Defining Predictors of Weight Loss Response to Lorcaserin

Abstract

Context

Individual responses to weight loss (WL) medications vary widely and prediction of response remains elusive.

Objective

We investigated biomarkers associated with use of lorcaserin (LOR), a 5HT_2cR agonist that targets proopiomelanocortin (POMC) neurons that regulate energy and glucose homeostasis, to identify predictors of clinical efficacy.

Methods

Thirty individuals with obesity were treated with 7 days of placebo and LOR in a randomized crossover study. Nineteen participants continued on LOR for 6 months. Cerebrospinal fluid (CSF) POMC peptide measurements were used to identify potential biomarkers that predict WL. Insulin, leptin, and food intake during a meal were also studied.

Results

LOR induced a significant decrease in CSF levels of the POMC prohormone and an increase in its processed peptide β-endorphin after 7 days; β-endorphin/POMC increased by 30% (P < .001). This was accompanied by a substantial decrease in insulin, glucose, and homeostasis model assessment of insulin resistance before WL. Changes in CSF POMC peptides persisted after WL (6.9%) at 6 months that were distinct from prior reports after diet alone. Changes in POMC, food intake, or other hormones did not predict WL. However, baseline CSF POMC correlated negatively with WL (P = .07) and a cutoff level of CSF POMC was identified that predicted more than 10% WL.

Conclusion

Our results provide evidence that LOR affects the brain melanocortin system in humans and that effectiveness is increased in individuals with lower melanocortin activity. Furthermore, early changes in CSF POMC parallel WL-independent improvements in glycemic indexes. Thus, assessment of melanocortin activity could provide a way to personalize pharmacotherapy of obesity with 5HT_2cR agonists.

obesity, lorcaserin, insulin, proopiomelanocortin, cerebrospinal fluid

As obesity rates continue to increase (1), the need for effective, targeted, nonsurgical treatments in obesity medicine is of paramount importance. Although several pharmacologic treatments of obesity have become available (2, 3), individual responses to these drugs vary widely, and at this time we have little ability to predict which patient will respond to which medication. We therefore investigated the effect of lorcaserin (LOR), a selective serotonin receptor (5HT_2cR) agonist previously US Food and Drug Administration (FDA) approved for weight loss (WL), to identify a biomarker profile that could predict response to this medication with the ultimate goal of achieving more effective personalized treatment of obesity. LOR was chosen as it targets the hypothalamic melanocortin system, which plays a critical role in regulating energy balance (4-9) and we have identified biomarkers of brain melanocortin activity in humans (10, 11). This system is composed of hypothalamic proopiomelanocortin (POMC) and agouti-related protein (AgRP) neurons whose peptide products interact at downstream melanocortin receptors (MC3/4Rs) to regulate food intake, energy expenditure, and glucose metabolism (4, 12-14). POMC-derived melanocyte-stimulating hormone (MSH) peptides decrease food intake while AgRP antagonizes the effects of MSH and is orexigenic. Both sets of neurons sense levels of energy stores and respond to a variety of nutrient, neuronal, and hormonal signals including leptin and insulin (15, 16). Disruption of the melanocortin system at multiple levels can lead to obesity in human and animal models (5, 17).

POMC neurons express 5HT_2cR and are activated by 5HT_2cR agonists (7), including LOR. Animal studies have shown that a functional brain melanocortin circuit is required for 5HT_2cR agonists to exert their effects both on energy balance and glucose metabolism (8, 9). Of note, changes in glucose homeostasis in animals could be achieved independently of WL (18-20). It is unknown if this occurs in humans as changes in glucose and insulin occur concurrently with diet and WL (21). Little is known about the effects of LOR on hypothalamic POMC in humans. We have therefore used cerebrospinal fluid (CSF) neuropeptide measurements to assess the short- and long-term effects of LOR on hypothalamic POMC in humans. In rodents CSF levels of POMC correlate with hypothalamic POMC messenger RNA (22). Human CSF contains high levels of the unprocessed POMC prohormone that correlate negatively with body mass index (BMI) and leptin and decline after diet-induced WL (10, 11, 23). Lower levels of the POMC-derived peptides β-endorphin (β-EP) and β-MSH are also detected in CSF and decline in parallel after WL (23). There is evidence that processing of the POMC prohormone to its biologically active peptides may also be regulated with respect to energy balance (24). We thus examined the ratio of β-EP to POMC as a potential indicator of changes in POMC processing. We have previously used CSF β-EP as a representative processed peptide as in contrast to α-MSH it can be reliably measured in CSF (23, 25). AgRP was also measured in plasma as there is evidence that plasma can serve as a biomarker of hypothalamic AgRP (23, 26), but this is not the case for plasma POMC peptides, which are of pituitary origin.

In this study we examined the effects of treatment with 1-week of LOR vs placebo while participants were weight stable before starting their diet; this was followed by treatment with LOR and diet for 6 months. We hypothesized that LOR would stimulate POMC neurons as reflected by an increase in CSF levels of the POMC prohormone as well as POMC-derived peptides after 1 week. However, a relative increase in CSF levels of the processed peptide β-EP compared to the POMC prohormone was also considered. We postulated that the observed changes in CSF POMC peptide levels might predict long-term WL response. Change in food intake during a test meal was examined as a potential predictor of WL response. Counterregulatory mechanisms that could limit the effectiveness of LOR including changes in levels of the melanocortin antagonist AgRP and the hypothalamic-pituitary-adrenal (HPA) axis were also examined. We also questioned whether CSF POMC levels at baseline might be predictive of response. The ratio of POMC to AgRP was examined as a potential measure of baseline melanocortin activity. The ability of LOR to affect glucose homeostasis was also examined with an emphasis on changes that occur before initiation of diet and subsequent WL. Before completion of our study, LOR was withdrawn from the market because of concern for possible increased risk of malignancy (27). However, the majority of our study was completed at the time, and we were able to demonstrate statistically significant changes in CSF levels of POMC peptides after 7 days of LOR before WL that were accompanied by substantial decreases in serum insulin and homeostasis model assessment of insulin resistance (HOMA-IR). Parameters at baseline that predicted greater WL at 6 months were also found.

Materials and Methods

Study Participants

A total of 32 individuals with overweight/obesity and without diabetes were recruited for this study. Thirty individuals completed the first phase of the study and were used for data analysis. Mean baseline BMI of the 30 participants (27 women, 3 men) was 34.3 (range, 28.3-42.7). Mean age was 36 years (range, 20-55 years). All participants were nonsmokers, had no history of substance abuse, and were not taking medications except for vitamins, thyroid hormone (n = 1), or contraception (estrogen + progesterone [n = 1]; progesterone only [n = 4]). Exclusion criteria included any clinically significant medical condition, eating disorder, prior bariatric surgery, or recent weight change ± 5% over the prior 6 months. All participants had a screening visit to confirm eligibility that included a physical exam, screening blood work (complete blood count and metabolic panel) and electrocardiogram. Diabetes was excluded by history and fasting blood glucose level less than 126 mg/dL. Premenopausal women were studied in most cases in the early follicular phase of the cycle. Pregnancy was ruled out by β-human chorionic gonadotropin test. This study was approved by the Columbia University Institutional Review Board and written informed consent was obtained from all individuals before their participation.

Study Protocol

This was a 2-phase study that examined responses both to short-term (7 days) and long-term (6 months) treatment with LOR (Fig. 1). Phase-1 was a double-blinded, crossover study in which participants were randomly assigned to receive placebo or LOR (10 mg twice a day) for 7 days and were then crossed over after a 3-week washout period to receive the other treatment for 7 days. Participants then continued to phase 2 and were treated with LOR (10 mg twice a day) for 6 months. Participants received nutritional counseling monthly to reduce daily caloric intake to 600 kcal below their calculated requirement (mean recommended caloric intake was 1530 kcal) and to exercise moderately for 30 minutes daily. The study was terminated early as LOR was withdrawn from the market. Thirty individuals completed phase 1 but only 19 participants (17 women, 2 men) completed 6 months of treatment: Six participants were withdrawn from the study once the FDA requested withdrawal of the drug; 2 participants were withdrawn because of headaches, which is a known side effect of LOR, and 3 participants withdrew for personal reasons. CSF and blood samples were collected, and test meals were performed at 3 study visits. There were 2 study visits in phase 1 after 7 days of placebo and 7 days of LOR and 1 study visit at the end of phase 2. Of the 19 individuals who completed 6 months of LOR, only 16 had a third lumbar puncture (LP).

Figure 1.

Study design (left panel). Phase-1: Double-blinded, crossover study. Participants were randomly assigned to placebo or lorcaserin (LOR) (10 mg twice a day) for 1 week, followed by a 3-week washout period, then crossed over to LOR or placebo for 1 week (n = 30). Phase 2: Participants were treated with LOR (10 mg twice a day) for 6 months (n = 19). Cerebrospinal fluid and blood samples were collected, and test meals were performed at the 3 study visits indicated by the arrows. Mean ± SEM monthly weight loss percentage (right panel) of phase 2 completers (n = 19). Average weight loss was 6.86% after 6 months of LOR.

Study Procedures

CSF (12 mL) was collected by LP between 0800 and 1000 hours after fasting from 2100 hours the previous night. The first 0.5 mL of CSF was discarded and the rest was pooled, centrifuged, and stored in aliquots at −80 °C. The procedure was well tolerated as reported previously (28). A blood sample was collected at the same time. A 24-hour urine sample was collected on the days preceding the 3 LPs.

Test Meals

Participants were fed a standardized breakfast (300 kcal) after the LP at each of the 3 study visits. A multi-item test meal consisting of a fixed buffet was performed 5 hours later (29). Individuals were seated at a table in front of the food array and told to eat as much as they wanted over a 1-hour period. Food energy and macronutrient consumption were calculated using the Nutrient Data Systems software.

Assays

POMC was assayed using an in house 2-site enzyme-linked immunosorbent assay (ELISA) with the capture monoclonal antibody directed against adrenocorticotropin (ACTH_10-18) and the detection antibody directed against γ-MSH (RRID: AB_2756529 and AB_2756530); there is no cross-reactivity with ACTH, α-MSH, β-MSH, or β-EP (10, 11). β-EP was measured by radioimmunoassay (RIA) as previously described (RRID: AB_2756516); there is 2.6% cross-reactivity with POMC on a molar basis (30). β-MSH was measured by RIA using an antiserum to human β-MSH and human β-MSH for tracer iodination and standards (Phoenix Pharmaceuticals); there is 10% cross-reactivity with POMC. AgRP was measured by ELISA and RIA with relative specificities for full-length AgRP and AgRP_83-132, respectively (11, 31).

Insulin, prolactin, and cortisol were measured in serum by Immulite1000 (Siemens Healthcare Diagnostics). Cortisol was measured in CSF by sensitive ELISA (Salimetrics) (32). Urine free cortisol was measured by the Irving Institute for Clinical and Translational Research Biomarkers Core Laboratory by liquid chromatography–mass spectrometry. Plasma and CSF leptin were measured by ELISA (R&D Systems) (11, 23). Glucose was measured by the hexokinase method.

Statistical Analysis

Data are expressed as mean ± SEM. Analyses were performed with GraphPad Prism version 9.4 for Mac (GraphPad Software). Hormone and neuropeptide levels after 7 days of placebo and LOR were analyzed by paired t test. N = 30 for all analyses except for insulin, glucose, and HOMA-IR, where n = 29 based on identification of an outlier by Grubbs outlier test; the outlier raised baseline placebo values. One-way repeated-measure analysis of variance with Dunnett multiple comparisons test was used for analysis comparing placebo to 7 days and 6 months of LOR. N = 19 for all analyses except for insulin, glucose, HOMA-IR, where n = 18 (same outlier removed as mentioned earlier), and CSF measurements, where n = 16 as only 16 participants underwent the third LP. Correlations were determined by linear regression analysis using Pearson correlation unless indicated differently with Spearman for nonparametric analysis. Statistical significance of the CSF POMC cutoff that predicted 10% WL was determined by Fisher exact test. Insulin resistance was calculated using HOMA-IR.

Results

Effects of Lorcaserin on Body Weight

There were no statistically significant changes in body weight, BMI, or waist circumference in phase 1 after 7 days of treatment with LOR vs placebo. Mean weight after placebo was 94.5 ± 2.7 kg (SEM) vs 94.5 ± 2.6 kg after LOR. Mean WL percentage throughout the 6-month LOR treatment period in the 19 participants who completed the study is shown in Fig.1; at 6 months WL was 6.86 ± 1.29%. Eleven individuals (58%) achieved more than 5% WL and 7 individuals (37%) achieved more than 10% WL.

Effects of Lorcaserin on Food Intake

There was no statistically significant difference in total caloric intake or caloric intake per kilogram of body weight during the test meals performed after 7 days of placebo vs 7 days of LOR (920 ± 119 vs 892 ± 109 kcal) or after 6 months of LOR (Tables 1 and 2), and changes in caloric intake after 7 days of LOR did not predict WL response to LOR at 6 months. Food intake was actually less during the second vs first test meal (822 ± 105 vs 990 ± 120) regardless of treatment (P < .01). However, at the end of the study there was a statistically significant negative correlation between caloric intake during the third test meal and percentage WL (r = −0.589; P = .016) and food intake during the third test meal vs placebo test meal was less in participants achieving more than 10% WL (P = .036).

Table 1.

Comparison of mean ± SEM baseline parameters measured after 1 week of placebo to mean values measured after 1 week of lorcaserin

	Placebo 1 wk ± SE	Lorcaserin 1 wk ± SE	P
Weight, kg	94.5 ± 2.7	94.5 ± 2.6	.675
Food intake, kcal	920 ± 119	892 ± 109	.646
Serum insulin, μIU/mL	17.6 ± 2.12	13 ± 1.17	.008
Serum glucose, mg/dL	98 ± 1.6	94 ± 1.3	.007
HOMA-IR	4.36 ± 0.57	3.06 ± 0.29	.009
Plasma leptin, ng/mL	55.2 ± 5.85	49.4 ± 5.16	.032
CSF leptin, pg/mL	403 ± 18.8	381 ± 19.7	.186
CSF POMC, fmol/mL	257 ± 18.5	225 ± 15.9	.0001
CSF β-endorphin, fmol/mL	14.4 ± 0.76	16.4 ± 0.95	.0017
CSF β-endorphin: POMC	0.061 ± 0.004	0.079 ± 0.005	< .0001
CSF β-MSH, fmol/mL	21.7 ± 1.18	20.7 ± 1.18	.242
Plasma AgRP, pg/mL (ELISA)	78.7 ± 7.66	79.3 ± 7.88	.819
CSF AgRP, pg/mL (ELISA)	25.9 ± 3.01	24.2 ± 3.14	.047
CSF AgRP, pg/mL (RIA)	45.4 ± 2.60	45.3 ± 2.78	.991
POMC: plasma AgRP	3.62 ± 0.30	3.19 ± 0.25	.017
CSF cortisol, ng/mL	5.7 ± 0.23	6.1 ± 0.33	.123
24-h urine cortisol μg/g creatine	11.0 ± 1.08	10.9 ± 0.83	.898
Serum cortisol, ng/mL	124.4 ± 9.8	112.9 ± 8.0	.299
Prolactin, ng/mL	12.7 ± 1.17	13.8 ± 1.17	.284

P values calculated by paired t test are included in the last column. n = 30 for all measures except for insulin, glucose, and homeostasis model assessment of insulin resistance (n = 29).

Abbreviations: AgRP, agouti-related protein; CSF, cerebrospinal fluid; ELISA, enzyme-linked immunosorbent assay; HOMA-IR, homeostasis model assessment of insulin resistance; MSH, melanocyte-stimulating hormone; POMC, proopiomelanocortin; RIA, radioimmunoassay.

Table 2.

Comparison of mean ± SEM baseline parameters measured after 1 week of placebo with mean values after 1 week and 6 months of lorcaserin in individuals who completed phase 2 of the study

	Placebo 1 wk ± SE	Lorcaserin 1 wk ± SE	Lorcaserin 6 mo ± SE
Weight, kg	96.8 ± 3.4	96.8 ± 3.3	90 ± 2.9c
Food intake, kcal	874 ± 105	869 ± 103	821 ± 133
Serum insulin, μIU/mL	19.1 ± 3.12	12.4 ± 1.35a	11.1 ± 1.48a
Serum glucose, mg/dL	99 ± 2.2	94 ± 1.6a	96 ± 1.9
HOMA-IR	4.77 ± 0.84	2.89 ± 0.33a	2.69 ± 0.40a
Plasma leptin, ng/mL	59.7 ± 8.26	52.3 ± 6.80	41.92 ± 5.81a
CSF leptin, pg/mL	406 ± 29.2	379 ± 25.8	329 ± 26.7a
CSF POMC, fmol/mL	253 ± 26	219 ± 21.4b	225 ± 20.1b
CSF β-endorphin, fmol/mL	15.1 ± 0.92	18.2 ± 1.25b	17.5 ± 1.08a
CSF β-endorphin: POMC	0.068 ± 0.006	0.092 ± 0.009b	0.085 ± 0.007c
CSF β-MSH, fmol/mL	20.7 ± 1.39	19.7 ± 1.48	21.6 ± 1.88
Plasma AgRP, pg/mL (ELISA)	73.6 ± 7.04	73.3 ± 7.05	80.5 ± 7.77
CSF AgRP, pg/mL (ELISA)	25.7 ± 2.61	23.9 ± 1.86	25.6 ± 2.50
CSF AgRP, pg/mL (RIA)	48.3 ± 4.00	49.4 ± 4.35	53.3 ± 4.95
POMC: plasma AgRP	3.79 ± 0.45	3.34 ± 0.37	3.00 ± 0.30a
CSF cortisol, ng/mL	5.4 ± 0.22	5.7 ± 0.43	6.1 ± 0.65
24-h urine cortisol μg/g creatine	11.8 ± 1.13	10.9 ± 1.12	11.4 ± 1.63
Serum cortisol, ng/mL	129 ± 13.4	113 ± 10.7	117 ± 9.3
Prolactin, ng/mL	13.2 ± 1.59	14.1 ± 1.49	13.1 ± 1.18

P values calculated by analysis of variance are indicated by the following: ^aP less than .05, ^bP less than .01, and ^cP less than .001. n = 19 for all measurements except insulin, glucose, and homeostasis model assessment of insulin resistance (n = 18) and for cerebrospinal fluid measurements (n = 16).

Abbreviations: AgRP, agouti-related protein; CSF, cerebrospinal fluid; ELISA, enzyme-linked immunosorbent assay; HOMA-IR, homeostasis model assessment of insulin resistance; MSH, melanocyte-stimulating hormone; POMC, proopiomelanocortin; RIA, radioimmunoassay.

Effects of Lorcaserin on Insulin, Glucose, and Other Hormones

There were highly significant decreases in serum insulin (P = .008), glucose (P = .007), and HOMA-IR (P = .009) after 7 days of LOR vs placebo despite no WL (Fig. 2). Insulin and HOMA-IR remained low after WL at 6 months with chronic LOR treatment and were significantly different from placebo but not from the values measured after 7 days of LOR (see Fig. 2). As described later, these changes in glycemic parameters were accompanied by statistically significant changes in CSF POMC peptides levels but there were no significant correlations between the degree of change in the glycemic parameters with the neuropeptide changes. There was a 10% decline in plasma leptin after 7 days of LOR vs placebo (P = .032) but CSF leptin levels did not change (see Table 1). After 6 months of LOR and statistically significant WL, there was a 30% decrease in plasma leptin (P = .017) and a 19% decrease in CSF leptin (P = .01) (see Table 2). There were no significant effects of acute or chronic treatment with LOR on serum or CSF cortisol, 24-hour urine free cortisol, or on serum prolactin levels (see Tables 1 and 2).

Figure 2.

Upper panel: mean ± SEM serum insulin (left) and homeostasis model assessment of insulin resistance (HOMA-IR) (right) in all participants after 1 week of lorcasrin (LOR) (teal bars) vs 1 week placebo (blue bars), before any weight loss (n = 29). Both decrease significantly after LOR. Lower panel: mean insulin (left) and HOMA-IR (right) in the study completers (n = 18). Both decreased after 1 week (teal bars) and 6 months (plum bars) of LOR vs 1 week placebo (blue bars) **P less than .01, *P less than .05.

Baseline Levels of Proopiomelanocortin, β-Endorphin, and β-Melanocyte–stimulating Hormone in Cerebrospinal Fluid (CSF) and Agouti-related Protein in CSF and Plasma

At baseline (after 7 days of placebo), CSF concentrations of POMC, β-EP, and β-MSH were highly correlated with each other (Fig. 3). These correlations persisted after LOR treatment (data not shown) despite the differential effects of LOR on POMC and β-EP levels as outlined later. CSF POMC levels at baseline correlated negatively with CSF leptin and positively with CSF and plasma AgRP levels measured by ELISA (P < .05) as shown in a previous cohort of lean and obese individuals (11). Similarly, CSF β-EP levels correlated negatively with CSF leptin (r = −0.494; P = .005) and positively with CSF (r = 0.538) and plasma (r = 0.619) AgRP (P < .002). CSF β-MSH also correlated positively with CSF and plasma AgRP levels (P < .01). Correlations of CSF POMC peptide levels with BMI and plasma leptin were not statistically significant in this overweight/obese cohort.

Figure 3.

Pearson correlations of cerebrospinal fluid (CSF) proopiomelanocortin (POMC) with CSF β-EP and CSF β-MSH measured at baseline after 1 week of placebo (blue) and of baseline CSF β-EP with β-MSH (plum) (n = 30).

Effects of Lorcaserin on Proopiomelanocortin, β-Endorphin, and β-Melanocyte–stimulating Hormone in Cerebrospinal Fluid (CSF) and on Agouti-related Protein in CSF and Plasma

There were consistent, highly significant changes in CSF levels of POMC peptides after 7 days of LOR vs placebo (Fig. 4). CSF POMC declined from 257 ± 18.5 to 225 ± 15.9 fmol/mL (P = .0001) but CSF β-EP increased from 14.4 ± 0.76 to 16.4 ± 0.95 fmol/mL (P = .0017). This resulted in a 30% increase in the ratio of β-EP/POMC (processed peptide/prohormone) (P < .0001) (see Fig. 4). There was no effect of time order on these responses, which occurred regardless of whether the participant received placebo or LOR first. Analysis of the individuals who completed 6 months of LOR treatment showed that the changes in CSF POMC peptides measured after 7 days of LOR persisted after 6 months of treatment (see Fig. 4). POMC decreased after 7 days of LOR vs placebo and remained lower after 6 months of LOR (P = .01). β-EP increased after 7 days of LOR and remained higher after 6 months of LOR (P = .016) despite WL. Compared to placebo, the ratio of β-EP/POMC increased by 35% after 7 days of LOR and remained increased by 25% after 6 months of LOR (P = .0005). There were no statistically significant changes in CSF levels of β-MSH measured after 7 days or 6 months of LOR (see Tables 1 and 2).

Figure 4.

Upper panel: mean ± SEM cerebrospinal fluid (CSF) proopiomelanocortin (POMC), β-EP, and the ratio of POMC/β-EP in all participants after 1 week of lorcaserin (LOR) (teal bars) vs 1 week on placebo (blue bars) (n = 30). CSF POMC decreased whereas β-EP and the ratio of POMC/β-EP increased after 1 week of LOR vs placebo. Lower panel: mean CSF POMC, β-EP, and the ratio of POMC/β-EP in study participants who completed 6 months of LOR and had 3 lumbar punctures (n = 16). As in the larger cohort, CSF POMC decreased whereas β-EP and the ratio of POMC/β-EP increased after 1 week of LOR (teal bars) vs placebo (blue bars). After 6 months of LOR (plum bars), CSF POMC remained lower and β-EP and the β-EP/POMC remained higher vs 1 week placebo (blue bars). ****P less than .0001, ***P less than .001, **P less than .01, *P less than .05.

There was a small decline in CSF AgRP measured by ELISA after 7 days of LOR (P = .047) but no change in plasma AgRP (see Table 1). No statistically significant changes in CSF AgRP measured by ELISA or RIA were noted after 6 months of LOR (see Table 2). However, plasma AgRP tended to increase after 6 months (see Table 2) and the increase was statistically significant (P = .03) if the 2 individuals who gained weight during the study were excluded. The percentage change in plasma AgRP at 6 months vs placebo correlated positively with percentage WL (r = 0.465; P = .045). This is similar to what has been reported after diet-induced WL (23).

Baseline Cerebrospinal Fluid Proopiomelanocortin and Weight Loss Response to Lorcaserin at 6 Months

Phase 1 changes in CSF POMC or POMC-derived peptides did not predict WL at 6 months. However, lower levels of CSF POMC (r = −0.425; P = .07) and lower POMC-to-AgRP ratio (calculated as a measure of melanocortin activity) (r = −0.486; P = .035) at baseline were associated with a better WL response to drug treatment. Interestingly, lower POMC-to-β-EP ratio at baseline was also associated with greater WL (r = −0.630; P = .004). Correlations of CSF POMC, POMC/AgRP, and POMC/β-EP with percentage WL at 6 months are shown in Fig 5. These correlations remained similar when corrected for BMI using partial Spearman correlation: POMC (r = −0.37; P = .128), POMC/AgRP (r = −0.46; P = .05), and POMC/β-EP (r = −0.63; P = .005). There was no statistically significant correlation between BMI at baseline and percentage WL. CSF β-EP and β-MSH did not correlate with percentage WL. CSF POMC was lower, as was the POMC/AgRP ratio and the POMC/β-EP ratio, in individuals achieving greater than 10% WL vs those with less than 10% WL (see Fig. 5). A CSF POMC cutoff of less than 220 fmol/mL at baseline was found to predict 10% WL at 6 months (P = .045 Fisher) (Fig. 6). Similarly, cutoffs of less than the median for POMC/AgRP (P = .045) and POMC/β-EP (P = .02) were predictive of 10% WL at 6 months.

Figure 5.

Upper panel: Spearman correlations of cerebrospinal fluid (CSF) proopiomelanocortin (POMC) and the ratios of POMC/AgRP and POMC/β-EP measured at baseline with the weight loss percentage achieved after 6 months of lorcaserin (LOR) (n = 19). Lower panel: mean ± SEM baseline POMC, POMC/AgRP, and POMC/β-EP in participants that achieved more than 10% weight loss (blue circles) vs less than 10% weight loss (teal squares). Baseline levels were lower in individuals that achieved more than 10% weight loss.

Figure 6.

Numbers of participants achieving more than 10% (blue bars) or less than 10% (plum bars) weight loss (WL) after 6 months of lorcaserin plotted using a cerebrospinal fluid (CSF) proopiomelanocortin (POMC) cutoff at baseline of 220 fmol/mL to separate into a high POMC (greater than 220 fmol/mL) group (left) and a low POMC (less than 220 fmol/mL) group (right). Baseline CSF POMC less than 220 fmol/mL predicted more than 10% weight loss with 6 months of lorcaserin treatment (P = .045 with Fisher exact test).

Discussion

The aim of this study was to identify biomarkers that would predict the WL response to LOR. With that intent, we analyzed CSF peptides as surrogate measures for brain melanocortin activity, as well as plasma hormones and food intake during a test meal. Despite early termination of the study due to medication withdrawal from the market, we had already completed phase 1 and had enough participants completing phase 2 to identify statistically significant effects. Participants achieved WL comparable to that in prior large LOR studies (6.9% in 24 weeks vs 5.6%-7% in 52 weeks in BLOOM (33), BLOOM-DM (34), and BLOSSOM studies (35)). The range of WL was also comparable to previous large LOR studies, with 58% achieving more than 5% WL and 37% more than 10% WL. Improvement in glycemic indexes accompanied by changes in the melanocortin system were detected before any WL. Statistically significant changes in CSF POMC peptides levels were also found after WL that were distinct from previously reported changes after WL induced by diet alone. In addition, our results show that LOR appears to more effective in individuals with lower baseline melanocortin activity.

LOR is thought to work through 5HT_2cRs on POMC neurons to achieve its effects both on energy balance and glucose homeostasis. Improvements in glucose homeostasis can be achieved irrespective of WL in animal models (13, 19, 20), with the melanocortin system being necessary to achieve that effect (13, 19, 20). In human studies, LOR has been shown to improve glycemic parameter and decrease insulin resistance, though this has been with concurrent diet and WL (21, 34, 36). In some studies, individuals also had type 2 diabetes for which they were already on medications. In contrast, our study participants did not have diabetes, were not on medications that would affect glucose homeostasis, and were asked to continue their preexisting eating behaviors while on 7 days of LOR or placebo. No WL was documented on either treatment. However, we saw statistically significant improvement in fasting glucose insulin and HOMA-IR after 7 days of LOR compared to placebo. Thus, LOR improved glucose homeostasis in participants without diabetes before any WL was achieved and without any confounding factors. This is consistent with mouse models showing that 5HT_2cR agonists improved glucose homeostasis at low concentrations that did not affect food intake or body weight; this required downstream activation of MC4-Rs (19, 20). Furthermore, selective loss of 5HT_2cR expression in POMC neurons caused hyperinsulinemia, hyperglucagonemia, and insulin resistance that occurred independently of changes in body weight (37). The improvements in glucose homeostasis that we report after 7 days of LOR were accompanied by statistically significant changes in CSF levels of POMC peptides consistent with the known role of POMC neurons in mediating the effects of 5HT_2cR agonists on glucose homeostasis in animals. These improvements in glucose homeostasis were sustained after WL.

Given the well-established effects of 5HT_2cR agonists in animal models, we hypothesized that LOR would activate POMC neurons in humans and that this would be reflected by changes in CSF POMC peptide levels after 1 week of treatment. Previous studies have shown that CSF POMC correlates with hypothalamic POMC in rodents and that both decrease in parallel after fasting (22). We have previously shown that human CSF contains high levels of POMC and that both POMC and the POMC-derived peptide β-EP decline concurrently after diet-induced WL (11, 23). We thus expected that CSF POMC and β-EP would increase after 1 week of LOR. While CSF β-EP did increase, this was accompanied by a decline in CSF POMC resulting in a highly significant increase in the ratio of the processed peptide β-EP to the unprocessed POMC prohormone. Furthermore, these changes in CSF POMC and β-EP persisted after 6 months of LOR in the setting of WL. Although this study did not include a group treated with placebo plus diet for 6 months, we have previously published data on individuals with comparable BMI (33.1 vs 34.3) and percentage WL (8.6% vs 6.9%) that can be used to compare the effects of WL induced by diet alone on CSF POMC peptides to the effects of diet plus LOR in the present study (23). In the previous diet study, we observed a 13.2% decline in CSF POMC, a 12.6% decline in CSF β-EP, and no changes in the β-EP-to-POMC ratio. In contrast, in this study we observed a 10.8% decline in CSF POMC, a 16% increase in CSF β-EP, and a 25% increase in the β-EP-to-POMC ratio. Thus, despite comparable WL with diet, CSF β-EP was increased with LOR treatment. There were, however, differences in the degree of calorie restriction and the time course of the 2 studies. In the prior study WL was achieved with an 800-kcal diet over a 6-week period, whereas in the present study daily caloric intake was reduced to 600 kcal below the individual calculated requirement (average 1530 kcal) over a 6-month period. These differences thus limit our comparison of the CSF POMC neuropeptide changes between the 2 studies. It should be noted that both of these studies enrolled primarily female participants although we attempted to recruit both sexes. It thus remains to be determined to what extent these results are applicable to men.

CSF β-EP was chosen as a representative processed peptide because, in contrast to α-MSH, it can be reliably measured in CSF and the assay has low cross-reactivity with POMC. CSF β-MSH was also measured and in contrast to β-EP did not change after LOR. This may be because the β-MSH RIA has more cross-reactivity with POMC. We have previously shown that β-MSH declines after diet-induced WL (unpublished observations), but this was not seen in the present study. Similar changes in leptin and AgRP were noted in both studies, with plasma and CSF leptin decreasing and plasma AgRP increasing after WL. The differential effects of LOR on CSF POMC and β-EP are consistent with an effect of LOR on POMC processing and/or on selective release of the processed peptides. The POMC prohormone has little biological activity and requires posttranslational processing for the generation of biologically active peptides, and there is evidence that POMC processing is regulated with respect to energy balance (24). In a previous study we showed that treatment with the opioid antagonist naltrexone, which stimulates POMC neurons in animals (38), caused an increase in CSF β-EP levels but POMC was unchanged resulting in an increased β-EP-to-POMC ratio (25). There are no animal data examining the effect of LOR on POMC peptides in the hypothalamus.

An important precision medicine goal was to try to identify potential predictors of WL with the use of LOR. Neither food intake nor changes in food intake during test meals performed after 7 days of LOR vs placebo were predictive of WL after 6 months of LOR. CSF cortisol, 24-hour urine cortisol, and serum prolactin levels were measured as serotonin has been shown to activate the HPA axis via stimulation of 5HT_2cR-expressing corticotropin-releasing hormone neurons (39) and to stimulate prolactin (40), and HPA activity and prolactin were not evaluated in previous LOR studies. There were no statistically significant changes in cortisol or prolactin or association with WL. Although LOR induced highly significant changes in CSF POMC and CSF β-EP, the degree of change after 1 week did not predict WL at 6 months. However, baseline levels of CSF POMC, as well as the ratios of POMC to AgRP and POMC to β-EP, were the most statistically significant predictors in our study. Specifically, lower POMC at baseline predicted more WL with LOR at 6 months, with a cutoff of less than 220 fmol/mL predicting more than 10% WL. The ratio of POMC to the MSH antagonist AgRP was calculated as a measure of melanocortin activity at baseline, demonstrating that lower calculated melanocortin activity at baseline was associated with a better WL response to a drug that stimulates the brain melanocortin system. Interestingly the ratio of POMC to β-EP was the best predictor of WL response for reasons that are not entirely clear.

The mechanisms responsible for individual differences in CSF POMC at baseline are not fully known but the concentration of POMC in CSF is remarkably constant when a given individual is studied twice under weight-stable conditions (unpublished data). Although there is a strong negative correlation between CSF POMC and BMI and leptin in a cohort of healthy lean, overweight, and obese individuals (11), we still find a wide range of POMC concentrations even within an obese cohort as in the present study. The correlation with BMI and POMC in the current cohort restricted to BMI greater than 28 was no longer statistically significant, but negative correlations both of POMC and β-EP with CSF leptin were still seen. However, it should be noted that there is considerable POMC neuronal heterogeneity and that POMC neurons that are sensitive to leptin are distinct from those that are sensitive to serotonin (14, 41, 42). While measuring CSF POMC levels is not a practical way to predict WL in the clinical setting, alternative ways to assess POMC and melanocortin activity could be found. As we continue to learn more about factors that regulate POMC expression and processing and melanocortin signaling, it may be possible to develop a genetic profile associated with low melanocortin activity and that would predict a better response to drugs that stimulate this pathway. Recently 11 loss-of-function variants in the 5HT_2cR were identified in individuals with severe obesity (43). Mice generated to express one of these variants developed obesity and had impaired activation of POMC neurons in response to LOR. The authors suggest that treatment with melanocortin agonists might be beneficial in such cases. On the other hand, treatment with 5HT_2cR agonists could be beneficial in cases that have a functional 5HT_2cR but have reduced melanocortin activity for other reasons.

LOR was withdrawn from the market by Eisai Inc in 2020 after an FDA recommendation out of concern for possible increased incidence of malignancy (27). However, since then, there has been debate over whether LOR was indeed responsible for the increase in malignancy risk (44), and given good WL outcomes with minimal side effects and a favorable cardiovascular profile (45-48), thoughts of ways to overcome potential risk of malignancy with LOR (49) have been entertained, including finding predictors of responders and minimizing the dose or developing alternative 5HT_2cR agonists (50). Our study provides evidence that assessment of baseline melanocortin activity could be used to predict the WL response to LOR and that LOR can have substantial glycemic effects regardless of WL, increasing its possible importance not only as an antiobesity but also as an antidiabetic medication. Thus, the role of 5HT_2cR agonists in obesity and perhaps diabetes is still of interest, maintaining the relevance of our study.

Acknowledgments

We thank the study participants for their participation and the staff of the Irving Institute for Clinical and Translational Research and the staff of the Ingestive Behavior Core of the NY Obesity Research Center for their help with implementing the clinical studies; Jaime Leskowitz for providing diet and nutritional counseling; and the Irving Institute for Clinical and Translational Research Biomarkers Core Laboratory for performing the urine cortisol assays.

Funding

This work was supported by the National Institutes of Health (grants R01-DK093920 to S.L.W.), Columbia University Clinical and Translational Science Award (UL1TR001873, T32 DK007559-31 to A.S.G.), and the Atkins Foundation (S.L.W).

Disclosures

The authors have nothing to disclose except for JK: Found Health-Chief Medical Advisor; GI Dynamics and Gila Therapeutics-Scientific Advisory Boards.

Data Availability

Some or all data sets generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on reasonable request.

Clinical Trial Information

Clinical trial registration number NCT03353220 (registered November 27, 2018).

Abbreviations

 
• β-EP

  β-endorphin

 
• AgRP

  agouti-related protein

 
• BMI

  body mass index

 
• CSF

  cerebrospinal fluid

 
• ELISA

  enzyme-linked immunosorbent assay

 
• FDA

  US Food and Drug Administration

 
• HOMA-IR

  homeostasis model assessment of insulin resistance

 
• HPA

  hypothalamic-pituitary-adrenal

 
• LOR

  lorcaserin

 
• LP

  lumbar puncture

 
• MSH

  melanocyte-stimulating hormone

 
• POMC

  proopiomelanocortin

 
• RIA

  radioimmunoassay

 
• WL

  weight loss