Cardiometabolic risk factors are associated with immune cell mitochondrial respiration in humans

Abstract

Modifiable cardiometabolic risk factors induce the release of proinflammatory cytokines and reactive oxygen species from circulating peripheral blood mononuclear cells (PBMCs), resulting in increased cardiovascular disease risk and compromised immune health. These changes may be driven by metabolic reprogramming of PBMCs, resulting in reduced mitochondrial respiration; however, this has not been fully tested. We aimed to determine the independent associations between cardiometabolic risk factors including BMI, blood pressure, fasting glucose, and plasma lipids with mitochondrial respiration in PBMCs isolated from generally healthy individuals (n = 21) across the adult lifespan (12 men/9 women; age, 56 ± 21 yr; age range, 22–78 yr; body mass index, 27.9 ± 5.7 kg/m²; blood pressure, 123 ± 16/72 ± 10 mmHg; glucose, 90 ± 14 mg/dL; low-density lipoprotein cholesterol (LDL-C), 111 ± 22 mg/dL; and high-density lipoprotein cholesterol (HDL-C), 62 ± 16 mg/dL). PBMCs were isolated from whole blood by density-dependent centrifugation and used to assess mitochondrial function by respirometry. Primary outcomes included basal and maximal oxygen consumption rate (OCR), which were subsequently used to determine spare respiratory capacity and OCR metabolic potential. After we corrected for systolic blood pressure (SBP), diastolic blood pressure (DBP), and blood glucose, LDL-C was negatively associated with maximal respiration (r = −0.56, P = 0.016), spare respiratory capacity (r = −0.58, P = 0.012), and OCR metabolic potential (r = −0.71, P = 0.0011). In addition, SBP was negatively associated with OCR metabolic potential (r = −0.62, P = 0.0056) after we corrected for DBP, blood glucose, and LDL-C. These data suggest a link between blood cholesterol, SBP, and mitochondrial health that may provide insight into how cardiometabolic risk factors contribute to impaired immune cell function.

NEW & NOTEWORTHY Independent of other cardiometabolic risk factors, low-density lipoprotein cholesterol, and systolic blood pressure were found to be negatively associated with several parameters of mitochondrial respiration in peripheral blood mononuclear cells of healthy adults. These data suggest that low-density lipoprotein cholesterol and systolic blood pressure may induce metabolic reprogramming of immune cells, contributing to increased cardiovascular disease risk and impaired immune health.

INTRODUCTION

Cardiometabolic risk factors including advancing age, hypertension, elevated body mass index (BMI) and fasting glucose, and dyslipidemia are associated with chronic inflammation (5, 10, 21, 32, 45, 48); however, the underlying mechanisms are incompletely understood. These risk factors not only impact cardiovascular function via an increased release of proinflammatory cytokines and reactive oxygen species (ROS) from circulating immune cells (i.e., peripheral blood mononuclear cells, PBMCs) (4, 8, 9, 15, 30, 33, 39, 42) but may also compromise immune cell metabolism (12, 29, 37, 54), making the immune system less capable of mounting an appropriate response to acute infection (37, 40). Accordingly, there is presently an urgent need to understand the relationship between cardiometabolic risk factors and immune cell function to develop more targeted therapies for improving both cardiovascular and immune health in humans.

The interaction between cardiometabolic risk factors and impaired immune cell function may be partially mediated by reduced mitochondrial respiration (11, 13, 44). In this regard, mitochondrial respiration is impaired in PBMCs of patients with early stage heart failure and is associated with elevated blood pressure and increased low-density lipoprotein cholesterol (LDL-C) concentration (29). In addition, mitochondrial dysfunction is associated with increased production of mitochondrial ROS and a greater release of proinflammatory cytokines (7, 54), contributing to oxidative stress and chronic, low-grade inflammation (11, 44). Whether these risk factors are similarly associated with reduced mitochondrial function in healthier adults, before the development of overt cardiometabolic disease, is an important question that remains to be determined. Therefore, the purpose of this study was to determine the associations between cardiometabolic risk factors and mitochondrial respiration in circulating immune cells from healthy individuals across the adult lifespan. We hypothesized that cardiometabolic factors including age, blood pressure, BMI, fasting glucose, and circulating lipids would be independently and negatively correlated with mitochondrial respiration and that these associations would occur independent of age.

METHODS

Subjects.

Twenty-one healthy adults between the ages of 18 and 79 yr were included in this study. Subjects were excluded if they were diagnosed with any chronic clinical disease or exhibited clinically abnormal blood chemistries indicative of abnormal renal, liver, thyroid, and adrenal function; past or present alcohol dependence or abuse; currently smoking; or pregnant or breastfeeding. All procedures were approved by the Institutional Review Board at the University of Delaware. The study rationale, procedures, risks, and benefits were explained to the subjects, and their written, informed consent was obtained before enrollment. Subjects were instructed to fast for ≥12 h and to refrain from prescription medications for 24 h before each study visit.

Cardiometabolic risk factors.

Seated brachial artery systolic (SBP) and diastolic (DBP) blood pressure were measured following 1 min of quiet rest in the nondominant arm using a validated semiautomated oscillometric sphygmomanometer (SunTech ADView 2, SunTech Medical) (35). Measurements were made in triplicate with 1 min of recovery in-between each measurement until three measurements were obtained that were within 5 mmHg of one another. These values were then averaged to determine resting SBP and DBP and were used to calculate brachial artery pulse pressure (SBP – DBP) and mean arterial pressure (MAP). BMI was calculated as body mass (kg)/height (m)². Fasting blood glucose and serum lipids were measured from a comprehensive metabolic panel and standard lipid panel, respectively, by a CLIA-certified clinical laboratory (Quest Diagnostics or LabCorp). The lipid panel included total blood cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and plasma triglycerides (TGs). LDL-C concentrations were calculated from the measured variables.

Peripheral blood mononuclear cell isolation.

PBMCs were isolated from whole blood collected in EDTA-coated Vacutainer tubes, as previously described (31). Briefly, whole blood was centrifuged for 20 min at 500 g and 23°C with the acceleration and brake set to low. The majority (~2/3) of plasma was replaced with room-temperature 1× phosphate-buffered saline (PBS) to increase extraction efficiency, and the sample containing red blood cells was carefully pipetted on top of 10 mL of room-temperature Histopaque-1077 (Sigma-Aldrich) and centrifuged at 400 g for 20 min at 23°C with the acceleration and brake set to low. PBMCs were extracted and transferred to a separate tube and washed twice in PBS for 10 min each at 400 g and 23°C. The PBMCs were then counted, and viability was measured with trypan blue using the Countess II-automated cell counter (Invitrogen, No. AMQAX1000).

Mitochondrial respiration.

Isolated PBMCs were seeded in a Seahorse XFp miniplate (Agilent) for assessment of whole cell respirometry. We combined the results from two separate studies that used slightly different protocols. Cells were either incubated overnight in 120 µL of RPMI-1640 medium (ATCC 30C2001) in XFp miniplate at a density of 400,000 or 200,000 live cells/well and then washed the following day with XF DMEM at pH 7.4 (Agilent, 103575-100) or seeded the same day in XF DMEM at a density of 200,000 live cells/well. Cell-seeding density and incubation time were assessed for their potential associations with mitochondrial respiration using Pearson correlations. Neither parameter explained the variance in respiration, therefore these factors were not included as covariates in subsequent statistical models; however, all mitochondrial respiration parameters were normalized to seeding density before statistical analysis and are presented as pmol O₂·min⁻¹·10⁻⁵ cells⁻¹.

Mitochondrial respiration was assessed using the Agilent Cell Mito Stress Test Kit (No. 103015-100) on a Seahorse XFp Analyzer (Agilent Technologies, Santa Clara, CA), which measures oxygen consumption rate (OCR) at basal and following three serial injections of oligomycin (1 µM), carbonyl cyanide-4(trifluoromethoxy)phenylhydrazone (FCCP, 2 µM), and rotenone and antimycin A (0.5 µM) to determine maximal respiration, spare respiratory capacity, coupling efficiency, and ATP-linked O₂ consumption (ATP_O2). Spare respiratory capacity is a measure of the cell’s capacity to increase O₂ consumption above basal respiration (2). Coupling efficiency is how well ATP_O2 is matched to basal respiration, and ATP_O2 is calculated by measuring the decrease in OCR following the inhibition of ATP synthase (2). We also calculated the OCR metabolic potential, which reflects the percent increase (from baseline) of OCR following the acute injection of FCCP into the respirometer and represents the ability of cells to increase energy production specifically through oxidative phosphorylation (1).

Statistical analyses.

Significant cardiometabolic correlates of mitochondrial respiration were identified using bivariate correlations and subsequently assessed for their independent associations with mitochondrial respiration using partial correlations while covariates constant. Bivariate and partial correlations were performed using GraphPad Prism (version 8.2.1) and R (version 3.6.1), respectively. Data are presented as Pearson’s correlation coefficients (R) with α set at P < 0.05 for all analyses.

RESULTS

Subject characteristics.

Subject characteristics are summarized in Table 1. Subjects were recruited across the adult lifespan (22–78 yr) and were generally healthy and free of chronic diseases. While as a group, all subject characteristics were within normal ranges, we observed the necessary variation in BMI, SBP and LDL-C to assess the independent association of these risk factors with mitochondrial function.

Table 1.

Participant characteristics

	Mean ± SD	Minimum-Maximum
Sex, men/women	12/9	-
Age, yr	56 ± 21	22-78
Height, cm	168.0 ± 12.6	149.5-196.2
Body mass, kg	79.1 ± 23.3	52.7-148.3
Body mass index, kg/m2	27.9 ± 5.7	21.9-40.8
SBP, mmHg	123 ± 16	100-152
DBP, mmHg	72 ± 10	58-89
PP, mmHg	51 ± 11	31-79
MAP, mmHg	89 ± 11	75-106
HR, beats/min	57 ± 8	40-71
Glucose, mg/dL	90 ± 14	61-131
TC, mg/dL	194 ± 31	143-256
HDL-C, mg/dL	62 ± 16	32-91
LDL-C, mg/dL	111 ± 22	73-161
TG, mg/dL	103 ± 60	38-315
Statins, n (%)	5 (24)
ACE inhibitor, n (%)	3 (14)
Angiotensin II receptor blocker, n (%)	1 (5)

Values are means ± SD and minimum-maximum values. Characteristics for all subjects (n = 21). BP, blood pressure; SBP, systolic blood pressure; DBP, diastolic blood pressure; PP, pulse pressure; MAP, mean arterial pressure; HR, heart rate; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; ACE, angiotensin-converting enzyme.

Cardiometabolic risk factors and mitochondrial respiration.

As reported in Table 2, maximal respiration (mean = 126.89 ± 82.31 pmol O₂·min⁻¹·10⁻⁵ cells⁻¹) was negatively associated with SBP (r = −0.51, P = 0.018), DBP (r = −0.46, P = 0.037), blood glucose (r = −0.43, P = 0.049), and LDL-C (r = −0.55, P = 0.010). Similarly, spare respiratory capacity (mean = 104.38 ± 74.27 pmol O₂·min⁻¹·10⁻⁵ cells⁻¹) was negatively associated with SBP (r = −0.52, P = 0.016), DBP (r = −0.44, P = 0.045), and LDL-C (r = −0.55, P = 0.0096). While nonsignificant, there was a strong association between spare respiratory capacity and blood glucose (r = −0.43, P = 0.054). Lastly, OCR metabolic potential (mean = 355.63 ± 141.55%) was negatively associated with LDL-C (r = −0.54, P = 0.011) and ATP_O2 (mean = 18.31 ± 13.72 pmol O₂·min⁻¹·10⁻⁵ cells⁻¹) was negatively associated with blood glucose (r = −0.45, P = 0.042). Basal respiration (mean = 22.50 ± 12.89 pmol O₂·min⁻¹·10⁻⁵ cells⁻¹) and coupling efficiency (mean = 81.68 ± 35.81%) were not associated with any of the cardiometabolic risk factors tested; therefore, no subsequent analyses were performed with these parameters.

Table 2.

Bivariate correlations between mitochondrial respiration and cardiometabolic risk factors

	Age	SBP	DBP	BMI	Glucose	TC	HDL-C	LDL-C
Basal respiration
r	−0.06	−0.25	−0.38	−0.13	−0.32	−0.12	0.20	−0.33
P value	0.78	0.27	0.087	0.59	0.16	0.59	0.39	0.14
Maximal respiration
r	−0.26	−0.51	−0.46	−0.16	−0.43	−0.23	0.31	−0.55
P value	0.25	0.018	0.037	0.50	0.049	0.31	0.17	0.010
Spare respiratory capacity
r	−0.28	−0.52	−0.44	−0.15	−0.43	−0.24	0.31	−0.55
P value	0.22	0.016	0.045	0.52	0.054	0.30	0.18	0.0096
ATPO2
r	0.05	−0.33	−0.42	−0.18	−0.45	−0.19	0.05	−0.32
P value	0.82	0.14	0.060	0.43	0.042	0.40	0.84	0.16
Coupling efficiency
r	0.18	−0.11	0.02	−0.08	−0.40	−0.24	−0.26	−0.18
P value	0.44	0.62	0.92	0.72	0.070	0.29	0.25	0.42
Metabolic potential
r	−0.22	−0.39	−0.21	0.02	−0.17	−0.28	0.26	−0.54
P value	0.34	0.079	0.36	0.93	0.47	0.21	0.25	0.011

Pearson’s correlation coefficient between each mitochondrial respiratory parameter and cardiometabolic risk factor. Abbreviations: SBP, systolic blood pressure; DBP, diastolic blood pressure; BMI, body mass index; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; ATP_O2, ATP-linked oxygen consumption. α was set at P < 0.05 (significant associations indicated by boldface).

After we corrected for covariates (e.g., SBP, DBP, and blood glucose), LDL-C was negatively associated with maximal respiration (Fig. 1A), spare respiratory capacity (Fig. 1B), and OCR metabolic potential (Fig. 1C) but was not associated with ATP_O2 (r = −0.13, P = 0.61). After we corrected for DBP, blood glucose, and LDL-C, SBP was also negatively associated with OCR metabolic potential (Fig. 1F), and there was a strong but nonsignificant negative association between SBP and maximal respiration (Fig. 1D) and spare respiratory capacity (Fig. 1E). There was no association between SBP and ATP_O2 (r = 0.03, P = 0.90). After we corrected for SBP, blood glucose, and LDL-C, DBP was no longer associated with maximal respiration (r = 0.01, P = 0.98), spare respiratory capacity (r = 0.06, P = 0.80), OCR metabolic potential (r = 0.40, P = 0.10), or ATP_O2 (r = −0.25, P = 0.32). Similarly, blood glucose was not associated with maximal respiration (r = −0.01, P = 0.97), spare respiratory capacity (r = 0.02, P = 0.93), OCR metabolic potential (r = 0.41, P = 0.09), or ATP_O2 (r = 29, P = 0.25) after we corrected for covariates (e.g., SBP, DBP, and LDL-C).

Fig. 1.

Cardiometabolic risk factors and mitochondrial respiration. Independent associations between low-density lipoprotein cholesterol (LDL-C) and systolic blood pressure (SBP) with maximal respiration (A and D), spare respiratory capacity (B and E), and oxygen consumption rate (OCR) metabolic potential (C and F). Statistical analyses based on partial correlation matrices with respiratory parameters adjusted for either subject SBP or LDL-C, diastolic blood pressure (DBP), and blood glucose. α was set at P < 0.05.

DISCUSSION

In the present study, we found that LDL-C is negatively associated with maximal oxygen consumption, spare respiratory capacity, and OCR metabolic potential of PBMCs, independent of age, blood pressure, BMI, and other blood lipids. Likewise, we observed near-significant associations between SBP with maximal respiration and spare respiratory capacity and a significant association with OCR metabolic potential. Collectively, these findings suggest a potentially important link between cardiometabolic risk factors and immune cell energy metabolism in healthy humans.

Our findings are also in agreement with previous findings in a group of adults that included patients with early stage heart failure (29). The negative associations of these risk factors with both maximal respiration and spare respiratory capacity suggest that these factors may interfere with the ability of immune cells to respond to increases in energy demand, such as during acute infection. In addition, the inverse association with OCR metabolic potential suggests that elevated risk factors may also shift immune cells toward more glycolytic energy production. Immune cell metabolism is complex and cell-type dependent. As demand for ATP increases during acute infection, some lymphocytes (e.g., T effector cells) rely more on glycolysis (55) and exhibit reduced oxidative metabolism as demand for oxygen exceeds its supply (40). Importantly, the dependence on glycolysis can be reversed by improving mitochondrial function, allowing these cells to maximize energy production through oxidative phosphorylation (40). Memory T cells, while inherently more oxidative than T effector cells, demonstrate improved survival in response to reexposure to an antigen when oxidative metabolism is increased, thus highlighting the importance of mitochondrial function in the regulation of adaptive immunity (34, 49). With regard to the innate immune system, monocytes also undergo metabolic reprogramming as they differentiate into either proinflammatory M1 monocytes or anti-inflammatory M2 monocytes (27). Predifferentiated monocytes and M2 monocytes are primarily aerobic, whereas M1 monocytes primarily use anaerobic metabolism (27). Collectively, the unique metabolic requirements of lymphocytes and monocytes implies an important role of oxidative metabolism in the cellular response to acute infection and/or inflammation. This is of particular importance to the present study as lymphocytes and monocytes make up the vast majority of PBMCs (26).

The mechanisms by which blood lipids influence immune cell metabolism requires further elucidation; however, one possibility is through oxidation of LDL-C. In isolated human macrophages, oxidized LDL-C induces mitochondrial dysfunction and promotes apoptosis via the formation of peroxidases, which decreases mitochondrial membrane potential (4b). Oxidized LDL-C also induces mitochondrial membrane dysfunction in human umbilical vein endothelial cells (53) and activates calcium-dependent mitochondrial pathways that result in apoptosis (51). In addition to oxidation of LDL-C, treatment of endothelial cells with LDL-C itself reduces ATP content and decreases the expression of genes that encode for respiratory complex proteins (19). Similarly, T cells incubated with LDL-C exhibit reduced mitochondrial mass and ATP production (36), and the loading of macrophages with free cholesterol decreases mitochondrial membrane potential, triggering apoptosis (56).

In addition to attenuating immune cell function directly, LDL-C-mediated reductions in immune cell respiration may also be deleterious to the cardiovascular system. The mitochondria are major sites of ROS generation (25, 44), which can be deleterious to the vasculature by inducing endothelial dysfunction (23), a risk factor for cardiovascular disease (24, 50). Because of their location within the blood, immune cells are in constant contact with endothelial cells and may release ROS, inflammatory cytokines, and other vasoactive molecules that attenuate vascular function and increased blood pressure (4, 9, 15, 33). In this regard, ameliorating mitochondrial-derived ROS with a mitochondrial-targeted antioxidant has been shown to restore endothelial function in older or hypertensive animals and healthy middle-aged/older adults (17, 20, 38). Thus, improving mitochondrial respiration in circulating immune cells may improve endothelial function and reduce blood pressure by reducing a major source of circulating ROS. Importantly, our results suggest that the effects of LDL-C on mitochondrial function occurs even at moderate concentrations and in healthy adults without overt CVD, potentially suggesting a need for early detection and management of LDL-C for maintenance of cardiovascular and immune health.

The relations between SBP and mitochondrial respiration may be explained by elevations in oxidative stress and inflammation, which have been shown to influence respiration (29) and are characteristic of individuals with hypertension (41). Healthy mitochondria are also important for maintaining vascular function, with increased oxidative stress leading to vascular endothelial dysfunction (16, 17, 38). In addition, various animal models of induced hypertension are associated with increased mtROS production, inflammation, impaired bioenergetics, and mitochondrial damage (13). While Li et al. (29) found that DBP was correlated with mitochondrial respiration in patients with heart failure, we found no association with DBP after correcting for other risk factors. This may be due to population differences or underlying effects of SBP and LDL-C. It should be noted that blood pressure was measured in triplicate during a single visit and may not be representative of true resting blood pressure in all individuals. The interaction between mitochondrial function and more definitive measures of blood pressure (e.g., 24-h ambulatory blood pressure) should be explored in future studies.

We did not see any associations between basal respiration or coupling efficiency and any of the cardiometabolic risk factors that we studied. While ATP_O2 was initially negatively correlated with blood glucose, this relationship was likely mediated by SBP, DBP, and LDL-C. Basal respiration is likely more representative of quiescent circulating immune cells, such as naïve T cells or macrophages, which are only activated when in contact with an antigen or during acute inflammation (18, 27). Thus, spare respiratory capacity and maximal respiration may be more important indexes of immune cell function, as these measures reflect the cells’ capacity to increase energy metabolism which is necessary when responding to acute infection or inflammation (18, 27). Both ATP_O2 and coupling efficiency reflect how much of the basal oxygen consumption is used to facilitate ATP production. Because our results suggest that LDL-C plays a greater role in inhibiting maximal respiration, a more appropriate experiment may be to investigate the coupling of oxygen consumption to ATP production during maximal respiration.

Because of the nature of partial correlations and the limitations of retrospective analyses, we cannot determine the causal relationship between these risk factors in immune cell respiration. We believe our results may have important implications for immune cell function in adults with cardiometabolic risk factors, such as hypertension or dyslipidemia, making these groups more vulnerable to secondary infection (6) and attenuated immune system function (3, 6, 28). However, we cannot rule out the possibility that reduced mitochondrial respiration contributes to an increased serum LDL-C concentration or blood pressure; however, the effects of inflammation on LDL-C concentration are equivocal with increases in inflammation associated with both increases and decreases in hepatic LDL receptor function, resulting in hypo- and hypercholesterolemia, respectively (14, 43). Likewise, we are unable to determine whether impaired immune cell mitochondrial respiration contributes to increased SBP or reflects the mitochondrial respiration of vascular tissue. Thus, our results should be interpreted as preliminary findings to guide future exploration of these hypotheses.

A limitation of this study is that we did not directly measure physical activity and only obtained self-reported physical activity in a subset of subjects; therefore, we were unable to definitively assess how physical activity may modulate these associations. Hedges et al. (22) recently demonstrated an increase in human skeletal muscle mitochondrial respiration following a 12-wk exercise training program; however, a similar improvement was not observed in PBMCs, suggesting that physical activity primarily affects tissues that are more metabolically active during exercise. Nevertheless, increased physical activity has been shown to protect the vasculature from the damaging effects of elevated blood LDL-C concentrations and lowers systolic blood pressure (4a, 52); therefore, it is worth investigating the potential influence of physical activity on the association between of these risk factors and immune cell respiration in future studies. Additionally, we were unable to assess the role of racial and ethnic differences on the associations between cardiometabolic risk factors and mitochondrial function because of a lack of diversity in our sample. Future studies should investigate these associations in a more racially diverse population that includes underrepresented minorities, as these groups are often at elevated risk for cardiometabolic diseases. Finally, our study did not control for the use of statins or cardiovascular acting medications because of the relatively small number of subjects taking these drugs; however, the influence of these drugs on immune cell respiration is worth consideration in future studies.

Conclusions and future directions.

In summary, we have demonstrated an independent association between LDL-C, SBP, and mitochondrial respiration in circulating immune cells after controlling for cardiometabolic risk factors. Collectively, our findings suggest a possible mechanism linking aberrant blood lipids and SBP to impaired immune system health. These results may have important implications for individuals with more severe cardiometabolic disease, as these factors may compromise immune system’s ability to respond to an acute infection. Moreover, even modest increases in LDL-C as observed in the present study may contribute to cardiometabolic disease risk by promoting the production of mitochondrial-derived reactive oxygen species and/or inflammatory cytokines. Based on this exploratory analysis, future studies should aim to identify the mechanisms by which LDL-C and SBP affect mitochondrial respiration and characterize the influence of blood lipids on mitochondrial function in patients with cardiovascular disease and chronic metabolic disorders (e.g., diabetes), as well as those with and without secondary infections and following treatment.

GRANTS

This work was supported by National Institutes of Health Grants P20GM113125 and K01AG054731.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

T.M.D., E.R.M., and C.R.M. conceived and designed research; T.M.D., F.S., and J.C.H. performed experiments; T.M.D. analyzed data; T.M.D. and C.R.M. interpreted results of experiments; T.M.D. prepared figures; T.M.D. drafted manuscript; T.M.D., E.R.M., F.S., J.C.H., and C.R.M. edited and revised manuscript; C.R.M. approved final version of manuscript.