Circulating Follistatin in Patients with Chronic Kidney Disease: Implications for Muscle Strength, Bone Mineral Density, Inflammation, and Survival

Abstract

Summary

Background and objectives

Follistatin mediates muscle growth and bone mineralization. At present, it is unknown whether circulating follistatin levels are altered in chronic kidney disease (CKD) or links to CKD risk factors and outcomes.

Design, setting, participants, & measurements

Plasma follistatin levels were cross-sectionally analyzed in relation to protein-energy wasting (PEW), handgrip strength (HGS), bone mineral density (BMD), and inflammatory markers in 280 CKD stage 5 patients, 32 CKD stage 4 patients, 16 CKD stage 3 patients, and 32 control subjects. In CKD stage 5 patients survival was prospectively investigated during a follow-up period of up to 5 years.

Results

The plasma follistatin concentration was not higher in CKD stage 5 patients than in other CKD stages or controls. In CKD stage 5 patients, circulating follistatin positively correlated with age, high-sensitivity C-reactive protein (hsCRP), and IL-6; negatively correlated with HGS, serum creatinine, and BMD; and was increased in patients with PEW. In a multivariate logistic regression model, lower HGS, lower BMD, and higher hsCRP independently correlated with higher follistatin levels. In a Cox regression model, follistatin levels were not associated with all-cause mortality.

Conclusions

Circulating follistatin levels were neither elevated nor predicted outcome in patients with CKD. However, increased follistatin levels occurred together with increased inflammation, reduced muscle strength, and low BMD, suggesting an involvement of a mechanism including follistatin in the uremic wasting process.

Introduction

Muscle wasting is an important feature in the syndrome of protein-energy wasting (PEW) present in patients with chronic kidney disease (CKD), which undoubtedly contributes to the increased mortality of this patient population (1–3). Although multiple catabolic or anabolic alterations have been shown to contribute to the mechanisms of CKD-related muscle wasting, the molecular pathways have not been fully elucidated.

Follistatin is an extracellular glycoprotein (the most abundant form with a size of 35 kD) (4) ubiquitously expressed and originally identified as an inhibitor of pituitary follicle stimulating hormone secretion (5). Follistatin exerts its actions by neutralizing biologic activities of members of the TGF-β superfamily (6–9). The best-characterized effect of follistatin pertains to the stimulation of muscle growth by direct inhibition of myostatin (6,8). Experimental evidence suggests that myostatin levels are overexpressed in uremic cachexia (10), and that strategies to correct uremic sarcopenia may be mediated, at least in part, by inhibition of myostatin expression (10,11). It is of interest that recent experimental studies suggest a potential therapeutic role for follistatin gene therapy in inducing muscle growth and muscle strength (7,8,12,13). Although less substantiated, emerging evidence also indicates that follistatin is involved in bone metabolism and development as well as in the inflammatory response by direct inhibition of activins and bone morphogenic proteins (9,14).

Because it is currently unknown if follistatin levels are dysregulated in the uremic milieu and involved in the muscle-wasting/bone demineralization processes, such study would be of interest because derangements at this level may be identified as targets for novel therapies. Therefore, we aimed at studying follistatin levels in a well characterized cohort of incident dialysis patients sampled close to initiation of dialysis therapy, with special emphasis on the muscle and bone axis. Additionally, comparisons were done with cohorts of CKD stage 3 to 4 patients and control subjects.

Materials and Methods

Subjects and Study Design

The local ethics committee of Karolinska Institutet at Karolinska University Hospital at Huddinge approved the study protocol, and informed consent was obtained from each individual. The main analysis of this study was conducted in consecutive patients from an ongoing prospective cohort study including incident patients who were close to beginning dialysis replacement therapy at the Renal Clinic of the Karolinska University Hospital Huddinge, Stockholm, Sweden (15). In the study presented here, post hoc analyses were performed in 280 CKD stage 5 patients (116 women, 41%) with a median age of 56 (range of 10th to 90th percentile, 35 to 68) years and a median GFR of 6 (4 to 9) ml/min per 1.73 m². Exclusion criteria for this patient population were age below 18 or above 70 years old, HIV or hepatitis B/C, signs of acute infection, and unwillingness to participate. Eighty-one patients had already started dialysis treatment at blood sampling, with a median time in dialysis of 8.5 (−1.5 to 43.5) days. However, after correction for age, follistatin levels did not differ in these patients (data not shown). Comorbidities at baseline were based on medical records. Diabetes was present in 90 (32%) patients. Cardiovascular disease (CVD; defined as cardiac, cerebrovascular [stroke], or peripheral vascular disease) was present in 106 (38%) patients. Most patients were on antihypertensive medications as well as phosphate and potassium binders; diuretics; and vitamin B, C, and D supplementation in accordance with clinical practice. Survival, censored for transplantation, was recorded from the day of examination and for a follow-up time of up to 5 years. For comparison purposes, we included 32 age- and sex-matched CKD stage 4 patients (17 women, median age of 58 [31 to 72] years, median estimated glomerular filtration [eGFR] of 20 [15 to 29] ml/min per 1.73 m²), 16 CKD stage 3 patients (4 women, median age of 65 [45 to 72] years, median eGFR of 50 [31 to 58] ml/min per 1.73 m²), and 32 controls (15 women, median age 57 [39 to 66] years, eGFR > 60 ml/min per 1.73 m²). Control subjects were recruited from a population-based group cohort of age- and sex-matched but otherwise randomly selected individuals in the Stockholm region. In the CKD stage 5 patients, GFR was calculated by the mean of urea and creatinine clearances, whereas GFR in CKD stages 3 to 4 patients and controls was estimated by means of the four-variable Modification of Diet in Renal Disease study equation (16).

Anthropometric Evaluation, Nutritional Status, and Bone Mineral Density

Subjective global nutritional assessment (SGA) (17) was performed on the same occasion as blood sampling and used as a surrogate of PEW (defined as SGA ≥ 2). Handgrip strength (HGS) was measured with a Harpenden handgrip dynamometer (Yamar, Jackson, MI) in the dominant and nondominant arms. Each measurement was repeated 3 times in each arm, and the highest value for each arm was noted. For our analysis, we used the result obtained, normalized with measurements from healthy subjects. Total bone mineral density (BMD) was estimated by means of dual-energy x-ray absorptiometry using a DPX-L device (Lunar Corporation, Madison, WI). BMD was expressed as total BMD (g/cm²) of the head, arms, legs, trunk, hip, pelvis, and spine. These assessments were completed at the time of or within 1 week of blood sample collection.

Blood Sampling and Laboratory Analysis

After an overnight fast, plasma samples were taken and stored at −70°C if not analyzed immediately. Plasma follistatin (FS288, FS300, and FS315) was determined with an ELISA kit (R&D Systems Europe Ltd., Abingdon, United Kingdom). High-sensitivity CRP (hsCRP) was measured by the nephelometry method. IL-6 was quantified in serum by an immunometric assay on an Immulite analyzer according to the instructions of the manufacturer (Siemens Medical Solutions Diagnostics, Los Angeles, CA). Determinations of serum albumin (bromcresol purple), hemoglobin, intact parathyroid hormone, and serum creatinine were performed by routine procedures in the Department of Clinical Chemistry at Huddinge Hospital.

Statistical Analyses

Values are expressed as mean ± SD or median (range of 10th to 90th percentile) or percentage, as appropriate. Statistical significance was set at the level of P < 0.05. Comparisons between two groups were assessed with the nonparametric Mann–Whitney test for continuous variables and a χ² test for nominal variables. Differences among three groups were analyzed using the nonparametric Kruskal–Wallis test. Spearman rank correlation analysis was used to determine associations between follistatin and selected parameters.

Determinants of higher follistatin levels were evaluated with a multivariate logistic regression model. Multiple regression models were fitted to assess independent predictors of HGS and BMD. These models included the variables significantly associated with follistatin levels, HGS, and BMD in univariate analysis (P < 0.05 as cutoff value) or other variables with a documented causal relationship (in this case age and sex). The effect of follistatin levels in predicting all-cause mortality or CVD-related mortality was determined using the Cox regression hazard models. The proportionality assumptions were checked through visual inspection of the log of the incidence rates. Finally, restricted cubic spline graphs were used to evaluate systematic relationships between follistatin levels and mortality using STATA version 11.1 (Stata Corporation, College Station, TX).

Results

The median plasma follistatin level in the 280 CKD stage 5 patients was 1.57 (0.89 to 2.81) ng/ml, which was not significantly different from the median level in CKD stage 4 patients (1.62 [0.76 to 3.42] ng/ml), CKD stage 3 patients (1.75 [1.04 to 2.72] ng/ml), and controls (1.28 [1.15 to 2.14] ng/ml; P = 0.13) (Figure 1).

Figure 1.

Plasma follistatin concentrations in 32 control subjects, 16 CKD stage 3 patients, 32 CKD stage 4 patients, and 280 incident dialysis patients. GFR was calculated by the mean of urea and creatinine clearances in incident dialysis patients and estimated using the Modification of Diet in Renal Disease study equation in CKD stage 3 to 4 subjects and controls.

Baseline Characteristics

Baseline clinical characteristics of 280 incident dialysis patients (CKD stage 5) are summarized in Table 1 according to median follistatin values. The patients with follistatin levels above the median were older and had a higher prevalence of PEW and clinically manifest CVD. These patients also showed lower serum albumin, serum calcium, and HGS as well as reduced BMD and elevated levels of CRP and IL-6.

Table 1.

Clinical characteristics of 32 controls, 48 CKD stage 3 to 4 patients, and 280 incident dialysis patients grouped according to median follistatin concentration (1.57 ng/ml), with univariate Spearman correlations (ρ) between the investigated parameters and serum follistatin levels in the incident dialysis patients

Parameter	Controls (n = 32)	CKD Stage 3 to 4 (n = 48)	Incident Dialysis Patients		Pa	ρ (P)b
			Follistatin below Median (n = 140)	Follistatin above Median (n = 140)
General
age (years)	57 (39 to 66)	60 (38 to 72)	54 (34 to 67)	58 (36 to 68)	0.02	0.18 (0.003)
gender (% men)	53	56	59	59	1.00	–
diabetes mellitus (%)	0	8	29	35	0.31	–
CVD (%)	0	19	31	44	0.03	–
GFR (ml/min per 1.73 m2)	>60	24 (16 to 56)	6 (4 to 9)	6 (4 to 9)	0.9	−0.05 (0.47)
hemoglobin (g/L)	140 ± 10	128 ± 15	105 ± 15	103 ± 15	0.11	−0.10 (0.09)
Nutritional parameters
PEW (SGA ≥ 2, %)	0	0	24	38	0.01	–
serum albumin (g/L)	39.2 ± 2.2	37.9 ± 2.9	33.7 ± 6.9	32.2 ± 5.9	0.03	−0.14 (0.02)
body mass index (kg/m2)	25.2 ± 4.2	26.4 ± 5.1	24.8 ± 4.2	24.7 ± 4.5	0.8	−0.01 (0.95)
HGS (%)c,d	–	–	82 ± 22	72 ± 25	<0.001	−0.28 (<0.001)
serum creatinine (μmol/L)	72 ± 11	216 ± 109	755 ± 249	681 ± 234	0.01	−0.12 (0.04)
Inflammation parameters
hsCRP (mg/L)	0.8 (0.4 to 5.5)	2.0 (0.6 to 8.5)	3.0 (0.5 to 25.1)	8.4 (1.6 to 59.8)	<0.001	0.35 (<0.001)
IL-6 (pg/ml)	–	–	5.2 (1.8 to 16.7)	7.3 (2.9 to 22.2)	<0.001	0.26 (<0.001)
Bone-related markers
calcium (mmol/L)	2.32 ± 0.09	2.38 ± 0.16	2.53 ± 0.27	2.44 ± 0.26	0.009	−0.18 (0.003)
phosphate (mmol/L)	0.99 ± 0.16	1.26 ± 0.36	2.01 ± 0.56	1.90 ± 0.59	0.11	−0.08 (0.17)
intact PTH (ng/L)	36 ± 12	103 ± 77	255 ± 256	281 ± 273	0.27	0.09 (0.13)
BMD (g/cm2)b	1.19 ± 0.07	1.19 ± 0.11	1.17 ± 0.11	1.11 ± 0.13	<0.001	−0.30 (<0.001)

Data presented as mean ± SD, median (range of 10th to 90th percentile), or percentage. PTH, parathyroid hormone.

^aComparison with Mann–Whitney test for continuous variables or χ² test for nominal variables in incident dialysis patients.

^bSpearman rank univariate correlation test.

^cHGS (%) was normalized based on the measurements among the healthy subjects.

^dn = 251.

^en = 221.

Univariate and Multivariate Correlates

In incident dialysis patients, follistatin levels positively correlated with age (ρ = 0.18; P < 0.01), hsCRP (ρ = 0.35; P < 0.001), and IL-6 (ρ = 0.26; P < 0.001) and negatively correlated with HGS (ρ = −0.28; P < 0.001), serum creatinine (ρ = −0.12; P < 0.05), serum albumin (ρ = −0.14; P < 0.05), serum calcium (ρ = −0.18; P < 0.01), and BMD (ρ = −0.30; P < 0.001). Patients with PEW (SGA ≥ 2) had higher median follistatin levels than those without PEW (SGA 1; 1.81 [1.00 to 3.17] versus 1.49 [0.83 to 2.62] ng/ml, P < 0.01; Figure 2). A logistic regression model showed that lower HGS (odds ratio [OR] = 3.09, 95% confidence interval [CI]: 1.47 to 6.50, P < 0.01), lower BMD (OR = 2.91, 95% CI: 1.49 to 5.67, P < 0.01), and higher hsCRP (OR = 2.13, 95% CI: 1.08 to 4.17, P < 0.05) were independent predictors of higher circulating follistatin levels (Table 2). Multiple regression models were applied to test whether the associations between follistatin and HGS or BMD were confounded by other clinical characteristics or comorbidities. HGS was independently associated with follistatin (β = −4.1, SEM = 1.47, P < 0.01; Table 3) in a model containing age, sex, CVD, diabetes, and hsCRP. BMD was also independently associated with follistatin (β = −0.03, SEM = 0.01, P < 0.01; Table 4) in a model including age, sex, SGA, and phosphate levels.

Figure 2.

Plasma follistatin concentrations in relation to the nutritional status as assessed by SGA in incident dialysis patients. Nonwasted, SGA = 1 (n = 195); wasted, SGA ≥ 2 (n = 85).

Table 2.

Logistic regression model of factors predicting higher follistatin level (>1.57 ng/ml) in 197 incident dialysis patients

Parameter	OR	95% CI	P
Intercept			0.005
Age, >56 years old	1.05	0.53 to 2.09	0.89
Gender, female	0.78	0.39 to 1.57	0.49
CVD, presence	0.64	0.30 to 1.40	0.27
HGS, <75.9%	3.09	1.47 to 6.50	0.003
BMD, <1.14 g/cm2	2.91	1.49 to 5.67	0.002
Inflammation, hsCRP >10 mg/L	2.13	1.08 to 4.17	0.03

Whole-model pseudo-r² = 0.16. Age, HGS, and BMD were dichotomized according to the median value. Eighty-three patients in whom HGS and/or BMD by dual-energy x-ray absorptiometry was missing were excluded from analysis.

Table 3.

Multiple regression model of factors predicting HGS (per % of increase) in 251 incident dialysis patients

Parameter	Parameter Estimate	SEM	P
Intercept	119.1	6.40	<0.001
Age, >56 years	−0.545	0.12	<0.001
Gender, female	−9.734	2.60	<0.001
Diabetes, presence	−11.264	2.79	<0.001
CVD, presence	−9.532	3.02	0.002
Inflammation, hsCRP >10 mg/L	−5.598	2.81	0.05
Follistatin, ng/ml	−4.140	1.47	0.005

Whole-model pseudo-r² = 0.35. Age was dichotomized according to the median value. Twenty-nine patients in whom HGS was missing were not included in the model.

Table 4.

Multiple regression model predicting for BMD (per g/cm² of increase) in 221 incident dialysis patients

Parameter	Parameter Estimate	SEM	P
Intercept	1.187	0.03	<0.001
Age, >56 years	−0.034	0.01	0.02
Gender, female	−0.093	0.01	<0.001
PEW, SGA ≥ 2	−0.051	0.02	0.002
Phosphate, mmol/L	0.034	0.01	0.005
Follistatin, ng/ml	−0.026	0.01	0.002

Whole-model pseudo-r² = 0.36. Age was dichotomized according to the median value. Fifty-nine patients with no available data on BMD were not included in the model.

Follistatin Levels and Risk of Mortality

Ninety-two (33%) patients died during a median follow-up of 25 (6 to 60) months. The main causes of death were CVD related (n = 50), infections (n = 11), and cancer (n = 6). In a multivariate age- and sex-adjusted Cox regression model, follistatin levels were not associated with all-cause or CVD-related mortality (data not shown). The spline curve analysis was used to assess the effect of circulating follistatin levels on mortality after adjustment for age and sex (Figure 3). It depicts a positive, albeit nonsignificant, trend toward increased risk for mortality across increasing follistatin concentrations.

Figure 3.

Restricted spline curve showing on the left axis the age- and sex-adjusted hazard ratios and 95% CIs (dashed lines) for all-cause mortality associated with circulating follistatin levels in 261 incident dialysis patients. The model is plotted as restricted cubic splines with two knots. On the right y-axis, the distribution of patients (counts) across the follistatin spectrum is shown. For the analysis presented here, we excluded 19 patients considered as outliers (follistatin levels > 3.11 ng/ml). Log HR, log-transformed hazard ratio; 95% CI, lower and upper 95% CIs, respectively.

Discussion

Given the detrimental consequences of muscle wasting in the increased mortality risk of CKD patients (1–3), therapeutic strategies and appropriate targets to prevent this process have been given high priority. Animal studies have identified myostatin blockade as a potential interventional target for renal wasting (10,11). We here provide original descriptive data on follistatin, a natural inhibitor of myostatin actions, in a cohort of incident dialysis patients.

A novel and unexpected finding of this study is that circulating follistatin concentrations did not differ between CKD patients at different stages and controls (Figure 1). In contrast, two previous smaller studies have reported elevated follistatin levels in hemodialysis (HD) patients (18,19). Thus, because heparinization during HD activates follistatin, and because we studied incident dialysis patients, the effects of recurrent heparin administration on follistatin levels in patients on maintenance HD need further consideration (19–21). Although it is yet unclear how follistatin is metabolized and eliminated in the body (22), our data may indicate no significant role for renal clearance or decreased follistatin production together with impaired clearance in CKD patients. The cross-sectional nature of this analysis precludes from explaining this observation. Although animal studies have suggested increased myostatin mRNA expression as a cause of uremia-induced muscle depletion (10,11), one study in HD patients showed that myostatin mRNA levels in skeletal muscle were similar to control subjects (23).

To the best of our knowledge, ours is the first clinical report on follistatin levels in incident dialysis patients. First, a negative association between follistatin and muscle strength and a positive correlation between follistatin and PEW were observed. This is not consistent with the reported effects of follistatin in inducing muscle hypertrophy via a proliferation of the satellite cell, a progenitor cell in muscle tissue (12,24), suggesting that the increased follistatin levels are not sufficient to counteract the catabolic effect of the uremic milieu and inflammation in this patient group. Interestingly, satellite cell dysfunction is thought to be one of the multiple causes of muscle atrophy in CKD (25). At the same time, follistatin is also capable of inhibiting the biologic effects of activin A, which displays multiple functions including the control of bone remodeling (9,26–28). Thus, experimental evidence suggests that follistatin contributes to the positive regulation of extracellular matrix mineralization by modulating the effects of activin A on the osteoblasts (27). In our analysis, we can observe that BMD correlated strongly and negatively with circulating follistatin in univariate and multivariate analysis. Finally, animal studies have recently demonstrated an anti-inflammatory potential of follistatin in LPS-induced systemic inflammation (29), chemical-induced colonic inflammation (30), and vascular inflammation triggered by oscillatory shear stress (31). In accordance with one report showing a direct association between CRP levels and follistatin in women with polycystic ovary syndrome (32), we found that inflamed patients exhibited the highest follistatin levels. Indeed, experimental evidence indicates that follistatin reacts like a positive acute phase reactant (29,33–35).

Although a Cox regression model showed that follistatin did not predict outcome, a weak effect of increasing follistatin concentration on the mortality risk of the patients was observed using cubic spline curve analysis (Figure 3). Thus, we cannot exclude that a larger sample size would have allowed us to see an effect on mortality. Because multiple pathways other than muscle wasting (36) are operative in the uremic milieu and contribute to the excessive mortality risk of CKD patients, our finding that circulating total follistatin is not a useful biomarker in risk stratification may not be surprising.

Several caveats and limitations of our study should be considered. First, the cross-sectional nature of our analysis precludes from inferring the causality of these associations. Our main limitation is perhaps the impossibility to concurrently study circulating free myostatin in this patient material because of the instability of its available detection assays (37). With the observation presented here that elevated (and not reduced) levels of this myostatin inhibitor were found in wasted and inflamed CKD patients, our observation adds yet another molecule to the growing list of substances with documented salutary vascular, nutritional, and/or metabolic effects (such as adiponectin (38) and osteoprotegerin (39,40)) for which paradoxically higher circulating levels are observed in CKD patients and predict poor outcome. It is becoming apparent that the presence of concurrent inflammation and PEW likely modifies and reverts these associations (41). Whereas the exact mechanisms for this observation may differ between the myostatin binder follistatin and the other above-mentioned molecules, the observed phenomenon is mind-boggling and definitely worthy of further studies and considerations.

Because we can only report on total follistatin levels (being unable to discern between free and bound forms), two scenarios can be hypothesized: although increased total follistatin may be the result of effective myostatin/activin neutralization, it is also possible that our observation reflects an insufficient compensatory mechanism (resistance to the action of follistatin). Despite these restraints, the study presented here provides the first clinical analysis of this regulator and suggests its potential role in the course of CKD-related comorbidities, specifically on the processes of muscle wasting, bone metabolism, and inflammation. It is our hope that this hypothesis-generating study will stimulate further studies aiming at increasing our insight into the role of this molecular pathway in uremic PEW because it represents a potential therapeutic target being currently studied in various animal models (7,8,12,13).

In conclusion, although follistatin levels do not seem to be affected by renal function, the levels were increased in wasted and inflamed CKD patients, negatively associating with muscle strength and BMD. Although we are still ignorant of the precise mechanism in which follistatin is increased in inflamed and wasted CKD patients, we speculate on the basis of reported experimental evidence that a mechanism involving follistatin may be activated to counter-regulate the detrimental effects of myostatin and activin in uremia.

Disclosures

Baxter Healthcare Corporation employs B.L. P.S. is a member of the scientific advisory board of Gambro AB. None of the other authors declare any conflicts of interest.