Serum levels of the cold stress hormones FGF21 and GDF-15 after cardiac arrest in infants and children enrolled in single center therapeutic hypothermia clinical trials

Abstract

Objective:

Fibroblast Growth Factor 21 (FGF21) and Growth Differentiation Factor-15 (GDF-15) are putative neuroprotective cold stress hormones (CSHs) provoked by cold exposure that may be age-dependent. We sought to characterize serum FGF21 and GDF-15 levels in pediatric cardiac arrest (CA) patients and their association with use of therapeutic hypothermia (TH).

Methods:

Secondary analysis of serum samples from clinical trials. We measured FGF21 and GDF-15 levels in pediatric patients post-CA and compared levels to both pediatric intensive care (PICU) and healthy controls. Post-CA, we compared normothermia (NT) vs TH (33°C for 72h) treated cohorts at <24h, 24h, 48h, 72h, and examined the change in CSHs over 72h. We also assessed association between hospital mortality and initial levels.

Results:

We assessed 144 samples from 68 patients (27 CA [14 TH, 13 NT], 9 PICU and 32 healthy controls). Median initial FGF21 levels were higher post-CA vs. healthy controls (392 vs. 40pg/mL, respectively, P<0.001). Median GDF-15 levels were higher post-CA vs. healthy controls (7,089 vs. 396pg/mL, respectively, P<0.001). In the CA group, the median change in FGF21 from PICU day 1–3 (after 72h of temperature control), was higher in TH vs. NT (231 vs. −20pg/mL, respectively, P<0.05), with no difference in GDF-15 over time. Serum GDF-15 levels were higher in CA patients that died vs. survived (19,450 vs. 5,337pg/mL, respectively, P<0.05), whereas serum FGF21 levels were not associated with mortality.

Conclusion:

Serum levels of FGF21 and GDF-15 increased after pediatric CA, and FGF21 appears to be augmented by TH.

Introduction

Therapeutic hypothermia (TH) to 33°C for 72h is standard of care for term newborns with hypoxic-ischaemic encephalopathy (HIE) [1]. The neuroprotective benefit of TH in children and adults after cardiac arrest (CA) is less clear; thus, targeted temperature management (TTM) is recommended encompassing a wider range of temperatures including strict fever prevention [2, 3]. Evidence suggests that neuroprotective cooling is more effective during early neurodevelopment, but factors that explain this observation remain to be defined.

Cold stress hormones (CSHs)—hormones upregulated by cold exposure–may mediate some aspects of neuroprotection during TH [4]. To our knowledge, no study has examined the effect of TTM protocols in the intensive care unit (ICU) on serums levels of CSHs. Moreover, brown adipose tissue (BAT) plays a major role in CSH secretion. Germane to the observation that TH is more protective in early neurodevelopment, BAT content is abundant during infancy and is inversely correlated with age in adults [5, 6]. Fibroblast Growth Factor 21 (FGF21) and Growth Differentiation Factor 15 (GDF-15) are two CSHs that show promise as putative neuroprotectants and are increased in adults after CA [4, 7–10]. We recently reported that the receptors for these ligands in the human brain are developmentally regulated (highest in newborns), supporting the concept that CSHs may mediate differences in the neuroprotective efficacy of TH in the very young vs. older cohorts with HIE [11, 12]. Specifically, the FGF21 co-receptor (β-Klotho) was expressed in hippocampus and cortex and highly age dependent, with robust expression in infants vs. children or adults [11]. The GDF-15 receptor (GFRAL) was minimal in these regions in the infant (<12-month) brain, but not seen in older children or adults [12]. We now investigate serum levels of FGF21 and GDF-15 in a cohort of pediatric post-CA patients at our institution, managed for 72h with either normothermia or 33°C TH.

Methods

Design and Setting:

This was a secondary analysis of prospectively collected serum samples from patients (n=27) enrolled in clinical trials of pediatric CA at a single institution. We also included: (a) pediatric ICU (PICU) controls (n=9) who were critically ill children admitted to the PICU for non-neurologic, non-CA aetiology with prospectively collected serum samples, and (b) healthy controls (n=32) who were healthy children recruited at the outpatient phlebotomy lab at our institution. All human studies, including this secondary analysis of samples, were approved by The University of Pittsburgh Institutional Review Board. Informed consent was obtained from the child’s parent or guardian and assent was obtained from the child when applicable. The protocols and results of the initial studies have been published [13–15].

Inclusion and Exclusion Criteria:

Resuscitated in-hospital cardiac arrest (IHCA) or out-of-hospital cardiac arrest (OHCA) patients admitted to the PICU between the ages of 1 week to 17-years-old were eligible for inclusion in the CA group. CA was defined as receipt of chest compressions for pulselessness by a healthcare worker. To meet inclusion criteria, patients in the CA group were required to have an arterial or venous catheter for blood withdrawal. Fourteen CA patients were enrolled in a randomized clinical trial of 24 vs. 72h of TH (NCT00797680), four CA patients were co-enrolled in a multicenter study comparing 48h of TH to normothermia (NCT00878644), and nine CA patients were enrolled in an observational study where the attending PICU physician dictated TTM strategy. We reported the post-resuscitation care protocol in our PICU [14]. From those studies, we identified subjects who 1) received targeted normothermia or TH (32.0–34.0°C) for 72h and 2) had serially collected serum samples for analysis of CSH levels.

PICU control eligibility included any child <18 years of age admitted to the PICU for non-neurologic, non-CA aetiology, and who had an indwelling central venous or arterial catheter. This cohort was from an exploratory study on biomarkers to predict unfavorable neurological outcome in critically ill children [15]. Healthy controls were recruited by a single investigator (RPB) before laboratory phlebotomy for Women Infants and Children (WIC) evaluation or surgery.

Clinical Data:

Clinical data on CA patients were collected from medical charts including patient characteristics, CA aetiology, resuscitation interventions, and clinical outcomes. Pediatric Cerebral Performance Category (PCPC) scores were obtained by the PI of the original study via telephone or in-person interviews with the parent/guardian. Data on PICU controls included demographics. Data on healthy controls included gender and age, consistent with the original IRB protocol to anonymize subjects.

Biomarker Measurements:

Blood samples were obtained twice daily from CA patients, days 1–4 post-ROSC, and on day 7 post-ROSC. Here we focused on the association of CSH levels during the cooling period, thus, we targeted for analysis serum obtained at <24h, 24h, 48h, and 72h post-ROSC. Blood from PICU controls was collected on d1–7 of PICU study enrollment; here we used the earliest serum collected (d1 of study enrollment). Blood from healthy controls was obtained at the outpatient phlebotomy lab. All samples were centrifuged and stored at −80°C.

CSH levels were measured using enzyme-linked immunosorbent assay (ELISA) kits via the manufacturer’s instructions (R&D Systems, Catalog: DF2100 and DGD150). We selected kits previously used to test human samples [16, 17]. All but two samples were run in duplicate (due to availability) and results averaged for analysis. Serial dilutions were performed if samples were above the upper limit of detection (ULOD). Samples remaining above ULOD were assigned the highest value detectable. For the FGF21 ELISA, the intra-assay coefficient of variance (CV) was <12% and inter-assay CV <10%. The lower limit of detection (LLD) of the assay is 4.67pg/mL. The GDF-15 ELISA had an intra-assay CV<10% and an inter-assay CV<10%. The LLD of the assay is 2.0pg/mL. Inter-assay controls were included on 8 of 10 plates.

Statistical Analysis:

Data were analyzed using StataIC 15 (StataCorp, College Station, TX, USA). For categorical values, we calculated frequencies and percentages. A Kruskal-Wallis test was used to compare initial FGF21 and GDF-15 levels post-CA, PICU controls and healthy controls, followed by a Wilcoxon Rank sum test for individual group comparisons. A Wilcoxon Rank sum test was used to investigate the association between <24h CSH levels post-arrest and hospital mortality as well as 6-month PCPC outcomes (dichotomized as favorable [PCPC: 1–3] and unfavorable [PCPC: 4–6 or increase from pre-CA PCPC if baseline PCPC >3]). To explore the potential effect of TH on CSH trajectories, we calculated the difference in FGF21 and GDF-15 levels at 72h vs. the initial samples (obtained within the first 24h of ICU admission) (i.e., the delta expressed as a positive or negative value). We chose 72h as the timepoint of interest since it marked the completion of TH. We used a Wilcoxon Rank sum to compare this difference between CA patients who received TH vs. normothermia. Since CSH receptors are developmentally regulated, we explored whether TH was associated with a greater change in patients <2 years of age vs. >2 years (Wilcoxon Rank sum test).

Results

The final analysis encompassed 56 samples from 14 CA patients receiving TH (32.0–34.0° C for 72h) and 47 samples from 13 CA patients in normothermia study arms, 9 samples from 9 PICU controls, and 32 samples from 32 healthy controls. The median age of CA patients was 7.8 yrs (IQR: 1.1–12.7) and did not differ vs. PICU controls (3.6 yrs; IQR: 2.5–10.6) or healthy controls (3.3 yrs; IQR: 1.0–7.1) (P=0.49). The CA cohort was 44% male and did not differ vs. PICU controls (67%) or healthy controls (42%--excluding one patient with unknown gender) (P=0.42). The CA cohort had more males in the normothermia group vs. the TH group (69% vs 21%, P=0.01), but otherwise the groups were well matched in terms of witnessed arrests, OHCA, bystander CPR, and arrest aetiology. Table 1 shows detailed pre-CA and CA characteristics in the normothermia and TH cohorts. The median time from PICU admission to <24h sample collection for CA patients was 10.5h (IQR: 6.5–13.0). Patient temperatures at the time each sample was obtained for the CA cohorts are shown in Figure 1.

Table 1:

Pre-/Post-arrest characteristics of the therapeutic hypothermia (33°C for 72h) and normothermia cohorts.

	CARDIAC ARREST		P-Value+
	NORMOTHERMIA (N=13)	HYPOTHERMIA (N=14)
Age (Q1-Q3)	7.8 (1.7–11.5)	5.0 (0.3–12.7)	0.59
Male (%)	9 (69%)	3 (21%)	0.01
Pre-arrest PCPC (Q1,Q3)	1 (1,2)	1 (1,1)	0.51
Witnessed (%)	7 (54%)	7 (50%)	0.84
Bystander CPR (%)	9 (69%)	11(79%)	0.59
Out-of-Hospital Cardiac Arrest	9 (69%)	11 (79%)	0.59
Asphyxia Arrest	11(85%)	11 (79%)	0.69
1st Rhythm
Asystole	3 (23%)	4 (29%)
PEA	5 (38%)	8 (57%)
Sinus	1 (8%)	1 (7%)
VF/VT	3 (23%)	1 (7%)
Unknown	1 (8%)	0
Adrenaline Doses (Q1,Q3)	2 (1,4)	2.5 (1,3)	0.75
Number of Defibrillations (Q1,Q3)	0(0,2)	0 (0,0)	0.31
Time CPR to ROSC in min (Q1,Q3)	30 (4,31)*	15 (10,24)	0.68
Mortality	5 (38%)	1 (7%)	0.05

PCPC=Pediatric Cerebral Performance Category score, +Wilcoxon Rank sum test

^*1 patient w/CPR during EMS (unknown duration)

Figure 1:

Temperatures of the two cardiac arrest cohorts at the time of blood sampling.

Median initial (<24h) FGF21 levels were higher in the CA group 392pg/mL (IQR: 66–612) vs. healthy controls 40pg/mL (IQR: 27–130) (P<0.001), though there was no difference between CA and median PICU control levels 817pg/mL (IQR: 178–4,535, P=0.12, Kruskal-Wallis between all groups P=0.0001) (Figure 2a). In contrast, median initial GDF-15 levels were increased in the CA group vs. both ICU and healthy controls (7,089pg/mL IQR: 3,805–13,306 vs. 2,122pg/mL IQR: 1,137–2,638 vs. 396pg/mL IQR: 332–547, P<0.001, Kruskal-Wallis P=0.0001) (Figure 2b).

Figure 2:

Box plot comparing initial FGF21 and GDF-15 levels in the cardiac arrest group to PICU and healthy controls (Figure 2a and 2b, respectively). FGF21 and GDF-15 plotted on a log scale. Levels initially compared using a Kruskal-Wallis test between all groups (P=0.0001 for FGF21 and GDF-15). Individual group differences were then compared using a Wilcoxon Rank sum test (CA: n=32, PICU control: n=9, healthy controls: n=32 for FGF21, n=31 for GDF-15)

Initial FGF21 levels post-CA did not differ in survivors 248pg/mL (IQR: 66–556) and non-survivors 757pg/mL (IQR: 392–1,187) (P=0.200) (Figure 3a). Initial FGF21 levels were also not associated with 6-mo outcome based on the 26 patients in whom long-term functional outcome data were available (favorable outcome: 206pg/mL [IQR: 111–569] vs. unfavorable outcome: 395pg/mL [IQR: 44–1,152], P=0.797). In contrast, initial levels of GDF-15 were associated with mortality (median survivors 5,337pg/mL [IQR: 3,355–8,382] vs. non-survivors 19,450 [IQR: 12,006–36,878], P<0.05) (Figure 3b). Similarly, initial GDF-15 levels were associated with 6-mo outcome (favorable outcome: 5,166pg/mL [IQR: 2,411–8,382] vs. unfavorable outcome: 10,878pg/mL [IQR: 7,336–25,895], P<0.05).

Figure 3:

Box plot comparing initial FGF21 and GDF-15 levels post-arrest between hospital survivors and non-survivors (Figure 3a and 3b, respectively). FGF21 and GDF-15 plotted on a log scale. Groups compared using a Wilcoxon Rank sum test (survivors: n=21, non-survivors: n=6).

FGF21 and GDF-15 in the normothermia and TH cohorts is shown in Figure 4a and Figure 4b, respectively. We examined the association between treatment with 72h of TH and serum FGF21 and GDF-15 trajectory by calculating the change in CSH levels (delta) over the first 72h. Based on levels from the patients (n = 23 for FGF21, n=22 for GDF-15) in whom data from <24 and 72h timepoints were available, FGF21 was augmented by 72h TH with a median difference of 231pg/mL (IQR: 33–1,006) in the TH group vs. a median difference of −20pg/mL (IQR: −307–71) in the normothermia group (P<0.05) (Figure 5a). Based off our prior report revealing that the β-Klotho co-receptor is expressed in children <2, the change in FGF21 levels was dichotomized by age <2 and >2 years. The median change in FGF21 for patients <2 years of age (n=7) was 1,006pg/mL (IQR: 33–1,373) vs. 82pg/mL (IQR: −1,131–384) for patients >2 years (n=7), although this trend was not significant (P=0.14). In contrast, GDF-15 levels were not augmented from initial levels to 72h by TH (median difference in TH −2,816pg/mL [IQR: −3,864 - −1,466] vs. normothermia −1,867pg/mL [IQR: −6,646 - −1,123], P=0.973) (Figure 5b). GDF-15 levels within the TH cohort were also not influenced by age. The median change in GDF-15 levels for patients <2 years (n=7) was −1,803pg/mL (IQR: −3,864- −1,954) vs. −3,109pg/mL (IQR: −6,970 - −1,688) for patients >2 years (n=6, P=0.39)

Figure 4:

FGF21 and GDF-15 levels at each measured timepoint (a and b, respectively). FGF21 and GDF-15 plotted on a log scale.

Figure 5:

Box plot examining the change in FGF21 and GDF-15 levels over the first 72h after cardiac arrest between the normothermia and therapeutic hypothermia (33°C for 72h) cohorts. Comparisons made using a Wilcoxon Rank sum test (hypothermia: n=14 for FGF21, n=13 for GDF-15, normothermia: n=9 both groups).

Discussion

To our knowledge this is the first characterization of FGF21 and GDF-15 levels in a pediatric CA population. Initial levels of FGF21 and GDF-15 significantly increased in serum from CA patients vs. healthy controls. Supporting our primary aim, TH was associated with augmented FGF21 levels over 72h cooling vs. normothermia, but not for GDF-15. FGF21 levels were not associated with either mortality or 6-mo outcome, while higher initial GDF-15 levels were associated with increased rates of hospital mortality and unfavorable 6-mo neurologic outcome. Finally, FGF21 and GDF-15 levels are increased in PICU controls (critically ill children with non-neurological conditions) vs. healthy control children.

The increase in GDF-15 and FGF21 levels seen after pediatric CA vs. healthy controls mirrors findings in adults after CA; in those studies increases in serum levels of GDF-15 and FGF21 were associated with unfavorable neurologic outcome [8–10]. The association between increased GDF-15 levels and increased hospital mortality/unfavorable 6-mo neurologic outcome in children may reflect initial insult severity, and suggests this hormone may be valuable for prognosticating outcomes after pediatric CA. Unlike GDF-15, FGF21 levels were not associated with hospital mortality or 6-mo outcome. This contrasts with findings by Pekkarinen and colleagues who showed that FGF21 was associated with unfavorable neurologic outcome in adults with CA [10]. The lack of an association here may relate to developmental differences, differences in arrest phenotype, or sample size limitations.

Mitochondrial stress stimulates the production/release of both hormones [18, 19]. Thus, increased secretion of GDF-15 and FGF21 after CA may be triggered by ischaemia-mediated mitochondrial damage or other signaling cascades. Other mitochondrial injury biomarkers markedly increase after CA [20]. GDF-15 is also upregulated by other cellular stressors (i.e. toxins and infections) [21, 22]. Given the likelihood of multi-organ injury in global ischaemia, a finding potentially shared with some PICU controls in our study, the robust increase above healthy controls is unsurprising. Consistent with this hypothesis, prior studies have revealed elevated FGF21 and GDF-15 levels in adult sepsis [22, 23].

Cold exposure increases FGF21 and GDF-15 serum levels [24, 25]. The sources of their secretion during cooling involve BAT and the liver [4, 25, 26]. Early studies revealed a rapid elevation (5–8h) of serum FGF21 after environmental cold exposure, but more recent studies suggested an initial decline in FGF21 serum levels during hypothermia [27, 28]. Applying environmental cold exposure to investigate thermogenic upregulation of CSHs in healthy human studies differs from TH in the ICU, where the complexity of background care (e.g., adrenergic agonists, sedatives, etc.) coupled with organ injury likely modifies the thermoregulatory response. It is likely that many of our patients were cold immediately post-resuscitation. As mentioned above, however, initial levels of CSHs are likely influenced by multiple factors beyond temperature—arrest duration, organ injury, infection, etc. Here we focused on the influence of prolonged cold exposure on FGF21 and GDF-15 levels post-arrest, as current recommendations for TH suggest cooling for at least 48h [2]. Consistent with that concept, increased FGF21 levels in the TH cohort were not apparent until 72h. Our study is unique in this regard since it examines whether CSH levels are associated with the use of prolonged TH during critical illness.

There are multiple CSHs [4]. The rationale for a combined analysis of FGF21 and GDF-15 during TH was that both have been shown to be neuroprotective (see below), and that FGF21 can induce GDF-15 during environmental cold exposure [25]. Thus, we expected increased FGF21 levels during ICU cooling would be associated with increased GDF-15. Our findings do not support their co-regulation during TH in the ICU. Minor changes in GDF-15 levels by TH may have been masked by marked increases seen in response to CA, i.e., that increases in GDF-15 levels had already produced a ceiling effect.

Pre-clinical studies confirm that exogenous FGF21 administration is neuroprotective in rodent models of stroke, traumatic neuronal injury, and newborn HIE [29–31]. Similarly, increased expression of GDF-15 in the CNS in vivo alters neuro-inflammation in a beneficial manner after trauma in rodents [7]. Thus, there is a strong rationale for understanding how TH impacts the levels of these CSHs. Also, FGF21 crosses the blood brain barrier, further establishing a framework to support the concept that TH-induced FGF21 may protect the injured brain [32].

Neurodevelopmental stage may also affect the strength of CSH mediated neuroprotection. We reported that the obligate co-receptor for FGF21 (β-klotho) is highly expressed in the human hippocampus and cortex in newborns, low in toddlers <2 years of age, and mostly absent in adults [33]. The wider distribution of β-klotho in the developing vs. the adult brain suggests the potential for a more robust benefit from circulating FGF21 in the young [11, 33]. Although not significant, our results suggest that children <2 years of age may release more FGF21 in response to TH vs. older children. Given the benefit of TH for HIE in term infants, promising results showing benefit in neonatal CA, and >7% improvement in 1-year neurodevelopmental outcome for pediatric OHCA, this putative developmental response merits assessment in a larger sample [1, 34, 35].

The cognate receptor for GDF-15 is GDNF-family receptor α-like (GFRAL) [25, 36]. In adults, GFRAL is expressed in the area postrema and nucleus of the solitary tract in the human brain [36]. Consistent with other reports, we failed to detect GFRAL in the hippocampus and prefrontal cortex of humans in older children or in adults [12]. We observed a faint (potential) GFRAL signal in the infant brain, but could not validate the authenticity of this band by mass spectrometry. Sabatini and colleagues described low, but detectable levels of GFRAL in the immature mouse brain suggesting developmental regulation of the receptor [37]. Thus, whether GDF-15 activates direct neuroprotective mechanisms in infants by stimulating neuronal GFRAL receptors remains unclear. GDF-15/GFRAL in the periphery may indirectly promote brain recovery by blunting neuroinflammation since multiple sclerosis patients with chronically elevated GDF-15 levels exhibited a stable disease course [38].

Several study limitations merit mention. This was an exploratory analysis of previously obtained samples from other studies, limiting sample size and our ability to control for confounders. This renders the findings susceptible to artifacts associated with multiple comparisons. The normothermia and TH cohorts were well-matched, but there were more males in the normothermia group. Prior studies found no difference in circulating baseline serum levels of FGF21 in prepubescent male vs. female children [39, 40]. However, potential sex-dimorphic effects on CSH induction during cooling merit future exploration. We also did not measure CSF levels. Finally, quantification of serum targets by ELISA is influenced by the sensitivity/specificity of the capture/detection as well as antibody cross-reactivity. Prospective studies are warranted of CSH levels across the insult spectrum in TH-treated patients across ages, and to confirm the ELISA results by mass spectrometry.

In summary, in a serum analysis of FGF21 and GDF-15 levels after 72h of TH vs. normothermia in pediatric CA patients, we identified robust initial increases in FGF21 and GDF-15 levels. A delayed increase in FGF21 but not GDF15 levels was observed in the TH vs. normothermia CA groups, suggesting divergent effects of TH on these CSHs. Further studies are needed to define the contribution of CSHs to hypothermic neuroprotection and determine whether manipulation of this response or administration of CSHs represent therapeutic opportunities after CA.