Re: 음식치료제 BHB + 페닐알라닌 -=＞ 식욕억제, BHB+Lysine --＞후성유전학적 변화

[문형철]

부작용이 없는 위고비를 향한 연구!!

음식치료제 베타-하이드로뷰티레이트 대사산물 --> bhb+페닐알라닌 ... 식욕억제 기전

β-hydroxybutyrate (BHB, 베타-하이드록시뷰티레이트)가

단순히 에너지원으로만 쓰이는 게 아니라,

새로운 2차 대사 경로(shunt pathway)를 통해 항비만 대사산물을 만든다

https://www.cell.com/cell/fulltext/S0092-8674(24)01214-5

이 논문은

β-hydroxybutyrate (BHB, 베타-하이드록시뷰티레이트)가

단순히 에너지원으로만 쓰이는 게 아니라,

새로운 2차 대사 경로(shunt pathway)를 통해 항비만 대사산물을 만든다는 사실을 밝혔습니다.

주요 내용

1. 새로운 BHB shunt pathway 발견
   • 효소 CNDP2(carnosine dipeptidase 2)가 BHB와 자유 아미노산(특히 페닐알라닌 등)을 결합시킴.
   • 이 과정에서 BHB-amino acids라는 새로운 계열의 케톤 대사산물이 생성됨.
   • 그중 가장 풍부하고 중요한 물질이 BHB-Phe (N-β-hydroxybutyryl-phenylalanine).
2. BHB-Phe의 항비만 효과
   • 비만 마우스에게 BHB-Phe를 투여하면:
     • 시상하부(hypothalamus)와 뇌줄기(brainstem)의 특정 신경세포를 활성화
     • 식욕을 억제하고 음식물 섭취를 줄임
     • 체중 감소 효과를 나타냄
   • 이 효과는 기존의 알려진 식욕 조절 경로(예: GLP-1 등)와는 다른 방식으로 작동하는 것으로 보임.
3. 유전적 검증
   • CNDP2를 제거한 마우스(CNDP2-KO)에서는:
     • BHB-amino acids 생성이 거의 사라짐
     • 케톤 상태(ketosis)에서 오히려 음식 섭취가 늘고 체중이 증가함
   • 즉, 이 경로가 정상적으로 작동해야 케톤 상태에서 식욕 조절이 제대로 이뤄짐.
4. 인간과 보존성
   • CNDP2 의존적인 BHB-아미노산 생성 경로는 인간에서도 확인됨.
   • 따라서 이 경로가 인간의 에너지 균형과 체중 조절에도 관여할 가능성이 높음.

의의

• 케톤체(BHB)의 새로운 생리적 역할을 밝힌 중요한 연구.
• 케토제닉 다이어트나 간헐적 단식 시 나타나는 식욕 억제 효과의 한 가지 분자적 기전을 제시.
• 향후 BHB-Phe나 CNDP2를 표적으로 한 새로운 비만 치료제 개발 가능성을 열어줌.

쉽게 말하면, 

“케톤체(BHB)가 아미노산과 결합해서 만든 새로운 물질(BHB-Phe)이

뇌를 자극해 식욕을 떨어뜨리고 살이 찌는 것을 막아준다”는 내용입니다.

BHB와 아미노산 결합 기전 및 BHB-Phe 식욕억제 효과

1. BHB가 아미노산과 결합하는 기전 (BHB-ylation)

기존에 알려진 BHB 대사는 주

로 에너지 중간체(아세토아세테이트 등)와의 상호 전환이었습니다.

이번에 밝혀진 새로운 경로는

이차 대사(shunt pathway)로, 다음과 같이 진행됩니다.

핵심 효소: CNDP2 (Carnosine Dipeptidase 2)

• 원래 carnosine 분해 효소로 알려졌으나, 새로운 기능이 확인됨.
• BHB의 카르복실기(-COOH)와 아미노산의 아미노기(-NH₂)를 효소적으로 결합(conjugation)시켜 N-β-hydroxybutyryl-아미노산 (BHB-amino acid)을 생성.

반응 개요:

text

BHB + 유리 아미노산 → (CNDP2 촉매) → BHB-아미노산 + H₂O

• 가장 풍부하게 생성되는 것은 BHB-Phe (N-β-hydroxybutyryl-phenylalanine)
• 다른 아미노산(Leu 등)과도 결합 가능 → BHB-amino acid family 형성

• 유도 조건: 케톤증(ketosis) 상태에서 강하게 유도됨 (케톤 생성 식이, 케톤 에스터 보충 시).
• 보존성: 마우스와 인간 모두에서 확인됨.

CNDP2를 유전자 제거(KO)하면 조직의 BHB-ylation 활성이 완전히 사라지고, BHB-아미노산 수치가 크게 감소합니다.

2. BHB-Phe의 식욕억제 효과

• 구조·기능적 유사체: Lac-Phe (lactate + phenylalanine)의 케톤 버전.
• 작용 부위:
  • 시상하부(hypothalamus)와 뇌간(brainstem)의 특정 뉴런을 활성화.
  • 이를 통해 섭식 행동 억제.
• 실험 결과 (비만 마우스 모델):
  • BHB-Phe 투여 시 → 음식 섭취량 감소 + 체중 증가 억제.
  • 에너지 소비(운동량, 산소 소비 등)에는 큰 영향 없음 → 주로 식욕 억제를 통해 항비만 효과 발휘.
  • CNDP2-KO 마우스: 케톤 에스터나 케톤 생성 식이 후에도 식욕이 잘 억제되지 않고, 체중이 더 많이 증가.

→ BHB-Phe는 케톤증 시 자연적으로 생성되어 에너지 균형 조절에 기여하는 생리활성 대사산물입니다.

3. 2023/2024 Cell 논문 이후 후속 연구

원논문은 Moya-Garzon et al., Cell (2025년 1월호, 온라인 2024년 11월) “A β-hydroxybutyrate shunt pathway generates anti-obesity ketone metabolites” 입니다. (사용자가 말씀하신 2023 Cell은 아마 Lac-Phe 관련 이전 연구나 시점 혼동으로 보입니다.)

후속 논문/관련 연구 현황 (2025~2026 초 기준):

• 아직 직접적인 대규모 후속 실험 논문은 많지 않습니다.
• 주로 리뷰·논평 형태로 등장:
  • Cell Metabolism 등에서 “Shunt-ing down obesity…” 식의 하이라이트/해설 논문.
  • 인간에서의 보존성과 비만 치료 가능성에 대한 논의가 활발.
• 관련 후속 방향:
  • CNDP2 유전형과 비만 관련성 조사
  • BHB-Phe의 정확한 수용체/신호전달 경로 규명
  • 외인성 BHB-Phe 또는 CNDP2 조절을 통한 치료제 개발 가능성

현재로서는 이 Cell 논문이 해당 경로를 처음 보고한 원천 논문이며, 후속 연구는 아직 초기 단계입니다

A β-hydroxybutyrate shunt pathway generates anti-obesity ketone metabolites

Maria Dolores Moya-Garzon1,6,8,17 ∙ Mengjie Wang15,17 ∙ Veronica L. Li1,2,6,8,17 ∙ … ∙ Steven M. Banik2,6 ∙ Yong Xu15 Send email to yongx@bcm.edu ∙ Jonathan Z. Long1,6,7,8,16,18 Send email to jzlong@stanford.edu … Show more

•  

Show Outline

Highlights

•

A ketone shunt derivatizes β-hydroxybutyrate by conjugation with amino acids

•

BHB-amino acids are endogenous mouse and human metabolites

•

BHB-Phe administration decreases food intake and body weight in obese mice

Summary

β-Hydroxybutyrate (BHB) is an abundant ketone body. To date, all known pathways of BHB metabolism involve the interconversion of BHB and primary energy intermediates. Here, we identify a previously undescribed BHB secondary metabolic pathway via CNDP2-dependent enzymatic conjugation of BHB and free amino acids. This BHB shunt pathway generates a family of anti-obesity ketone metabolites, the BHB-amino acids. Genetic ablation of CNDP2 in mice eliminates tissue amino acid BHB-ylation activity and reduces BHB-amino acid levels. The most abundant BHB-amino acid, BHB-Phe, is a ketosis-inducible congener of Lac-Phe that activates hypothalamic and brainstem neurons and suppresses feeding. Conversely, CNDP2-KO mice exhibit increased food intake and body weight following exogenous ketone ester supplementation or a ketogenic diet. CNDP2-dependent amino acid BHB-ylation and BHB-amino acid metabolites are also conserved in humans. Therefore, enzymatic amino acid BHB-ylation defines a ketone shunt pathway and bioactive ketone metabolites linked to energy balance.

β-Hydroxybutyrate(BHB)는

풍부한 케톤체이다.

현재까지 알려진 모든 BHB 대사 경로는

BHB와 주요 에너지 중간체 간의 상호 전환을 포함한다.

여기서 우리는

CNDP2 의존적 효소 반응을 통해 BHB와 유리 아미노산이 결합하는,

이전에 알려지지 않았던 BHB 이차 대사 경로를 확인하였다.

이 BHB 분지 경로는

항비만 케톤 대사산물 가족인 BHB-아미노산을 생성한다.

마우스에서

CNDP2를 유전적으로 제거하면 조직의 아미노산 BHB화(BHB-ylation) 활성이 사라지고 B

HB-아미노산 수치가 감소한다.

가장 풍부한 BHB-아미노산인

BHB-Phe는 케톤증에 의해 유도되는 Lac-Phe의 유사체로서,

시상하부와 뇌간 뉴런을 활성화하고 섭식을 억제한다.

반대로 CNDP2-KO 마우스는 외인성 케톤 에스터 보충이나 케톤 생성 식이 후 음식 섭취량과 체중이 증가한다.

CNDP2 의존적 아미노산 BHB화와 BHB-아미노산 대사산물은 인간에서도 보존되어 있다.

따라서

효소적 아미노산 BHB화는 케톤 분지 경로를 정의하며,

에너지 균형과 관련된 생리활성 케톤 대사산물을 나타낸다.

Graphical abstract

Keywords

1. BHB
2. metabolomics
3. enzyme
4. obesity
5. ketone
6. metabolite

Introduction

Mammals have evolved complex nutrient-responsive pathways that link the availability of external energy sources to internal metabolic homeostasis. These pathways involve changes to cellular energy metabolites, which function both as metabolic fuels and as downstream effectors. A key example is the metabolite β-hydroxybutyrate (BHB), a ketone body whose levels rise during periods of low carbohydrate availability, such as during starvation, intermittent fasting, or with consumption of a ketogenic diet.1,2 BHB is used as a metabolic fuel that can be oxidized by metabolic tissues like the brain and heart for ATP production. In addition, BHB is also a signaling molecule that can activate G-protein-coupled receptors,3,4 post-translationally modify proteins,5 or inhibit nuclear histone deacetylases6 to alter cellular and metabolic processes.

The classical primary metabolic pathways that mediate hepatic BHB production and extrahepatic BHB oxidation are well established. In hepatic ketogenesis, fatty acid oxidation leads to the production of acetyl-coenzyme A (CoA), which, via the sequential enzymatic action of HMGCS2 (3-hydroxymethylglutaryl-CoA synthase 2), HMGCL (3-hydroxy-3-methylglutaryl-CoA lyase), and BDH1 (3-hydroxybutyrate dehydrogenase 1), ultimately results in generation of BHB.7 Once produced and exported into the circulation,8 BHB can then be taken up by extrahepatic tissues, where it undergoes oxidation via BDH1, SCOT (succinyl-CoA:3-ketoacid CoA transferase), and the TCA (tricarboxylic acid) cycle.9 Importantly, all known metabolic pathways of BHB involve the same metabolic interconversions of BHB to primary intermediates that are directly used for ATP production. Metabolic pathways of BHB outside of primary metabolism have not been reported to date.

Jansen et al. previously showed that CNDP2 (carnosine dipeptidase 2) catalyzes the condensation of lactate and amino acids in vitro,10 and we established this biochemical pathway to be a physiologically relevant synthetase reaction in vivo.11 The most abundant member of the N-lactoyl amino acids, N-lactoyl-phenylalanine (Lac-Phe), is an exercise- and metformin-inducible metabolite that suppresses food intake and body weight.12,13 These data show that the metabolic pathways of lactate extend beyond glycolysis and primary metabolism and include secondary metabolic shunt pathways that produce lactate-derived signaling metabolites. From a chemical perspective, BHB and lactate exhibit a high degree of structural similarity: both are hydroxy fatty acids that differ only by a single methylene. We therefore considered the possibility that BHB, like lactate, might also be enzymatically conjugated to amino acids. This predicted metabolic pathway would represent a previously unknown pathway of BHB secondary metabolism and produce a class of orphan metabolites, the N-β-hydroxybutyryl amino acids (BHB-amino acids). However, we were unable to find any evidence for enzymatic BHB-ylation of free amino acids in the published literature. We also were unable to find prior annotation of any BHB-amino acids as endogenous metabolites in public databases such as METLIN,14 the Human Metabolome Database (HMDB),15 or Global Natural Product Social Molecular Networking (GNPS).16

Here, we show that the enzyme CNDP2 catalyzes BHB-ylation of free amino acids in vitro and in vivo. The product of this biochemical reaction, the BHB-amino acids, are endogenously present in mouse and human plasma and exhibit ketosis inducibility and genetic regulation. N-β-hydroxybutyryl phenylalanine (BHB-Phe), the most abundant BHB-amino acid, is a structural and functional congener of Lac-Phe that suppresses feeding upon administration to obese mice. Conversely, CNDP2-knockout (KO) mice exhibit increased food intake and body weight on a ketogenic diet or following ketone ester administration. Lastly, CNDP2-mediated amino acid BHB-ylation and the BHB-amino acid metabolites are conserved in humans. These data establish a BHB metabolic shunt pathway linked to energy balance.

ResultsCNDP2 catalyzes BHB-ylation of amino acids in vitro

To determine whether CNDP2 can catalyze the BHB-ylation of amino acids in vitro (Figure 1A), we incubated BHB and phenylalanine (20 mM each) with cell lysates from HEK293T cells that were transiently transfected with FLAG-tagged mouse CNDP2 or GFP control. Overexpression‎ of CNDP2 protein was confirmed by western blotting (Figure S1A). After incubation at 37°C for 1 h, we used liquid chromatography-mass spectrometry (LC-MS) to measure the expected condensation product, BHB-Phe. As shown in Figure 1B, CNDP2-transfected cell lysates exhibited >140-fold greater phenylalanine BHB-ylation activity compared with GFP-transfected cell lysates. CNDP2 can therefore synthesize BHB-Phe, in addition to its previously reported Lac-Phe synthesis activity. We performed several control experiments to examine the substrate specificity of the CNDP2-dependent BHB-ylation reaction. CNDP2 exhibited similar rates of amino acid BHB-ylation and N-lactoylation but could not accept shorter (e.g., acetate, C2) or longer (e.g., octanoate, C8) organic acids as substrates (Figure 1C). On the amino acid side, CNDP2 exhibited the fastest BHB-ylation activity using phenylalanine as a substrate (Figure 1D). Some other hydrophobic amino acids were also accepted, but with lower activity (<20%) compared with phenylalanine, and no activity was observed with many of the amino acids tested (Figure 1D). CNDP2 did not exhibit BHB-ylation of the free N terminus of a peptide substrate (Figure S1B). CNDP2 was also unable to accept the BHB isomer 3-hydroxyisobutyrate (3-HIB) acid as a substrate, though we did observe minor CNDP2 condensation activity with 2-hydroxybutyrate (2-HB) (Figure S1C).17,18

Figure 1 CNDP2 catalyzes amino acid BHB-ylation in vitro

Show full captionFigure viewer

Figure S1 Additional in vitro characterization of CNDP2 activity, related to Figure 1

Show full captionFigure viewer

To quantitatively compare the kinetics of CNDP2-dependent BHB-ylation and N-lactoylation, we generated purified recombinant mouse CNDP2-FLAG protein for in vitro kinetic assays. The resulting kinetic data, which were fit to Michaelis-Menten kinetics, revealed that BHB is in fact a higher-affinity substrate for the CNDP2 active site than lactate (Km for BHB = 8.8 mM; Km for lactate = 33.4 mM, Figures 1E and 1F). By contrast, CNDP2-dependent Lac-Phe synthesis is faster than that of BHB-Phe synthesis under conditions of saturating substrate concentrations (Vmax for lactate = 62.7 nM/min/mg, Vmax for BHB = 3.5 nM/min/mg, Figures 1E and 1F).

Lastly, we docked the product BHB-Phe into the CNDP2 active site (Figure 1G; see STAR Methods). This modeling predicted several interactions of BHB-Phe with key active site residues, including E166, D195, and H455. We generated single-point mutations of each of these residues by transient transfection to HEK293T cells (Figure S1D). In each case, mutations of any of these active site residues concomitantly reduced both CNDP2-dependent BHB-Phe and Lac-Phe production (Figures 1H and 1I). Therefore, BHB-ylation and N-lactoylation activities are both entirely encoded within the CNDP2 polypeptide. In addition, these two catalytic activities cannot be readily dissociated by single-point mutations in the active site pocket.

CNDP2-dependent BHB-ylation in mouse tissues

To determine whether endogenous CNDP2 catalyzes amino acid BHB-ylation in mouse tissues, we first examined the tissue expression‎ of CNDP2 by western blot using an anti-CNDP2 antibody. The specificity of this antibody was confirmed using tissues from CNDP2-KO mice (Figure S2A). CNDP2 protein levels were highest in kidney and gut and lower in many of the other tissues examined (Figure 2A). Next, we performed in vitro BHB-Phe synthesis activity using crude total lysates of kidney, gut, brain, liver, and quadriceps tissues from wild-type (WT) mice (see STAR Methods). We selected these tissues because of their wide range of CNDP2 protein expression‎. As shown in Figure 2B, the highest BHB-Phe synthesis activity was observed in kidney and gut (∼10–30 pmol/min/mg), whereas brain, liver, and quadriceps exhibited lower, but detectable, BHB-Phe synthesis activity (∼1–5 pmol/min/mg). This pattern of BHB-Phe synthesis across tissues largely paralleled the protein expression‎ of CNDP2 in these same tissues. In addition, the temperature dependence of the renal BHB-Phe synthesis activity also paralleled that of recombinant CNDP2 protein (Figures S2B and S2C). Therefore, the tissue expression‎ pattern and temperature profile of CNDP2 protein both correlate with the tissue BHB-Phe synthesis activity.

Figure 2 CNDP2 is the principal BHB-amino acid synthetase in mouse tissues

Show full captionFigure viewer

Figure S2 Additional characterization of tissue CNDP2 activity, related to Figure 2

Show full captionFigure viewer

We used tissues from CNDP2-KO mice to determine the contribution of CNDP2 to the tissue BHB-Phe synthesis activity. As shown in Figure 2B, both kidney and gut BHB-Phe synthesis activity was largely abolished (>95% reduced) in tissues from CNDP2-KO mice. The smaller BHB-Phe synthesis activity in other tissues was also greatly diminished (>85% reduced in brain, >75% reduced in liver, and >60% reduced in quadriceps). Using leucine or valine as substrates for in vitro BHB-ylation, a similar pattern of BHB-Leu and BHB-Val synthesis across WT and CNDP2-KO tissues was observed (Figures 2C and 2D): kidney and gut tissues from WT mice both exhibited the highest BHB-amino acid synthesis activity, and this activity was largely abolished in tissues from CNDP2-KO mice. In additional control experiments, we assayed tissue Lac-Phe synthesis activity, which again exhibited a similar pattern and CNDP2 dependence to that of BHB-amino acid synthesis (Figure 2E). By contrast, carnosine hydrolysis across tissues exhibited a distinct pattern, with highest activity in liver and quadriceps and little activity in the kidney, gut, and brain (Figure 2F). Importantly, the carnosinase activity was not altered in CNDP2-KO tissues (Figure 2F). We conclude that CNDP2 is the principal enzyme responsible for BHB-amino-acid-synthesis activity in mouse tissues. In addition, despite its previously annotated in vitro activity, CNDP2 is not a major tissue carnosinase.

BHB-amino acids are endogenous mouse metabolites

The product of the CNDP2-catalyzed BHB-ylation reaction, BHB-amino acids, have not been previously reported as endogenous metabolites. We therefore developed a targeted metabolomics approach to determine whether BHB-amino acids can be detected in mouse plasma. We first synthesized an authentic BHB-Phe standard by classical amide coupling between BHB and phenylalanine (Figure 3A; see STAR Methods). Fragmentation of the authentic BHB-Phe standard revealed a major daughter ion corresponding to Phe (m/z = 164) and a second, smaller daughter ion corresponding to the decarboxylation product (m/z = 206). Next, we developed a targeted multiple reaction monitoring (MRM) method on a high-performance liquid chromatography coupled to triple quadrupole mass spectrometry (QQQ-LC/MS) to monitor the parent to phenylalanine transition for BHB-Phe. In mouse plasma, we identified an endogenous peak that eluted at an identical retention time with the authentic standard (Figure 3A). To exclude the possibility that this method may also be detecting isobaric 2-HB- and 3-HIB-phenylalanine isomers, we synthesized authentic standards of 2-HB-Phe and 3-HB-Phe and developed MRM methods that could distinguish between each of the three molecules. Although 2-HB-Phe and 3-HIB-Phe also yielded daughter ions corresponding to phenylalanine, we also identified unique transitions for each of these isomers (2-HB-Phe: 260 > 102, fragmentation at N-Cα; 3-HIB-Phe: 250 > 220, loss of CH3O, Figures S3A and S3B). For BHB-Phe, >99% of the total signal was detected using the 250 > 164 transition, 2-HB-Phe was detected using both 260 > 102 and 250 > 164 (in a 6:1 ratio), and 3-HIB-Phe was detected using both 250 > 220 and 250 > 164 (in a 1.6:1 ratio) (Figure S3C). The signal from the endogenous peak was found to be comprised >99% of the 260 > 164 transition (Figure S3C). Therefore, the endogenous signal is BHB-Phe; in addition, 2-HB-Phe and 3-HIB-Phe are not endogenous metabolites.

Figure 3 Detection and ketosis inducibility of BHB-amino acids in mouse plasma

Show full captionFigure viewer

Figure S3 Additional characterization of mouse metabolites, related to Figure 3

Show full captionFigure viewer

We next synthesized authentic standards for BHB-Leu, BHB-Val, and BHB-Met, which also exhibited the same characteristic amino acid daughter ion (Figures 3B–3D). Using a similar MRM approach, we also detected endogenous peaks with transition and retention time identical to that of the authentic standards (Figures 3B–3D), demonstrating that these other BHB-amino acids are also endogenous metabolites.

Because of their biosynthetic origin from BHB, circulating BHB-amino acids would be predicted to rise with increasing BHB levels, such as those achieved by nutritional or physiologic ketosis. Levels of BHB-amino acids in mouse blood plasma were therefore measured after 1 week of ketogenic diet, a 24-h fast, or oral administration of a ketone ester drink (3 g/kg of body weight). We confirmed that plasma BHB levels were elevated by each of these conditions (Figure S3D). All these stimuli consistently produced robust 2- to 10-fold elevations in each of the BHB-amino acids (Figures 3E–3H). Ketogenic diet produced greater variation in induction of BHB-amino acid levels, which may reflect the more chronic nature of this perturbation (1 week) compared with the two other acute ketosis stimuli (≤24 h). The levels of phenylalanine, Lac-Phe, and lactate were not consistently changed across the three ketosis stimuli (Figure S3D). Tissue levels of BHB-Phe were also detectable and elevated after ketone ester oral gavage (Figure S3E). We conclude that BHB-amino acids are endogenous, ketosis-inducible mouse metabolites.

Genetic regulation of BHB-amino acids by CNDP2 and HMGCL

Figure 4A shows a schematic of the CNDP2-dependent ketone metabolic shunt pathway in the context of known ketogenesis and ketolysis pathways. To directly test the physiologic contribution of CNDP2 to BHB-amino acid biosynthesis, we measured BHB-amino acids in blood plasma from CNDP2-KO mice after a ketone ester drink challenge or after 1 week of ketogenic diet. After acute administration of ketone ester drink, CNDP2-KO mice exhibited >90% depletion of multiple plasma BHB-amino acids (Figure 4B). Reductions in multiple BHB-amino acids were also observed in CNDP2-KO mice after 1 week on a ketogenic diet (Figure 4C). In both experiments, levels of plasma BHB, lactate, or phenylalanine were not consistently changed between WT and CNDP2-KO mice, and levels of Lac-Phe were, as expected, reduced in CNDP2-KO mice (Figures S4A and S4B).

Figure 4 Genetic regulation of BHB-amino acids by CNDP2 and HMGCL

Show full captionFigure viewer

Figure S4 Additional characterization of mouse metabolites in the genetic models, related to Figure 4

Show full captionFigure viewer

We performed additional metabolomic profiling of organic acid-amino acid conjugates in CNDP2-KO mice. N-acetyl-Phe, N-acetyl-Val, N-acetyl-Leu, and N-acetyl-Met, as well as their corresponding free amino acids, were not changed between WT and CNDP2-KO mice (Figure S4C). We were unable to detect N-propyl, N-butyryl, or N-octanoyl-amino acids (C3, C4, and C8, respectively, Figure S4C), likely reflecting the low circulating abundance of the corresponding organic acids compared with acetate, lactate, and BHB.

Next, we examined the effects of liver-specific deletion of HMGCL, a critical upstream enzyme in hepatic ketogenesis (Figure 4A). We obtained plasma from liver-specific KOs of HMGCL (Alb-Hmgcl−/− mice), which were previously generated by crossing Albumin-cre mice with Hmgclfl/fl mice.7 Several BHB-amino acids, such as BHB-Met, BHB-Leu, and BHB-Val, but not BHB-Phe, were reduced by ∼50%–80% in plasma from these animals (Figure 4D). We confirmed ∼30% reductions in BHB levels in plasma from Alb-Hmgcl−/− mice and, in addition, found no changes in lactate, Lac-Phe, or phenylalanine levels compared with Hmgclfl/fl controls (Figure S4D). Together, these data establish the genetic and biochemical requirement for two upstream enzymes, HMGCL and CNDP2, in the regulation of circulating BHB-amino acids levels.

Role of BHB-Phe in feeding behaviors and body weight regulation

BHB-Phe is the most abundant BHB-amino acid (Figure 3E). This metabolite is a congener of Lac-Phe that shares both chemical similarity as well as a common biosynthetic pathway via CNDP2. We therefore considered the possibility that BHB-Phe might also be functionally similar to Lac-Phe and regulate food intake and body weight. We first used gain-of-function approaches to determine whether BHB-Phe is sufficient to reduce food intake and body weight. In an initial study of diet-induced obese (DIO) mice in metabolic chambers, BHB-Phe (50 mg/kg, intraperitoneally [i.p.]) reduced food intake without affecting movement, oxygen consumption, or carbon dioxide production (Figures 5A–5E). A reduction in respiratory exchange ratio (RER) was also observed, consistent with a suppression of food intake (Figure 5E). Under these conditions, plasma BHB-Phe levels peaked at ∼20 μM at the 1-h time point and returned to baseline values by 3 h (Figure S5A). In an independent experiment in home cages, we found that acute administration of BHB-Phe (50 mg/kg, i.p.) to DIO mice suppressed food intake without affecting water intake, establishing specific suppression of food vs. fluid ingestion (Figure S5B). Lastly, plasma levels of other feeding-regulating hormones, such as ghrelin, leptin, and GDF15, were also unaltered in mice following a single administration with BHB-Phe (50 mg/kg, i.p., Figure S5C).

Figure 5 BHB-Phe suppresses food intake and body weight

Show full captionFigure viewer

Figure S5 Additional metabolic characterization of mice treated with BHB-Phe and CNDP2-KO mice, related to Figure 5

Show full captionFigure viewer

We performed chronic studies of BHB-Phe in DIO mice. Daily administration of BHB-Phe (50 mg/kg/day, i.p.) resulted in a durable suppression of daily food intake and, as expected, a concomitant reduction in body weight gain (Figures 5F and 5G). At the end of the experiment, BHB-Phe-treated mice exhibited reductions in aspartate aminotransferase (AST), alanine aminotransferase (ALT), and total triglycerides (TGs); no changes were found in high-density lipoprotein (HDL)- or low-density lipoprotein (LDL)-cholesterol (Figure S5D). BHB-Phe-treated mice lost the same amount of weight as pair-fed controls (Figure 5H), demonstrating that the observed suppression of food intake explains the observed change in body weight in BHB-Phe-treated mice.

To understand the structural requirements of BHB-Phe that are important for its body weight-lowering effects, we tested the effect of additional, structurally related metabolites. First, although BHB-Phe (50 mg/kg/day, i.p.) efficiency suppressed body weight and food intake in DIO mice, either BHB alone or phenylalanine alone at the same doses were without effect (Figure 5I). Second, the related metabolites BHB-Lys (50 mg/kg, i.p.), as well as the dipeptides Phe-Phe (50 mg/kg, i.p.) and Leu-Leu (50 mg/kg, i.p.), also failed to reduce food intake or body weight (Figures 5J and 5K). Third, we tested other CNDP2-regulated BHB conjugates, including BHB-Met, BHB-Leu, and BHB-Val. These metabolites consist of BHB conjugated to hydrophobic amino acids; notably, two of them, BHB-Leu and BHB-Val, represent BHB conjugated to branched-chain amino acids (BHB-BCAAs). In this case, we observed body-weight-lowering activity for each BHB-Met, BHB-Leu, and BHB-Val (Figure S5E). Therefore, BHB-Phe and other CNDP2-regulated BHB-hydrophobic amino acid conjugates, including BHB-BCAAs, have anorexigenic and anti-obesity effects in mice, whereas BHB conjugates to other essential amino acids such as BHB-Lys, as well as other dipeptides, do not exhibit the same bioactivity.

Lastly, we used CNDP2-KO mice to examine the physiologic contributions of BHB-amino acids to energy balance. We previously reported a gene-by-environment interaction of the Cndp2 gene and glycolytic stimuli: CNDP2-KO mice have normal body weights after a standard high-fat-diet feeding protocol; however, upon glycolytic stimulus challenge to increase Lac-Phe levels (by treadmill exercise or by metformin treatment), KO animals exhibit an increased food intake and body weight phenotype compared with WT controls.11,12 To determine the contribution of BHB-amino acids, in an initial experiment we administered ketone esters (3 g/kg/day, per os [p.o.]) to WT and CNDP2-KO mice that had been rendered obese by high-fat-diet feeding for 11–19 weeks. At the beginning of the experiment, body weights were not different between genotypes. By the end of the 12-day ketone ester treatment, WT mice on average lost −0.3 ± 0.4 g, whereas CNDP2-KO mice gained +1.0 ± 0.2 g (mean ± SEM, p < 0.05, Figure 5L). CNDP2-KO mice also exhibited greater cumulative food intake than WT mice (Figure 5M). In a second experiment, we placed WT and CNDP2-KO mice on a ketogenic diet. Initial body weights were once again not different between genotypes. By the end of the experiment, CNDP2-KO mice on a ketogenic diet gained more weight and ate more food than WT mice (Figure S5F). We confirmed that Lac-Phe and BHB-Phe were independently induced following sprint treadmill exercise and ketone ester or ketogenic diet treatment, respectively (Figure S5G). Therefore, ketone esters and a ketogenic diet represent environmental perturbations that uncover the effects of Cndp2 genotype on body weight and food intake.

BHB-Phe activates neural populations in the hypothalamus and brainstem

To better understand the neurobiological mechanisms by which BHB-Phe suppresses feeding, we first used pharmacological and genetic approaches to determine whether the effect of BHB-Phe might be mediated by hypothalamic melanocortin signaling, glucagon-like peptide-1 receptor (GLP-1R), or brainstem GDNF family receptor alpha-like (GFRAL) pathways. The effect of BHB-Phe (50 mg/kg, i.p.) on food intake and body weight was similar in WT or melanocortin 4 receptor (MC4R)-KO mice (Figures S6A and S6B). Similarly, the GLP-1R antagonist Exendin-3 (0.1 mg/kg/day, i.p.) efficiently blocked the anorexigenic and anti-obesity effects of GLP-1 (2 mg/kg/day, i.p., Figures S6C and S6D); however, under these same Exendin-3 did not alter the effect of BHB-Phe on food intake and body weight (Figures S6E and S6F). We obtained a neutralizing anti-GFRAL antibody19,20 (Eli Lilly clone 8A2) and verified that this antibody (10 mg/kg, subcutaneously [SQ]) completely blocked the activity of recombinant GDF15 to suppress food intake and body weight (4 nmol/kg, SQ) (Figure S6G); however, the effect of BHB-Phe on feeding and body weight was unaffected by anti-GFRAL antibody administration (Figures S6H and S6I). We conclude that the anorexigenic activity of BHB-Phe is independent of these known pathways of feeding control.

Figure S6 Role of MC4R, GLP-1R, and GFRAL pathways in the anorexigenic effects of BHB-Phe, related to Figure 6

Show full captionFigure viewer

Next, we used an activity-dependent genetic labeling strategy21 (TRAP, targeted recombination in active populations) to identify neurons activated following pharmacological dosing of BHB-Phe. In this approach, upon BHB-Phe treatment, c-Fos-dependent recombination of a reporter cassette (tdTomato) enables permanent genetic labeling of BHB-Phe. We treated TRAP2 mice (TRAP2/Rosa26-LSL-tdTomato) with BHB-Phe (50 mg/kg, i.p.), followed by 4-hydroxytamoxifen 30 min later, to initiate cre-dependent recombination in BHB-Phe-activated neurons. 2 weeks later, the same mice received Lac-Phe (50 mg/kg, i.p.) and were sacrificed 90 min later for c-Fos immunostaining (Figure 6A). This experimental approach therefore enables concurrent identification of both BHB-Phe-activated (e.g., TRAP+, marked by tdTomato) and Lac-Phe-activated (e.g., c-Fos+) neurons in the same animal. We examined multiple hypothalamic and brainstem regions for TRAP+ or c-Fos+ neurons. Compared with vehicle treatment, BHB-Phe and Lac-Phe both activated neuronal populations in multiple brain regions, including the paraventricular hypothalamic nucleus (PVH), the suprachiasmatic nucleus (SCN), the dorsomedial hypothalamic nucleus (DMH), the ventromedial hypothalamic nucleus (VMH), the arcuate nucleus of the hypothalamus (ARH), the lateral hypothalamus (LH), the lateral parabrachial nucleus (LPBN), and the nucleus of the solitary tract (NTS) (Figure 6B). Detailed analyses of TRAP+ neurons and c-Fos+ neurons in each of these regions revealed that only a small fraction (∼2%–30%) of neurons were activated by both BHB-Phe and Lac-Phe (e.g., double TRAP+/c-Fos+), whereas the vast majority of activated neurons were distinct (Figure 6C). Representative sections from the indicated regions are shown in Figure 6D. Because TRAP recombination and c-Fos immunoreactivity may have different sensitivities to label-activated neurons, we repeated this experiment in an independent cohort of TRAP2/Rosa26-LSL-tdTomato mice but now reversed the BHB-Phe/Lac-Phe sequence. In this reversed experiment, Lac-Phe-activated neurons were identified by TRAP recombination, whereas BHB-Phe-activated neurons were identified by c-Fos immunoreactivity (Figure S7A). Once again, although both Lac-Phe and BHB-Phe activated neural populations in the hypothalamus and brainstem (Figure S7B), detailed analysis of each region showed that the two populations of activated neurons were largely distinct (Figure S7C). We conclude that pharmacological administration of BHB-Phe activates several hypothalamic and brainstem regions implicated in feeding behaviors in a manner overlapping but distinct from that of Lac-Phe.

Figure 6 TRAP/c-Fos mapping of BHB-Phe- and Lac-Phe-activated neurons in the brain

Show full captionFigure viewer

Figure S7 Additional characterization of TRAP/c-Fos mapping BHB-Phe- and Lac-Phe-activated neurons in the brain, related to Figure 6

Show full captionFigure viewer

Conservation of CNDP2 enzyme activity and BHB-amino acids in humans

Lastly, we sought to understand the conservation of the CNDP2-dependent amino acid BHB-ylation pathway to humans. First, we obtained recombinant human CNDP2, which exhibited the expected in vitro BHB-ylation activity, using BHB and phenylalanine as substrates (Figure 7A). Michaelis-Menten kinetics using increasing concentrations of BHB also revealed similar substrate affinity and maximal velocity to the mouse CNDP2 enzyme (Figure 7B). Next, we identified three human cell lines expressing hCNDP2: U937 macrophage cells, Caco-2 gut epithelial cells, and PANC-1 pancreatic ductal cells. For each human cell line, we generated control and hCNDP2-KO lines via CRISPR-Cas9. Complete loss of hCNDP2 was validated by western blotting using our anti-CNDP2 antibody (Figures 7C–7E). Knockout of hCNDP2 in each cell line resulted in near complete ablation of cell lysate BHB-amino-acid-synthesis activity. These data show that an endogenous amino acid BHB-ylation activity is present in human cells and primarily mediated by CNDP2 (Figures 7C–7E). Lastly, to determine whether BHB-amino acids are endogenous human metabolites, we measured plasma levels of BHB-amino acids from a subset of participants in a trial of exogenous ketone supplementation.22 After fasting overnight (>8 h), participants consumed a ketone monoester drink (0.3 g/kg HVMN Ketone Ester) and plasma was collected 1 h later. BHB-Phe, BHB-Leu, BHB-Val, and BHB-Met were also detectable in baseline plasma samples and elevated after ketone ester drink (Figure 7F). As expected, levels of BHB were increased by the ketone ester drink, whereas levels of phenylalanine, Lac-Phe, and lactate remained unchanged (Figure 7G). Therefore, both CNDP2-mediated amino acid BHB-ylation and BHB-amino acid metabolites are conserved in humans.

Figure 7 Human CNDP2 activity and BHB-amino acids in human plasma

Show full captionFigure viewer

Discussion

Here, we show that CNDP2 controls a secondary pathway of BHB metabolism leading to the production of a family of BHB-derived metabolites, the BHB-amino acids. In addition, BHB-Phe, the most abundant BHB-amino acid, is a structural and functional congener of Lac-Phe that reduces food intake and body weight. These data establish that the biochemical pathways of BHB extend beyond primary metabolic intermediates and include BHB-derived signaling metabolites that regulate energy homeostasis.

That CNDP2 can accept either BHB or lactate as a substrate represents an unusual biochemical mechanism that directly couples a metabolic state with the production of bioactive metabolites. To the best of our knowledge, such a multi-functional, flux-dependent enzyme-coupling mechanism has not been previously described. One interpretation of these data is that CNDP2 functions a “sensor” of glycolytic or ketosis flux, depending on whether lactate or BHB levels are elevated. Interestingly, the effector is simply a metabolic derivative of the substrate and, therefore, represents one of the simplest models by which a signal can be converted to an effector.

The chemical logic of CNDP2-dependent BHB-amino acid biosynthesis mirrors that of other signaling molecule: in every case, lower-abundance bioactive species are produced from higher-abundance precursors. For instance, steroid hormones are produced from cholesterol, thyroid hormones are produced from tyrosine, prostaglandins are produced from fatty acids, and histamine is produced from histidine. Our data show that the pool of abundant precursors is not limited to amino acids, cholesterol, or lipids but can also include other abundant metabolic fuels such as BHB. In addition, this CNDP2-dependent mechanism is operational in Cndp2+ cells (e.g., macrophages, other immune cells, and epithelial cells of kidney and gut), which are cell types not classically associated with BHB metabolism.

Although past studies have reported the anorexigenic and anti-obesity effects associated with elevated ketones in mice23,24 and in humans,25 this is by no means a consolidated phenomenon. Our data demonstrate that the effects of BHB also extend to BHB-derived metabolites. Therefore, potential variations in levels of BHB-amino acids may be an important contributor to the conflicting associations of ketosis and energy balance reported in previous studies. Indeed, our own data demonstrate that changes in the circulating BHB-amino acid levels are correlated to, but not linearly determined by, changes in circulating levels of BHB itself; consequently, control for variation in BHB-amino acid levels (and potentially CNDP2 genotype) should be potentially considered in future studies of ketosis and obesity. From a teleological point of view, high ketone levels are a product of both increased hepatic ketogenesis and adipose lipolysis to provide fatty acid substrates. Consequently, a high ketone state demonstrates that sufficient adipose lipid stores are available for ketogenesis. Therefore, one potential interpretation of the association of high ketones and food intake suppression may be that ketones, and BHB-amino acids, signal a state of fat sufficiency. In addition, our data point to hypothalamic and brainstem neural populations as potential downstream targets of BHB-Phe.

Our studies here also expand our understanding of the gene-by-environment interactions in energy balance. Previously, we had shown that CNDP2-KO mice only exhibit a body weight phenotype following treadmill running or metformin treatment but not under “standard” high-fat-diet feeding conditions. Therefore, phenotypes associated with the Cndp2 gene are only revealed when the appropriate and specific environmental stimulus is provided. Our studies identify ketone ester administration and a ketogenic diet as environmental contributors to the CNDP2-dependent phenotype and suggest that additional nutritional or physiologic perturbations that increase ketogenesis may be relevant environmental stimuli that interact with the Cndp2 gene.

There are two reasons why BHB-amino acids were robustly detectable in our mass spectrometry analysis but not annotated in prior metabolomic studies. First, our enzymological studies of CNDP2, and the close chemical parallels between lactate and BHB, provided a compelling and directly testable biochemical hypothesis for the biosynthetic origins of BHB-amino acids. Second, our chemical synthesis of BHB-amino acid standards, which are otherwise not commercially available, enabled confirmation of the retention time and fragmentation of the endogenous peaks. Our strategy for detecting BHB-amino acids suggest that metabolome space might be more generally annotated by combining authentic metabolite standards with hypotheses about the chemical similarity of substrates and promiscuity of biochemical reactions.

Although our studies here only examined the role of BHB-amino acids in the context of energy homeostasis, the physiologic functions of BHB-amino acids may extend to other physiologic contexts as well. For instance, ketosis is being explored in a variety of other contexts, such as in neurodegenerative diseases,26 inflammation,27 muscle resilience,28 cancer treatment,29 and several other age-associated diseases. In addition, elevated BHB is observed in other pathophysiologic conditions, such as diabetic ketoacidosis. Our data demonstrate that BHB-amino acids are also produced when levels of BHB are high, raising the possibility that the effects of ketosis and BHB in these other contexts might also be, at least in part, mediated by concomitant production of BHB-amino acids.

Limitations of the study

There are four main limitations of this study. First, we show that BHB-Phe activates neural populations in the hypothalamus and brainstem. However, we do not further characterize the molecular identities of these neurons nor do we probe the functional consequences of their activation. In the future, such experiments would define which feeding-associated brain regions and cell types are important downstream effectors of the anorexigenic action of BHB-Phe. Second, our gain-of-function sufficiency studies with pharmacological administration of BHB-amino acids achieved supraphysiologic levels of metabolites in circulation. Although lower, more physiologically relevant concentrations were not tested here, such experiments may reveal more about the subtleties of how these ketone metabolites influence feeding and neural circuits in a physiological context. It is possible that lower, more physiologically relevant concentrations could selectively activate different subpopulations of neurons or produce more nuanced effects on feeding behavior or body weight. Such experiments might clarify the presence of potential dose-dependent “entourage effects,” where combinations of BHB-amino acids act synergistically at lower doses to modulate neural or metabolic responses. Third, we have not yet identified CNDP2 point mutants that can only accept either BHB or lactate as substrates. Such mutants would enable functional dissection of these multiple biochemical branches of CNDP2 activity in vitro and in vivo. Fourth, our study uses global CNDP2-KO mice, which does not enable specific assignment of cell types or tissues that contribute to total BHB-amino acid synthesis in vivo. The use of conditional CNDP2-KO mice might establish certain organs, such as the gut, which are more predominant contributors to whole-body BHB-amino acid synthesis.

Resource availabilityLead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jonathan Long (jzlong@stanford.edu).