Acyl-ghrelin Is Permissive for the Normal Counterregulatory Response to Insulin-Induced Hypoglycemia

Insulin-induced hypoglycemia is prevalent in type 1 and advanced type 2 diabetes (1,2). Given the high risk of harm associated with hypoglycemia, a highly integrated counterregulatory response (CRR) system exists to prevent, minimize, and reverse hypoglycemia. As reviewed by Cryer (3), in humans without diabetes, the traditional CRR includes a decrease in insulin secretion, increases in glucagon, epinephrine, cortisol, and growth hormone (GH) release, and activation of the sympathoadrenal system. In patients with diabetes experiencing hypoglycemia due to over-insulinization, these counterregulatory defenses are often compromised (3). For instance, the normal fine-tuning of insulin release is not an option due to β-cell failure, and as a result of α-cell dysregulation, an attenuated glucagon response to hypoglycemia also may be present (4).

Although this CRR response in contexts of both health and diabetes has been long appreciated, it remains incompletely understood. Other regulatory factors that may participate in the CRR continue to be characterized. Acyl-ghrelin is one such factor that may work to link together the other, more traditional CRR hormones. For instance, increased sympathoadrenal tone is a potent signal stimulating ghrelin secretion (5). Also, acyl-ghrelin administration reduces insulin sensitivity, restricts insulin secretion, and raises plasma glucagon, GH, and cortisol, as previously reviewed (6,7). These interactions, as well as those affecting GLP-1 and somatostatin release, arcuate AgRP and hindbrain Phox2B neuronal engagement, hepatic gluconeogenesis, and hepatic autophagy, likely contribute to acyl-ghrelin’s overall glucoregulatory effects, which are emphasized by the actions of administered acyl-ghrelin to increase blood glucose (6,7) and, conversely, by the blood glucose–lowering effects of acyl-ghrelin deletion or blockade. Regarding the latter, ghrelin deletion improves hyperglycemia in ob/ob mice (8). Also, administration of a ghrelin receptor (GH secretagogue receptor [GHSR]) antagonist normalizes blood glucose in otherwise hyperglycemic HNF1α-deficient mice, presumably via blocking the glucose-raising actions of increased plasma ghrelin (9). Furthermore, elevated ghrelin occurs in the streptozotocin model of type 1 diabetes, in which studies with ghrelin knockout (KO) mice, GHSR-KO mice, and GHSR antagonist–treated wild-type (WT) mice suggest ghrelin involvement in streptozotocin-induced hyperphagia and hyperglycemia (10–16).

An intact ghrelin system also is required to block the development of severe hypoglycemia and death in a mouse starvation model. Indeed, ghrelin-KO mice exhibit a progressive decline in fasting blood glucose to the point of near death after a week-long 60% caloric restriction regimen that depletes body fat to <2% (17). Marked hypoglycemia under this regimen also occurs in mice with ablated ghrelin cells, mice deficient in ghrelin O-acyltransferase, mice overexpressing the endogenous GHSR antagonist liver-expressed antimicrobial peptide 2 (LEAP2), GHSR-null mice, and mice with ghrelin cell-selective deletion of β₁-adrenergic receptors ([18] and as previously reviewed [7]). Here, we test the hypothesis that ghrelin’s glucoregulatory actions are also central to the CRR to insulin-induced hypoglycemia.

Experiments were approved by the UT Southwestern Medical Center Institutional Animal Care and Use Committee. Male C57BL/6N mice, 8 to 10 weeks old, were used for Figs. 3 and 6. All other experiments used 8- to 10-week-old male ghrelin-KO and WT littermates on a C57BL/6N background generated by pairing mice heterozygous for the ghrelin-KO allele. Derivation of the ghrelin-KO line used (GKO1) was recently described (19). Its validation included demonstrating absent stomach ghrelin-immunoreactivity in three individuals, using previously described methods (19) and as indicated in Fig. 1, and undetectable plasma acyl-ghrelin levels (Figs. 4 and 5). Mice were housed at 21.5–22.5°C using a 12-h light-dark cycle and were provided ad libitum access to water and regular chow (2016 Teklad Global 16% Protein Rodent Diet; Envigo, Indianapolis, IN), except as indicated. Mean body weights of ghrelin-KO and WT littermates were similar.

Mice were fasted 3 h before testing, beginning at 8:00 a.m. Fasting blood glucose (t = 0 min) was measured from nicked tails using a Contour Next EZ monitoring system (Bayer, Parsippany, NJ). Insulin (Humulin-R; Eli Lilly, Indianapolis, IN) was diluted in sterile saline and injected at 2.5 units/kg body wt i.p. at t = 0 min. Blood glucose was measured from nicked tails multiple times postinjection. To assess the trajectory of plasma acyl-ghrelin and insulin during the protocol, blood samples (∼35 μL) were taken from nicked tails of a separate cohort at t = 0, 30, 120, and 240 min.

Clamps were performed in conscious, unrestrained mice using previously published recommendations (20). Briefly, 4–5 days prior to performing clamps, anesthetized (2% isoflurane) mice were implanted with a right jugular vein catheter (0.20-inch × 0.37-inch Silastic tubing) (Instech Laboratories, Plymouth Meeting, PA). The free end of the cannula was exteriorized from the dorsal intrascapular region, and the incision sites were sutured closed. Mice were fitted with a vascular harness (Instech Laboratories). Only mice returning to within 10% of their presurgical body weight by the clamp day were studied. On the clamp day, mice were fasted for 5 h (starting at 8:00 a.m., with access to water until 12:30 p.m.). Food and water were restricted during the clamp procedure. Insulin was infused at 4 mU/kg/min i.v. over 2 h. A 20% glucose solution was simultaneously infused i.v. at a variable rate to achieve hypoglycemia of 35–45 mg/dL during the final 20 min (steady-state period). Blood glucose was measured via tail nicks every 5 min. Blood samples to measure acyl-ghrelin were taken from tail nicks at t = −5 and 120 min (“Basal” and “Clamp” values, respectively, in Fig. 4C). At 120 min, blood samples to measure insulin, glucagon, epinephrine, norepinephrine, corticosterone, and GH were collected by cardiac puncture from mice following isoflurane anesthetization (“Clamp” values in Fig. 4D–I).

A separate cohort of mice underwent the identical surgery and preparation identical to those for clamps described above but did not undergo the clamps, so as to collect sufficient blood to determine “Basal” traditional CRR hormone levels. Blood glucose levels measured by tail nick were found to be similar to the “Basal” values included in Fig. 4A for the above cohort (data not shown). At 1:00 p.m., animals were anesthetized with isoflurane, and blood samples were collected by cardiac puncture.

Hyperinsulinemic-hypoglycemic clamps also were performed in two independent cohorts of ghrelin-KO mice and C57BL/6N mice given the GHSR agonist HM01 (30 µg/g body wt s.c.) (Helsinn Healthcare SA, Lugano, Switzerland) or an equivalent volume of saline 5 min before insulin was started.

For acyl-ghrelin, blood was collected into ice-cold EDTA-coated microfuge tubes. P-hydroxymercuribenzoic acid (final concentration 1 mmol/L) (Sigma-Aldrich) was added, plasma was isolated following centrifugation, and HCl was added to achieve a final concentration of 0.1 N. For other hormones, blood was collected into three different EDTA-coated microfuge tubes. For glucagon, aprotinin (final concentration 250 KIU/mL) (Sigma-Aldrich) was added. For catecholamines, EDTA-glutathione solution (9% w/v EDTA and 6% w/v glutathione, pH 7.4; 2 μL per 100 μL blood) was added. For insulin and corticosterone, no reagents were added.

ELISA kits were used for acyl-ghrelin (Millipore-Merck, Burlington, MA), insulin (Crystal Chem, Downers Grove, IL), glucagon (Mercodia AB, Uppsala, Sweden), and corticosterone (Enzo Life Sciences, Farmingdale, NY). Calorimetric assays were performed using a BioTek PowerWave XS Microplate spectrophotometer (BioTek, Winooski, VT) and BioTek KC4 junior software. Plasma catecholamines were determined using high-performance liquid chromatography at the Vanderbilt University Medical Center Hormone Assay and Analytical Services Core.

All statistical analyses and graph preparations were performed using GraphPad Prism 7.0. A Student t test, one-way ANOVA, followed by the Tukey multiple comparisons test, or two-way ANOVA was used to test for significant differences among test groups. Data with significant unequal variance were log transformed prior to performing analyses. Outliers were detected by the ROUT (robust regression and outlier removal) test. Notably, blood glucose curves and glucose infusion rates during the clamps were compared during the steady-state period. Statistical significance was defined as P < 0.05, and 0.05 ≥ P < 0.1 indicated a statistical trend.

The data that support the findings of this study are available from the corresponding author upon request.

We performed insulin bolus–induced hypoglycemia tests, in which a single bolus of insulin was delivered i.p. to induce hypoglycemia, in ghrelin-KO and WT littermates. Although ghrelin-KO and WT mice exhibited similar starting blood glucose levels, they fell more dramatically and remained lower for a longer period in ghrelin-KO mice (Fig. 2A), contributing to a lower blood glucose area under the curve (Fig. 2B).

Using a separate cohort of C57BL/6N mice, we measured plasma insulin, blood glucose, and plasma acyl-ghrelin four times during the insulin bolus–induced hypoglycemia test. A spike in circulating insulin following the i.p. insulin bolus occurred at 30 min and returned to baseline by 240 min (Fig. 3A). Blood glucose fell by 30 min, reached its nadir at 120 min, and then rebounded slightly by 240 min (Fig. 3B). Plasma acyl-ghrelin fell by 30 min, reached its nadir at 120 min, and then rose again in a nonstatistically significant manner by 240 min (Fig. 3C).

Next, ghrelin-KO and WT littermates underwent hyperinsulinemic-hypoglycemic clamps. A low insulin dose (4 mU/kg/min) was selected to minimize direct inhibitory effects of insulin on ghrelin secretion (21,22) in WT mice, while still achieving hypoglycemia. This low-dose protocol reduced blood glucose into the target hypoglycemic range of 35–45 mg/dL in WT mice with only a minimal glucose infusion rate (GIR) (Fig. 4A and B). Despite exhibiting similar basal blood glucose levels (WT: 160 ± 7 mg/dL vs. ghrelin-KO: 162 ± 8 mg/dL, P = 0.88) and blood glucose values within the target hypoglycemic range during the final 20 min of the clamps nearly identical to those of WT littermates, ghrelin-KO mice required a markedly elevated GIR (∼10-fold higher) by the end of the 2-h protocol (Fig. 4B).

We also determined plasma acyl-ghrelin levels and levels of traditional CRR hormones at the start and end (identified as “Basal” and “Clamp” values, respectively, in Fig. 4C–I) of the clamps. The insulin infusion reduced plasma acyl-ghrelin by ∼41% in WT mice, although it remained far higher than the essentially undetectable levels in ghrelin-KO littermates (Fig. 4C). Importantly, there were no differences in basal plasma insulin between ghrelin-KO and WT littermates, and insulin levels rose equivalently in both genotypes by the end of the 2-h exogenous infusion (Fig. 4D). A similar pattern was observed for plasma glucagon (Fig. 4E). Basal plasma epinephrine and norepinephrine were similar in both genotypes and did not change during the clamps (Fig. 4F and G). In contrast, basal plasma corticosterone was similar in both genotypes but became elevated during the clamps only in WT littermates (Fig. 4H). Basal GH also was similar in both genotypes, but its rise during the clamps was blunted in ghrelin-KO mice (Fig. 4I).

Also, we performed low-dose hyperinsulinemic-hypoglycemic clamps in ghrelin-KO mice following administration of the GHSR agonist HM01 versus vehicle (saline). HM01 is a partial GHSR agonist exhibiting high GHSR binding affinity, high bioavailability, high central nervous system permeability, a longer plasma half-life than acyl-ghrelin, and several actions mimicking those of acyl-ghrelin (23–26). The target hypoglycemic range of 35–45 mg/dL was achieved in both groups (Fig. 5A). During the final 20 min of the clamp, HM01-administered ghrelin-KO mice required about a fourfold lower GIR than those given saline (Fig. 5B). Notably, the GIR needed by HM01-administered ghrelin-KO mice was similar to that required by WT mice from the above study (Fig. 4B). HM01 administration did not affect end-of-clamp plasma acyl-ghrelin, insulin, glucagon, or epinephrine (Fig. 5C–F), although it lowered plasma norepinephrine and raised corticosterone and GH compared with saline-administered ghrelin-KO mice (Fig. 5G–I).

The low-dose hyperinsulinemic-hypoglycemic clamp procedure with HM01 versus saline was repeated in a separate cohort of C57BL/6N mice. The target hypoglycemic range of 35–45 mg/dL was achieved in both groups (Fig. 6A). There was no significant effect of HM01 administration on the GIR during the final 20 min of the clamp (Fig. 6B).

Several facets of ghrelin biology had suggested that acyl-ghrelin may participate in the CRR to insulin-induced hypoglycemia. For example, ghrelin release from cultured ghrelin cells is stimulated by low glucose (see below), and acyl-ghrelin interacts with several components of the traditional CRR to hypoglycemia (6,7). Indeed, a prior study showed that 18-month-old GHSR-KO mice require a higher GIR during hyperinsulinemic-hypoglycemic clamps to maintain hypoglycemia (27). That study did not address the mechanistic underpinnings of the increased GIR, nor is one able to distinguish the effects of loss of ghrelin action versus loss of ghrelin-independent GHSR activity when using GHSR-KO mice (7). Here, we used two different models of insulin-induced hypoglycemia in ghrelin-KO and WT mice and measurements of circulating glucoregulatory hormones to characterize the contributions of acyl-ghrelin to the CRR to hypoglycemia. We now show that ghrelin-KO mice not only exhibit more pronounced and prolonged hypoglycemia than WT littermates exposed to the same insulin bolus–induced hypoglycemia test but also require markedly higher GIRs to maintain the same target blood glucose as WT littermates during low-dose hyperinsulinemic-hypoglycemic clamps. These data support previous studies suggesting increased insulin sensitivity upon ghrelin deletion (7). Other new findings suggest that acyl-ghrelin—which differs from the traditional CRR hormones in that its levels do not rise (but rather, fall) in the setting of hypoglycemia resulting from overinsulinization—helps integrate several components of the traditional CRR to insulin-induced hypoglycemia. These include corticosterone and GH, as evidenced by attenuated plasma corticosterone and GH responses in hypoglycemia-clamped ghrelin-KO mice and normalized GIR and corresponding increased plasma corticosterone and GH in hypoglycemia-clamped ghrelin-KO mice administered the GHSR agonist HM01. Thus, these data suggest that the presence of circulating acyl-ghrelin is permissive for the normal CRR to insulin-induced hypoglycemia. Overall, these data provide a clearer understanding of the extent of acyl-ghrelin’s glucoregulatory efficacy and the factors permitting the normal CRR.

This role for acyl-ghrelin during the CRR to insulin-induced hypoglycemia aligns with its other well-described glucoregulatory actions, which include effects of administered acyl-ghrelin and endogenous plasma acyl-ghrelin elevations to increase blood glucose (6,7), and which are emphasized by the blood glucose–lowering effects associated with acyl-ghrelin deletion or blockade. As mentioned, ghrelin-KO mice and similar mouse lines develop life-threatening hypoglycemia when subjected to a week-long caloric restriction protocol modeling starvation (7,17,18). Notably, the setting of insulin-induced hypoglycemia investigated here differs from the starvation model in that plasma insulin becomes nearly undetectable in the latter (18).

Published studies investigating the counterregulatory effect of the ghrelin system specifically in the setting of insulin-induced hypoglycemia are few in number. These include the above-mentioned mouse study using GHSR-KO mice (27) and a few human trials (see below). More common are studies in which ghrelin-KO, GHSR-KO, brain-selective GHSR-KO, and/or GHSR antagonist–administered mice have undergone standard insulin tolerance tests, hyperinsulinemic-euglycemic clamps, hyperglycemic clamps, and glucose tolerance tests (8,27–31). For instance, ghrelin-KO mice exhibit greater blood glucose drops during standard insulin tolerance testing (8). They also require increased GIRs during hyperinsulinemic-euglycemic clamps due to greater suppression of endogenous glucose production and increased glucose disposal (8). Similarly, during hyperinsulinemic-euglycemic clamps, GHSR-KO mice exhibit enhanced insulin-mediated suppression of endogenous glucose production and increased glucose disposal, including more insulin-stimulated glucose uptake into white adipose tissue and skeletal muscle, while during hyperglycemic clamps, they require reduced insulin (30,31). These effects of GHSR deletion may be age dependent, as the higher GIR required by GHSR-KO mice during hyperinsulinemic-euglycemic clamps was present in 3-month-old but not 18-month-old mice (27). Thus, lack of ghrelin or GHSR sensitizes the body to insulin administration not only during standard insulin tolerance tests and hyperinsulinemic-euglycemic clamps but also as shown here and by others (27) and during hypoglycemic conditions resulting from overinsulinization.

It also is worthwhile to discuss the contributions of the traditional CRR hormones to acyl-ghrelin’s CRR effects. The mean plasma GH achieved during the newly reported hyperinsulinemic-hypoglycemic clamps was 41% lower in ghrelin-KO mice than in WT littermates. In contrast, HM01 administration raised plasma GH in clamped ghrelin-KO mice while also reducing the GIR to a level similar to that needed for WT mice. These findings align with acyl-ghrelin’s long-known actions as a potent GH secretagogue and studies identifying GH as a key mediator of acyl-ghrelin’s glucoregulatory effects during the above-described week-long caloric restriction protocol modeling starvation (7,17,18,32). Importantly, GH rises gradually over the course of the week-long caloric restriction protocol in WT mice but not in ghrelin-KO mice; replacement of GH by infusion during the protocol averts hypoglycemia and death in mice lacking acyl-ghrelin (17,18). Thus, the hyperinsulinemic-hypoglycemic clamp model can be added as another setting in which GH likely mediates ghrelin’s glucoregulatory actions.

Although the patterns of plasma corticosterone were not identical to those of GH, they suggest a role for corticosterone as well in mediating ghrelin’s effects during insulin-induced hypoglycemia. In particular, corticosterone rose by the end of the hypoglycemic clamp in WT mice but not in ghrelin-KO mice. In contrast, HM01 administration to clamped ghrelin-KO mice raised plasma corticosterone while also reducing the GIR to a level similar to that needed for WT mice.

The genotype-dependent and treatment-dependent effects on catecholamines observed here during the hyperinsulinemic-hypoglycemic clamps were not necessarily as expected. First, while we had predicted plasma epinephrine and norepinephrine to rise in WT mice over the course of the clamp as part of the CRR, based on their trajectories during the CRR in humans (3), they did not. That said, in prior work, C57BL/6J mice—a substrain different yet related to the mice used here—exhibited only modest rises in epinephrine and no changes in norepinephrine during hyperinsulinemic-hypoglycemic clamps; other strains ranged from showing marked increases in catecholamines to none at all (33). Thus, the literature does not support the notion that catecholamine levels increase during hypoglycemia in all strains of mice. It is also possible that any small, albeit physiologically significant, changes to catecholamines, if present, could be masked by the relatively high levels that were detected. These relatively high levels, which were in the ng/mL range compared with the pg/mL of other studies (34), in turn may have been influenced by the methods used here for blood collection. For instance, prior work showed that blood sampling from tail nicks compared with indwelling arterial cannulas induces a rise in plasma catecholamines (34). Second, plasma norepinephrine was lower in HM01-treated, clamped ghrelin-KO mice than in saline-treated mice. Although only speculative, perhaps HM01’s effects to raise plasma corticosterone and GH were sufficient in the setting of the clamp such that only a minimal norepinephrine response was needed.

Of interest, links between acyl-ghrelin and corticosterone and between acyl-ghrelin and the catecholamines previously have been established in the literature (35). These mainly have been in the context of stress. For instance, administered acyl-ghrelin induces c-fos expression in hypothalamic corticotropin releasing factor (CRF) neurons, increases CRF mRNA in those neurons, induces CRF release from hypothalamic explants, and increases plasma ACTH, corticosterone, and epinephrine in rodent and/or human subjects (36–44). Additionally, elevations in plasma ACTH and/or corticosterone are attenuated in ghrelin-KO mice following a 15-min restraint stress protocol and in GHSR-deficient mice following exposure to a 10-day chronic social defeat stress protocol (45,46).

We also must acknowledge another potential caveat of our studies, regarding the use of ghrelin-KO mice—in particular, the possibility that long-term absence of ghrelin may have resulted in compensatory changes during development that altered the true effects of absence of ghrelin action in the setting of insulin-induced hypoglycemia. That said, HM01 was able to normalize the GIR phenotype of ghrelin-KO mice and, furthermore, increased the corticosterone and GH responses that were otherwise reduced in ghrelin-KO mice. Also, as reviewed in ref. 7, mice in which ghrelin cells were ablated only after adulthood was attained exhibited the same profound hypoglycemia when subjected to the above-described starvation model as similarly treated ghrelin-KO mice, suggesting no compensatory masking of ghrelin’s glucoregulatory effects by germline ghrelin deletion, at least under that protocol.

As a final discussion point, for acyl-ghrelin to play a key role in the CRR to insulin-induced hypoglycemia, we might have expected its plasma level to elevate upon hypoglycemia induction, as occurs with the traditional CRR hormones. Such also might be expected based on studies demonstrating increased ghrelin release from cultured gastric mucosal cells upon their incubation in media with 0 or 1 mmol/L glucose (21,47). However, in the insulin bolus–induced hypoglycemia model, plasma acyl-ghrelin fell in WT mice by the time plasma insulin reached its peak, was even lower when blood glucose reached its nadir, and did not rise in a statistically significant manner by 2 h following the nadir in blood glucose. Plasma acyl-ghrelin levels also fell over the course of the hyperinsulinemic-hypoglycemic clamps in WT mice.

Effects of insulin to decrease plasma acyl-ghrelin have been described before. For instance, insulin reduces ghrelin release from cultured gastric mucosal cells, likely via direct interaction with insulin receptors, which are highly expressed in ghrelin cells (21,22). Also, at least three human studies report changes to plasma total ghrelin (acyl-ghrelin plus desacyl-ghrelin) in response to insulin-induced hypoglycemia. In the first, following a 10-min insulin infusion in which blood glucose fell to the hypoglycemic range or was maintained in the euglycemic range using infused dextrose, ghrelin dropped by the same amount in both conditions (48). In the second, a rapid fall in ghrelin was observed during a hyperinsulinemic-euglycemic clamp procedure (49). These two human studies suggest that insulin suppresses ghrelin release independently of the degree of glycemia. In a continuation of the second study, subsequent hypoglycemia, as achieved by lowering the GIR, did not affect plasma total ghrelin levels, suggesting that hypoglycemia does not reverse the fall in total ghrelin induced by hyperinsulinemia (49). In the third, a single i.v. insulin bolus to induce hypoglycemia lowered total ghrelin; however, it then rebounded 1 h later to a level higher than that observed in subjects given saline instead of insulin and in subjects given insulin but remained euglycemic due to infused dextrose (50). This suggests that hypoglycemia resulting from a bolus of insulin can stimulate ghrelin release in humans, becoming apparent in a delayed fashion after the initial spike in insulin dissipates. That said, we did not observe a hypoglycemia-induced plasma acyl-ghrelin elevation in the insulin bolus–induced hypoglycemia mouse model of the current study.

Yet, despite the observed drop in plasma acyl-ghrelin in WT mice subjected to insulin-induced hypoglycemia, our data suggest that the remaining circulating acyl-ghrelin nonetheless continues to substantively exert key glucoregulatory effects, as emphasized by the exaggerated hypoglycemia of ghrelin-KO mice. Similarly, despite the observed drops in plasma acyl-ghrelin in WT mice subjected to hyperinsulinemic-hypoglycemic clamps, the remaining acyl-ghrelin retains its crucial permissive functions during the CRR, as highlighted by the requirement for a markedly higher GIR in ghrelin-KO mice, the attenuated changes to the traditional CRR hormones corticosterone and GH in ghrelin-KO mice, and their rescue upon administration of the GHSR agonist HM01.

The authors thank the Vanderbilt University Medical Center Hormone Assay and Analytical Services Core for performing the catecholamines assays.

Funding. This work was supported through research grants from the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases (R01-DK-109408 to E.D.B., R01-DK-119341 to E.D.B. and J.M.Z., and R56-DK-071320 and R01-DK-103884 to J.M.Z.), a gift from the David & Teresa Disiere Foundation (to J.M.Z.), the Diana and Richard C. Strauss Professorship in Biomedical Research, the Mr. and Mrs. Bruce G. Brookshire Professorship in Medicine, and the Kent and Jodi Foster Distinguished Chair in Endocrinology, in Honor of Daniel Foster, M.D. (to J.M.Z.).

Duality of Interest. C.P. is employed by Helsinn Healthcare SA. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.S. conceptualized and performed the experiments, analyzed and interpreted the data, and helped write the manuscript. D.G. performed the experiments and helped analyze and interpret the data. B.K.M. and C.L. conceptualized and performed some experiments. B.G.F., C.C.L., S.O.-L., and N.P.M. performed the experiments. C.P. provided HM01. E.D.B. and J.M.Z. conceptualized the experiments, secured funding, interpreted the data, supervised the research activity, and helped write the manuscript. K.S., E.D.B., and J.M.Z. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.