movement, and the isonation angle was maintained at ≤60°. After obtaining concurrent measures of baseline brachial artery diameter and blood flow velocity, reactive hyperemia was produced by inflating the blood pressure cuff to 250 mmHg to produce arterial occlusion. Occlusion was maintained for 5 min, followed by rapid deflation. Prior to the release of arterial occlusion, Doppler blood flow velocity was acquired and recorded until ∼15 s after the release of the occlusion. B-mode ultrasound brachial artery diameter images were recorded continuously for 2 min. Brachial artery diameter and blood flow velocity were analyzed using commercially available software (Vascular Analysis Tools 5.5.1, Medical Imaging Applications). Baseline blood flow velocity was reported as the average of at least five cardiac cycles, and hyperemic blood flow velocity was reported as the first complete velocity envelope following deflation of the cuff. This velocity envelope was also used to determine the peak velocity-time integral (VTI), as previously described (31, 32). All images were coded by number, blinded to age, and analyzed by the same investigator (K.L.M.). All procedures conformed with recently published guidelines for assessing FMD in humans (33). The coefficient of variation of brachial artery FMD (%) from our laboratory is 1.5% (34).

Data and Statistical Analysis

To determine whether differences in arterial diameter between younger and older males necessitated allometric scaling of FMD, a linear regression analysis was performed to determine the slope and 95% confidence intervals (95% CI) of the correlation between the natural log of peak diameter (lnD_peak) and baseline diameter (lnD_base) across all participants. Allometric scaling of FMD has been recommended when the unstandardized β-coefficient deviates from 1 and/or the 95% CI has an upper limit <1, as this indicates peak diameter is not increased as a constant proportion of resting diameter (35). When indicated by these criteria, we allometrically scaled FMD by using a linear mixed model with the difference of lnD_peak and lnD_base (lnD_diff) as the dependent variable, group (younger, older) as a fixed factor, and lnD_base as the covariate. Fisher’s least significant difference test was used for pairwise post hoc comparisons, as recommended (36).

The statistical approaches reported here were informed by recent guidelines for statistical reporting of cardiovascular research (37). Data were tested for normality using the Shapiro-Wilk normality test. Participant characteristics, RHI, and FMD were analyzed using Student’s t-test; in the case of nonnormally distributed data, the Mann-Whitney rank sum test was used. The relationship between age and measures of vascular function were tested using Pearson product-moment correlations. Statistical significance was set at P < 0.05, and the results are expressed as means ± SD. Nonnormally distributed data are presented as median (interquartile range). Data analysis was performed with IBM SPSS Statistics, version 26.0 and GraphPad Prism, version 8.4.1.

RESULTS

Participant characteristics are reported in Table 1 Older participants were significantly heavier and had higher BMI, as well as greater percent body fat, waist circumference, waist-to-hip ratio, as well as higher blood pressure (all P < 0.01). However, younger and older participants had similar total and LDL cholesterol and triglycerides.

Table 1.

Baseline characteristics of participants

Variable	Younger (n = 20)	Older (n = 20)	P Value
Age, yr	29 ± 4	65 ± 4	<0.001
Height, cm	180 ± 7	180 ± 7	0.709
Body mass, kg	78.6 ± 9.0	87.8 ± 12.0	0.009
BMI, kg/m2	24.2 ± 2.5	27.3 ± 3.5	0.003
Body fat, %	22.5 ± 5.5	29.4 ± 6.0	<0.001
Waist circumference, cm	87 ± 7	102 ± 8.6	<0.001
WHR	0.89 ± 0.06	0.94 ± 0.05	0.005
Systolic BP, mmHg	118 ± 9	127 ± 8	0.012
Diastolic BP, mmHg	73 ± 5	82 ± 7	<0.001
Mean BP, mmHg	88 ± 6	97 ± 7	<0.001
Total cholesterol, mg/dL	159 ± 27	170 ± 30	0.236
LDL cholesterol, mg/dLa	59 (45-102)	66 (56-73)	0.952
Triglycerides, mg/dL	79 ± 31	88 ± 35	0.410

Data were analyzed using Student's t-test and are expressed as means ± SD, unless otherwise noted. BMI, body mass index; BP, blood pressure; LDL, low-density lipoprotein; WHR, waist-to-hip ratio.

^aData were analyzed using the Mann-Whitney rank sum test and are expressed as median (interquartile range).

Supine measures of systolic and diastolic blood pressures and mean arterial pressure were higher in older participants (all P < 0.01); however, pulse pressure, heart rate, baseline blood flow velocity, and hyperemic blood flow velocity were similar between age groups (Table 2). As expected, blood flow velocity increased markedly from baseline to hyperemia (baseline = 6.0 ± 2.2, hyperemia = 49.8 ± 17.9 cm/s; P < 0.001), and this increase was similar between age groups (younger = Δ43.8 ± 16.4, older = Δ43.8 ± 19.5; P = .988). Resting brachial artery diameter, assessed via ultrasound, was larger in older, compared with younger males (P = 0.008; Table 2). Brachial artery FMD, expressed as either the absolute (Table 2) or percent change (Fig. 1) in brachial artery diameter from resting diameter, was greater in younger participants (P < 0.001). Allometric scaling of the diameter change did not alter these results (P = 0.001; Table 2). Similarly, adjusting FMD for peak hyperemic blood flow velocity did not abolish the difference in FMD (younger = 0.016 ± 0.004, older = 0.011 ± 0.006 s−¹; P = 0.003). Older males had greater RHI than younger males (P = 0.014; Fig. 2); natural log transforming RHI (lnRHI) did not change these results (P = 0.02; Table 2).

Table 2.

Vascular and hemodynamic parameters

Variable	Younger (n = 20)	Older (n = 20)	P Value
Supine SBP, mmHg	124 ± 9	135 ± 13	0.004
Supine DBP, mmHg	70 ± 9	80 ± 8	<0.001
Supine MAP, mmHg	88 ± 8	99 ± 9	<0.001
Pulse pressure, mmHg	54 ± 7	55 ± 8	0.793
Heart Rate, beats/min	54 ± 8	57 ± 5	0.150
Baseline brachial artery diameter, mm	4.13 ± 0.49	4.57 ± 0.52	0.008
Absolute FMD, mm	0.32 ± 0.06	0.22 ± 0.09	<0.001
Allometrically Scaled FMD, %	7.6 ± 1.8	5.2 ± 1.8	0.001
Baseline blood flow velocity, cm/s	8.2 ± 2.7	7.5 ± 1.5	0.348
Hyperemic blood flow velocity, cm/s	50.2 ± 15.8	49.4 ± 19.7	0.892
Peak VTI, cm	87.4 ± 18.1	84.8 ± 21.8	0.682
lnRHI, AU	0.53 ± 0.26	0.76 ± 0.29	0.012

Data were analyzed using Student's t-test and are expressed as means ± SD DBP, diastolic blood pressure; FMD; flow-mediated dilation; lnRHI, natural log transforming RHI; MAP, mean arterial pressure; SBP, systolic blood pressure; VTI, velocity-time integral.

Figure 1.

Brachial artery flow-mediated dilation (FMD) in younger (open bars) and older (solid bars) men. Older men had lower FMD compared to younger men. Data were analyzed using Student's t-test and are expressed as means ± SD with individual data points.

Figure 2.

Reactive hyperemia index (RHI) in younger (open bars) and older (solid bars) men assessed via peripheral artery tonometry (PAT). Younger men had attenuated RHI compared to older men. Data were analyzed using Student's t-test and are expressed as means ± SD with individual data points.

Correlations between brachial artery FMD, peak VTI, RHI, and age are displayed in Table 3. Brachial artery FMD (r = −0.56; P < 0.01) and allometrically scaled FMD (r = −0.47; P < 0.01) had significant negative correlations with age. In contrast, RHI (r = 0.32; P = 0.04) and lnRHI (r = 0.31; P = 0.05) were positively correlated with age. Neither measure of brachial artery FMD was related to RHI or lnRHI (P > 0.05). Although the relation between VTI and age was not significant (r = −0.13; P = 0.42), VTI was correlated with FMD (r = 0.48; P < 0.01), and scaled FMD (r = 0.45; P < 0.01), but not RHI (r = 0.18; P = 0.27) or lnRHI (r = 0.16; P = 0.32).

Table 3.

Pearson product-moment correlation coefficients among different measures of vascular function

	FMD	Scaled FMD	Peak VTI	RHI	lnRHI
Age	−0.56 (<0.01)	−0.47 (<0.01)	−0.13 (0.42)	0.32 (0.04)	0.31 (0.05)
FMD	–	0.83 (<0.01)	0.48 (<0.01)	−0.15 (0.35)	−0.17 (0.28)
Scaled FMD	–	–	0.45 (<0.01)	0.01 (0.97)	−0.03 (0.84)
Peak VTI	–	–	–	0.18 (0.27)	0.16 (0.32)
RHI	–	–	–	–	0.98 (<0.01)

Data were analyzed using Pearson product-moment correlations and are expressed as correlation coefficient (P value).

FMD, flow-mediated dilation; lnRHI, natural log transformed reactive hyperemia index; RHI, reactive hyperemia index; VTI, velocity-time integral.

Bold font indicates a significant association.

On a preliminary basis, participants were divided into those with a “normal” EndoScore (RHI >1.67) and compared to those with an “abnormal” EndoScore (i.e., endothelial dysfunction) (RHI ≤1.67) per the manufacturer’s recommendations. Twelve men had an RHI ≤1.67 (age = 41 ± 19 yr) compared with 28 with an RHI >1.67 (age = 50 ± 19 yr; P = 0.14). As expected, RHI was significantly lower in men with an “abnormal” EndoScore (1.36 ± 0.20 AU) compared with men with a “normal” EndoScore (2.26 ± 0.49 AU; P < 0.001). However, there was no significant difference in brachial artery FMD when participants were divided on the basis of EndoScore (“abnormal” = 7.02 ± 2.20% vs. “normal” = 5.83 ± 2.3%; P = 0.13).

DISCUSSION

Consistent with previous findings, the current data indicate a reduction in endothelial function with aging (18–23). However, the reduction in vascular endothelial function in older healthy men was only observed when assessed in the large conduit arteries using brachial artery FMD, but not in the microvasculature when assessed via finger plethysmography. In fact, microvascular function, as indicated by RHI using PAT was higher in older adults. In contrast to previous findings (15, 17), the present data indicate no relation between macrovascular and microvascular function, as assessed using PAT.

The contrasting findings between the present study and previous studies (15, 17), which have shown a correlation between RHI and FMD, can be explained by several important differences, including study population and methodology. For example, previous studies (15, 17) included both male and female participants, whereas the present study only included men. Menopausal status has a significant influence on vascular function (22, 29) and neither of these previous studies indicated the menopausal status of the women in their cohorts. Further, Kuvin et al. (15) included participants who were being evaluated for chest pain. Thus, it is possible that RHI and brachial artery FMD are correlated in conditions associated with endothelial dysfunction. Finally, methodological differences may account for these discrepancies, such as the use of upper, rather than lower, arm occlusion to induce reactive hyperemia (15) and calculating FMD from diameters measured at a predetermined time window after cuff deflation (i.e., 45 s) rather than from continuous recordings as in the present study (17).

We also examined the differences in brachial artery FMD when participants were divided by an abnormal (RHI ≤1.67) versus normal (RHI >1.67) EndoScore, rather than divided by age. Although differences in brachial artery FMD were not significantly different (P = 0.13), the mean value was ∼1.2% greater in the group with an abnormal EndoScore, compared with those with a normal EndoScore. Importantly, this was true despite the fact that age was similar between these groups. While both PAT (16) and brachial artery FMD (8–11) are related to coronary artery function, the disagreement between RHI and brachial artery FMD raise concerns about using these measures interchangeably. It should be noted that one recent study using updated guidelines for measuring brachial artery FMD (38, 39) demonstrated a strong correlation between FMD and ACh-induced coronary artery vasodilation (r = 0.77), demonstrating the clinical utility of brachial artery FMD for assessing coronary artery disease risk (9).

Although a number of reports have established that both brachial artery FMD and RHI are lower in the presence of cardiovascular risk factors (40–43), the present study is in agreement with a large cohort study of middle-aged adults that suggests no relation between measures of FMD and RHI (44). There are several possible explanations for the differences observed between brachial artery FMD and RHI. The automated EndoPAT analysis uses a fixed time frame during hyperemia to determine RHI, which may not be appropriate in aging. A careful study of brachial artery FMD responses in aging indicates that peak responses are delayed in older, compared with younger adults and that using an inappropriate time frame obscures differences that exist between groups (45).

It is also possible that the higher RHI observed in the present study reflect an adaptation of the microvasculature in aging to offset the reduction in conduit artery function. However, brachial artery blood flow peak VTI post-cuff occlusion, a well-established measure of microvascular function (31, 32), was not different between groups and was not correlated with RHI. Similarly, a number of studies using invasive methodology to study cutaneous microvascular function have reported a reduction in microvascular function with aging (24–28). For example, Minson et al. (28) used laser-Doppler flowmetry to demonstrate that older adults (69–84 yr old) have an impaired cutaneous vasodilatory response to local heating compared to younger adults (18–24 yr old). Importantly, by inhibiting nitric oxide synthase via local infusion of N^G-nitro-l-arginine methyl ester (l-NAME), these investigators determined that the nitric oxide contribution to cutaneous vasodilation is reduced in older adults (28). Because brachial artery FMD and PAT are considered to represent endothelium-depended dilation, which depends partly, at least, on nitric oxide bioavailability (46–49), it seems unlikely that macrovascular and microvascular function would exhibit these differential patterns in aging.

Although the cutaneous circulation has been identified as a model for generalized microvascular dysfunction (50–52), complexities in the regulation of cutaneous vascular tone add difficulty in interpreting these data. Blood vessels in most areas of the skin are similar to other vascular beds; however, acral skin of the fingers and palms contain a high proportion of arteriovenous anastomoses, and vascular tone is primarily regulated by sympathetic adrenergic nerves (53). While nitric oxide plays a minimal role in regulating resting digital blood flow (54, 55), one study using an infusion of l-NAME to inhibit nitric oxide production demonstrated that ∼50% of the reactive hyperemia response in the fingertip assessed via PAT is mediated by nitric oxide (49). It remains unknown, however, whether the remaining 50% of the hyperemic response is regulated by sympathetic adrenergic nerves or some other, yet unidentified, mechanism. It is possible that age-related changes in sympathetic activity may account for the differential responses between brachial artery FMD and PAT observed in the present study, and future studies using adrenergic blockade may provide more insight into this potential mechanism.

Instead, we propose that the differences between measures of brachial artery FMD and PAT presented herein indicate that PAT may not be suitable for assessing age-related endothelial dysfunction in men. Although previous literature has demonstrated relatively low variability between repeated measures of RHI over several weeks (56), other data indicate that PAT was unable to detect acute changes in endothelial function induced by cigarette smoking or hyperglycemia (56). Additionally, these data question the utility of PAT as a less technically demanding surrogate measure of endothelial function that can replace the use of brachial artery FMD. Previous cross-sectional studies in middle-aged adults have demonstrated conflicting relations between FMD and RHI, with some reports indicating a positive relation between the two measures (15, 17, 57), while others demonstrated no relation (43, 44, 58).

Limitations

In the present study, EndoPAT was performed before brachial artery FMD. It is possible, therefore, that there may be an order effect in the present study; however, tests were separated by ≥20 min, and it is unlikely that an order effect would influence the outcomes of these tests differentially.

Additionally, manufacturer recommendations for the EndoPAT recommend using upper, rather than lower arm occlusion. We chose to use lower arm occlusion to maintain a consistent cuff placement, and, therefore, area of ischemia as the brachial artery FMD. Previous literature comparing the RHI response to lower and upper arm occlusion indicated no differences in the response (59). Therefore, it is unlikely that the use of lower arm occlusion had a significant effect on our findings.

At least one previous report has indicated that hypohydration may affect vascular function as measured by brachial artery FMD (60). We did not collect urine samples from participants on the day of testing and, thus, are unable to comment on hydration status. However, all participants received a 100-mL bolus infusion of normal saline before testing, and the infusion was maintained throughout testing, making hypohydration less likely. Additionally, all comparisons are within-subject with measures separated by ∼25 min, making it unlikely for hydration status to have changed much between measures.

Data presented here are a part of a larger study aimed at investigating the cardiovascular outcomes of low testosterone in aging males; therefore, we did not include females in the present study. Accordingly, these relations should be examined in older women, as vascular aging is largely affected by sex hormones and significant sex differences may exist.

Finally, our relatively small sample size may have affected our ability to detect correlations between these measures. However, the fact that brachial artery FMD and RHI had opposite relations to age make it unlikely that a larger sample would have yielded different results.

Conclusions

In conclusion, we demonstrate a reduction in endothelium-dependent vasodilation in older men, a phenomenon that was observed when assessed via brachial artery FMD (i.e., in the large conduit arteries), but not when assessed using PAT (i.e., in the microvasculature) in men. Although we recognize the possibility that RHI may be higher in older adults to compensate for a reduction in FMD, we speculate that these differences represent the limitations of PAT to accurately evaluate endothelial dysfunction that occurs with aging in men.

GRANTS

This research was supported by National Institutes of Health R01AG049762 and National Institutes of Health T32AG000279, Colorado Clinical and Translational Sciences Institute UL1 TR001082, Colorado Nutrition and Obesity Research Center P30 DK048520 and Eastern Colorado GRECC.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

B.L.S., K.L.H., and K.L.M. conceived and designed research; M.C.B., L.E.D., T.L.W., A.B., B.L.S., and K.L.M. performed experiments; M.C.B. and K.L.M. analyzed data; M.C.B., B.L.S., K.L.H., and K.L.M. interpreted results of experiments; M.C.B. prepared figures; M.C.B. drafted manuscript; M.C.B., L.E.D., T.L.W., A.B., B.L.S., K.L.H., and K.L.M. edited and revised manuscript; M.C.B., L.E.D., T.L.W., A.B., B.L.S., K.L.H., and K.L.M. approved final version of manuscript.