Differential effects of L‐ and D‐phenylalanine on pancreatic and gastrointestinal hormone release in humans: A randomized crossover study

To investigate the effects of L‐phenylalanine on gastroenteropancreatic hormone release, glucose levels, subjective appetite and energy intake in humans, and to determine whether these effects were stereoisomer‐specific by comparing them with D‐phenylalanine.

K E Y W O R D S appetite, calcium-sensing receptor, glycaemia, humans, L-phenylalanine

| INTRODUCTION
High-protein diets (HPDs) improve long-term glycaemic control (as measured by HbA1c) in people with type 2 diabetes 1 and body composition in adults. 2,3 Accordingly, HPDs may benefit individuals with type 2 diabetes (T2D) and/or obesity, 4,5 but are difficult to adhere to. 6 Understanding how HPDs mediate their beneficial effects may facilitate development of agents to treat or prevent T2D and obesity.
Protein stimulates the release of gastroenteropancreatic hormones involved in glycaemic control and appetite regulation, 7 Following ingestion, protein is hydrolysed within the gastrointestinal tract into amino acids, which can be sensed via cellular mechanisms, including promiscuous amino-acid sensing G-protein-coupled receptors such as the calcium-sensing receptor (CaSR). 9 The CaSR is widely expressed in tissues not associated with calcium metabolism, 10 including pancreatic beta cells, 11 gastrointestinal enteroendocrine cells 12 and vagal afferents, 13,14 suggesting physiological roles in glucose and energy homeostasis and gastroenteropancreatic hormone release. With regards to nutrient detection, the CaSR is allosterically modulated by amino acids, showing a stereoselective response to their L-isomers. 15 Evidence suggests that the magnitude of change in both calcium mobilization and CaSR sensitivity varies depending on the L-amino acid utilized, with L-phenylalanine regarded as the most potent amino acid allosteric enhancer of the CaSR. 15 Human interventions show that acute ingestion of $10 g Lphenylalanine reduces the postprandial glucose area under the curve (AUC) by $65%, 16 decreases energy intake at a meal by $2000 kJ and stimulates CCK release. 17 Non-significant trends on PYY and GLP-1 levels have also been reported. 18 Previous research suggests the glycaemic and appetite-suppressing effects of L-phenylalanine in animals are mediated by the CaSR, activation of which stimulates the release of GLP-1, GIP and PYY. 19,20 Relative to humans, these animal models use comparatively high doses of L-phenylalanine ($3-12 mmol/kg, equivalent to $35-140 g of L-phenylalanine for a 70 kg human). However, comparing preclinical in vivo doses with human doses is difficult because of interspecies differences in body weight, size and metabolism that alter pharmacokinetics and pharmacodynamics, 21 and is particularly challenging when administering an agent believed to act within the gut lumen, where local exposure of the nutrient-sensing mechanism to the agent would be expected to matter more than systemic dose. It is therefore important to investigate whether these effects occur at doses feasible for human intervention.
Identifying whether effects on hormone release are specific to the L-phenylalanine stereoisomer would provide evidence that they are mediated by a specific biological mechanism, and would be in accord with a role for the CaSR in energy and glucose homeostasis in humans. This can be achieved by the inclusion of the synthetic isomer D-phenylalanine as a comparator, which exhibits weak modulatory activity of the CaSR 15 but provides an isonitrogenous, isosmotic and isocaloric control. Previous research shows that 10 g L-but not D-phenylalanine induces anorectic effects and CCK release, implying a role for the CaSR or another stereoisomer-specific mechanism in this response. 17 However, the comparative release of other gastroenteropancreatic hormones following D-and L-phenylalanine administration is unknown. Additionally, for L-phenylalanine to have possible therapeutic utility, it is helpful to establish whether it has wider, and possibly detrimental, metabolic effects beyond those expected on downstream pathways such as tyrosine synthesis.
The present study therefore aimed to investigate the effects of L-phenylalanine on gastroenteropancreatic hormone release, glucose levels, subjective appetite and energy intake in humans, and to determine whether these effects were stereoisomer-specific by comparing them with D-phenylalanine. Nuclear magnetic resonance analyses were also used to investigate possible wider effects on metabolism.

| Ethics
This study was granted ethical approval (London -West London & GTAC Research Ethics Committee, UK; reference number: 16/LO/1040) and conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to study commencement.

| Study design
Two studies were conducted: (a) a dose-finding study to determine the tolerability of L-phenylalanine and (b) an acute study to investigate and compare the effects of L-and D-phenylalanine on gastroenteropancreatic hormone release, glycaemia, subjective appetite and energy intake.
L-phenylalanine and D-phenylalanine powders were purchased from Euro OTC Pharma and encapsulated in opaque hypromellose capsules (Blackburn Distributions, UK) using a capsule machine (The Capsule Machine, Capsule Connection, USA).

| Dose-finding study
The dose-finding study had a five-visit, unblinded, ascending dose design. The five visits involved participants consuming increasing doses of L-phenylalanine at sequential visits: (a) 0, (b) 0, (c) 3, (d) 6 and (e) 10 g. The inclusion of two 0 g visits facilitated participant familiarization with the study protocol. The maximum dose (10 g) was based on the highest single dose used by prior studies investigating oral Lphenylalanine. 16,17,[22][23][24] All participants received doses in the same order with study visits separated by a minimum of 1 week.

| Acute study
The acute study had a four-visit, double-blind, placebo-controlled crossover design. The first visit for all participants was an acclimatization visit, during which participants consumed the placebo treatment to enable their familiarization with the study protocol. Participants were then randomly allocated to a sequence comprising the following treatments: (a) placebo, (b) 10 g D-phenylalanine or (c) 10 g L-phenylalanine. Study visits were separated by a minimum of 1 week.
Randomization Treatment sequences were generated using randomisation.com and allocated to participants using sealed opaque envelopes upon commencement of the first visit.
Blinding Treatments were blinded by an external staff member; participants and investigators were unaware of treatment. Visual analogue scales (VAS) were scored, energy intake calculated and assay performed while blinded. Treatment allocation was then revealed to permit statistical analysis.

| Sample size
The dose-finding study was a descriptive pilot study and thus a formal power calculation was not used. Pancreatic hormones were measured in the dose-finding study; we were particularly interested in the effects of L-phenylalanine on glucagon, which appeared to reflect the effects of a high-protein meal on glucagon release. 25 Sample size for the acute study was therefore calculated based on glucagon data from the dosefinding study ( Figure S1D). It was estimated that nine participants would provide an 80% chance of detecting a statistically significant difference with P less than .05. Consequently, we aimed to recruit 12 participants to account for an estimated dropout rate of 30%.

| Dose-finding study
Participants arrived at the Imperial Clinical Research Facility at about 08:30 AM following an $12-hour overnight fast. Participants were instructed to purchase and consume the same ready meal of their choice before 08:00 PM the evening prior to each study visit. Participants were also instructed to refrain from alcohol consumption and strenuous exercise for the 24 hours preceding the study. Upon arrival, an intravenous cannula was inserted into the antecubital vein of the participant to allow serial blood sampling. Following the collection of baseline measurements, participants ingested either (a) 0, (b) 0, (c) 3, (d) 6 or (e) 10 g L-phenylalanine at T = 0 (in 0, 0, 5, 9 and 13 capsules, respectively; size 00). Blood samples and 100 mm VAS were collected at 15-minute intervals (T = 0, 15,30,45,60,75,90,105,120,135 and 150) for 150 minutes following ingestion ( Figure S2).

| Acute study
The study design of the acute study was identical to the dose-finding study except that following baseline measurements, participants The ad libitum test meal was a tomato and mozzarella pasta bake ready meal (Sainsbury's, UK) with an energy density of 473 kJ/100 g and containing 3.3 g fat (26% of total energy), 14.9 g carbohydrate (53% of total energy) and 4.5 g protein (16% of total energy) per 100 g. Participants were given 20 minutes to consume the meal in an enclosed area and were instructed to 'eat until comfortably full'. The use of mobile telephones, laptops or other distractions was not permitted during this period.

| Dose-finding study
Blood samples were analysed for insulin and glucagon concentrations at all time points. Similarly, tolerability was assessed using VAS comprising five questions relating to nausea, warmness, irritability, anxiety and sleepiness at all time points.

| Acute study
Blood samples were analysed for insulin, glucagon, GIP and glucose concentrations for time points T = 0 to T = 120. Glucose, GLP-1 and PYY measurements were performed for an additional 30 minutes (T = 120 to T = 150) to investigate possible extended effects of treatments. Non-targeted metabolomic analysis using nuclear magnetic resonance was performed on samples at T = 0 and T = 45 to investigate the wider acute effects of L-phenylalanine on metabolism, in particular those pathways not typically associated with phenylalanine metabolism.
Subjective appetite was assessed using VAS comprising four questions relating to hunger, desire to eat, prospective consumption and fullness. This was converted to a composite appetite score (CAS) as previously described. 8 Absolute energy intake was calculated by measuring the mass of the test meal before and after consumption (Super-SS Bench Scale) and using the difference to calculate total energy consumption using the energy density of the meal. Relative energy intake was also calculated by adding the energy content of the treatments to the energy intake at the test meal.

| Assays
Insulin, glucagon and total GIP concentrations were measured using a MILLIPLEX MAP human metabolic hormone magnetic bead panel (Merck, USA). Glucose concentrations were measured using a colorimetric enzymatic activity assay (GL364, Randox, UK). Total GLP-1 and total PYY were measured using established in-house radioimmunoassays. 26,27 Assay protocols (including detection limits and intraassay coefficients of variation) and details of the non-targeted metabolomic analysis are described in appendices S2 and S3, respectively.

| Statistical methods
Total AUC for the preprandial (0-70 minutes) and postprandial periods (70-120/150 minutes) were calculated for all outcomes excluding metabolite and energy intake data. Energy intake and AUC data were analysed using a one-way ANOVA with repeated measures.
Time-series data for insulin, glucagon, GIP, GLP-1, PYY, glucose and VAS scales were analysed by fitting a mixed effects model (treatment x time). If a significant effect was detected, multiple pairwise comparisons were performed using t-tests with the Tukey correction being employed. Separate analyses for time-series data were performed for preprandial (0-70 minutes) and total (0-120/150 minutes) study periods for the acute study to prevent postprandial changes masking smaller preprandial changes. To avoid confusion, a single graphical representation for each outcome is presented, which combines the results from the preprandial and total study periods.
Mixed effects models were also used to assess associations between ad libitum energy intake and postprandial hormone and glucose concentrations.
Quantified small metabolites were analysed using Wilcoxon matched-pairs signed rank test and Bonferroni correction was used for multiple test correction.
All statistical analyses were conducted using Stata 16 (StataCorp, USA) and Prism 8.3.0 (Graphpad, USA) with statistical significance based on 95% limits (P < .05). All data are mean ± SEM unless otherwise stated.

| Participant flow
The numbers of participants who were enrolled, received the intended treatment and were analysed are presented in Figure S3 (dose-finding study) and Figure S4 (acute study).

| Acute study
Twelve healthy participants were recruited into the acute study. Of the 12 participants recruited, 11 completed all four visits (three males, eight females) with one participant being withdrawn from the study as they no longer satisfied the inclusion criteria. Data from this participant were not included in statistical analyses. The participant characteristics (N = 11; mean ± SD) were age 30.4 ± 13.1 years, height 1.68 ± 0.09 m, weight 64.5 ± 9.1 kg and BMI 22.2 ± 2.1 kg/m 2 .
3.3 | Dose-finding study: tolerability of L-phenylalanine L-phenylalanine was well tolerated at all doses and by all participants.
No side effects were reported during or between visits. No significant main or interaction effects for nausea 0-150mins (P > .05) and nausea AUC 0-150mins (P > .05; Figure S5), or for warmth, irritability, anxiety and sleepiness (P > .05; data not shown), were detected.

| Acute study: the effect of L-phenylalanine on gastroenteropancreatic hormones
Insulin, glucagon and GIP data are available for nine of the 11 participants because of a machine error during analysis of samples from two participants.

| Insulin
A significant interaction effect (treatment x time) was detected for insulin 0-70mins (P = .0089). Multiple pairwise comparisons identified significant differences between L-phenylalanine and other treatments at several time points ( Figure 1A). Insulin AUC 70-120mins also showed a significant main effect (P = .0216) with multiple pairwise comparisons showing a significant difference between placebo and Dphenylalanine (Table 1). However, no significant main or interaction effects were detected when analysing the data for insulin 0-120mins (P > .05) or insulin AUC 0-70 (P = .0646; Table 1).  (Table 1).

| Acute study: the effect of L-phenylalanine on circulating metabolites
Phenylalanine concentrations significantly increased from T = 0 to T = 45 following L-phenylalanine (P = .0215) or D-phenylalanine

| Acute study: the effect of L-phenylalanine on subjective appetite
No statistically significant differences between treatments in the CAS were detected (P > .05; Figure S9).

| Acute study: the effect of L-phenylalanine on ad libitum energy intake
Absolute energy intake at the ad libitum test meal (mean ± SEM) was not significantly different between the three treatments: (a) placebo  Figure 4A).
Relative energy intake at the ad libitum test meal corrected for the energy content of the treatments (mean ± SEM) was not significantly different between the three treatments: (a) placebo 3310 ± 344 kJ, (b) D-phenylalanine 3281 ± 325 kJ and (c) Lphenylalanine 3094 ± 309 kJ (P > .05; Figure 4B).

| DISCUSSION
This study confirms that acute ingestion of 10 g L-phenylalanine is well tolerated and stimulates insulin and glucagon release. We also show for the first time that L-phenylalanine stimulates GIP secretion in humans. As these effects were not observed with the ingestion of D-phenylalanine, they appear to be stereoisomer-specific. Both Lphenylalanine and D-phenylalanine reduced postprandial glycaemia, although this effect appeared prolonged for L-phenylalanine. However, L-phenylalanine at this dose and using this particular study design did not affect subjective appetite or short-term ad libitum energy intake.
The initial dose-finding study suggested that 10 g L-phenylalanine was tolerable. We did not administer D-phenylalanine in the dosefinding study, which may have provided greater insight into the effects of D-phenylalanine, although there were no reports of any side effects in the acute study. The metabolomics analyses found Lphenylalanine increased circulating tyrosine levels, as would be expected given the importance of L-phenylalanine in tyrosine synthesis, 28 but did not seem to have major effects on other metabolic variables. Minor differences in various lipids appear to reflect the effects of time across groups, which generally show the same pattern, although some changes only achieve significance following placebo treatment. This is reassuring regarding the specificity of the effects of L-phenylalanine, although it would be necessary to analyse samples at later time points and following chronic ingestion to confirm that Lphenylalanine does not have detrimental effects.

| L-phenylalanine stimulates the release of insulin, glucagon and GIP
L-phenylalanine stimulating the release of insulin and glucagon is consistent with previous human studies. 16,18,29 To our knowledge, this is the first study to show that acute L-phenylalanine appears to increase GIP concentrations in humans ( Figure 1C). There was a clear trend for higher GIP concentrations 30-70 minutes following Lphenylalanine ingestion, although only one time point reached statistical significance (T = 70) and further studies are therefore required to confirm this effect. Unfortunately, we were unable to measure GIP in our dose-finding study because of the limited sample volume.
L-phenylalanine has been shown to stimulate GIP release in swine and rodent models partly via the CaSR. 20,30 Congruent with this, no GIP response was observed following D-phenylalanine ingestion ( Figure 1C). Similarly, no significant changes in insulin or glucagon concentrations were observed following D-phenylalanine ingestion in the preprandial period ( Figure 1A,B), suggesting these responses may also be CaSR-mediated. These effects are also in accord with previous work showing a stereoselective CCK response with L-but not D-phenylalanine. 17 GIP promotes insulin and glucagon release via the GIP receptor on pancreatic beta-and alpha-cells 31,32 and may be involved in the L-phenylalanine-induced release of these hormones. Alternatively or additionally, L-phenylalanine may directly interact with the pancreas, as CaSR activation has been shown to stimulate insulin and glucagon secretion from beta-and alpha-cells in vitro, 33 or signal via the vagus nerve to modulate nervous control of pancreatic hormone release, given CaSR is expressed in vagal afferent neurons. 13,14 In accord with a recently reported human study, 18 Lphenylalanine did not significantly affect GLP-1 or PYY concentrations following ingestion (Figure 2A,B). However, postprandial PYY concentrations following D-phenylalanine were significantly elevated ( Table 1). As this is the first study to measure PYY concentrations following D-phenylalanine ingestion, the reason for this increase (and for the absence of a significant response following L-phenylalanine ingestion) is unclear. D-isomers of amino acids are intestinally absorbed more slowly than their L-isomers. 34 It is therefore possible that L-  reported to have several actions that could feasibly mediate this effect, 43 including directly inhibiting insulin secretion from the pancreas, increasing insulin sensitivity of skeletal muscle and adipose tissue, and slowing gastric emptying, thereby decreasing the rate of appearance of glucose into systemic circulation and the consequent insulin response. 44 GLP-1 also inhibits gastric emptying. 43 However, while we did not measure gastric emptying rate in the present study, a previous human intervention suggests L-phenylalanine does not affect gastric emptying. 18 It is important to note that postprandial glucose concentrations were measured following an ad libitum meal, at which participants ate varying quantities of food. The effects of D-and L-phenylalanine on postprandial glucose control may therefore be secondary to their carbohydrate intake, and not a direct effect of the treatment on glycaemic control. However, this is improbable. There was no significant association between ad libitum energy intake and postprandial glucose (β = 0.004; P = .7333). While there was a slight trend for reduced food intake following L-and Dphenylalanine ingestion, these treatments were associated with higher rather than lower GLP-1 and PYY levels, suggesting that differences in ad libitum food intake were not responsible for the differences observed.
4.3 | L-phenylalanine does not affect subjective appetite or short-term ad libitum energy intake L-phenylalanine did not significantly affect subjective appetite or energy intake at an ad libitum test meal served 70 minutes postingestion. Despite no significant difference being detected, energy intake following L-phenylalanine was $400 kJ lower than with placebo ( Figure 4A). Previous investigations into the anorectic properties of L-phenylalanine have also reported non-significant reductions in energy intake ($500 kJ) following L-phenylalanine relative to placebo, 45  L-phenylalanine-induced satiety has been reported to be modulated by menstrual cycle phase. 23 We did not record the cycle stage when conducting study visits, which may feasibly have affected energy intake, and therefore increased the variation in response to Lphenylalanine. However, when relative energy intake is calculated (i.e. including the 167 kJ provided by L-phenylalanine), the possible clinical significance of the reduction in energy intake with Lphenylalanine seems small ( Figure 4B).
In conclusion, our results show that phenylalanine-induced increases in insulin, glucagon and GIP prior to meal ingestion are dependent on its chiral form, in accord with the CaSR mediating these effects. We also show that L-phenylalanine and D-phenylalanine decrease postprandial glucose concentrations following an ad libitum test meal, with L-phenylalanine producing a more prolonged suppression. Although it is impractical to regularly consume 10 g of encapsulated amino acids, fortification of foods with L-phenylalanine may represent a possible strategy to prevent or treat glucose dysregulation and diabetes. L-phenylalanine did not significantly affect subjective appetite or short-term ad libitum energy intake, but this may be related to sample size and/or the time point at which the meal was offered. Furthermore, D-phenylalanine-and possibly L-phenylalanine-stimulated the release of PYY. Further work is needed to identify the mechanisms underlying the glucose-lowering effects of L-phenylalanine, for example, by co-administering GLP-1 and GIP receptor antagonists, and appropriately powered studies are required to clarify its effects on energy intake and GLP-1 release. Our work suggests L-phenylalanine plays a role in gastroenteropancreatic hormone secretion and potentially in glycaemic control, and that targeting the systems which sense ingested L-phenylalanine may be therapeutically useful.