Targeted proteomics reveal α-MSH-like Enterobacterial ClpB protein in gut microbiota of patients with anorexia nervosa Authors: Lisa Bremard, Lisa Wallart, Emilie Lahaye, Mohamed Amine Ben Mlouka, Jochen Seitz, Beate Herpertz-Dahlmann, Sergueï O Fetissov Published: 2025 ABSTRACT Abstractα-melanocyte-stimulating hormone (α-MSH) is an anorexigenic peptide. In this exploratory study, a targeted proteomic approach was used to detect α-MSH-like bacterial proteins in the gut microbiota of patients with anorexia nervosa (AN) and healthy controls. Enterobacteriaceae chaperone DnaK was identified as an α-MSH-like protein common to both study groups, while caseinolytic protease B (ClpB) was present in patients with AN. These data further link anorexigenic ClpB to the AN pathophysiology. Subject terms: Diseases, Microbiology Main text Anorexia nervosa (AN) is a psychiatric disease of unknown origin characterized by several psychological, behavioral, and metabolic impairments leading to insufficient food intake and weight loss1. Absence of efficient pharmacological treatment emphasizes the need for a better understanding of the molecular mechanisms underlying the AN pathophysiology2. Our approach to this problem is focused on the research of possible mechanisms leading to the overactivation of the melanocortin (MC) peptide signaling, a key anorexigenic system. Indeed, α-melanocyte-stimulating hormone (α-MSH) is a neuropeptide of the MC family which inhibits feeding by stimulation of the MC type 4 receptor (MC4R)3. While deficiency of the α-MSH/MC4R signaling, ex. Due to mutations of the α-MSH precursor proopiomelanocortin (pomc) or the mc4r gene are known to cause hyperphagia and obesity in humans4,5, these genes are not significantly affected in AN6. An alternative mechanism leading to enhancement of the α-MSH/MC4R signaling may involve α-MSH-reactive immunoglobulins G (IgG) antibodies, which play a role of an α-MSH carrier protein in the blood7. Indeed, IgG from patients with AN was able to enhance the MC4R signaling in vitro8. Importantly, plasma levels of α-MSH-reactive IgG correlate positively with the core psychopathological traits in patients with AN9. Determining the antigenic origin of α-MSH-reactive IgG may, hence, provide the cues for a better understanding of the AN pathophysiology. An emerging field of research links gut microbiota with the immune system and the brain function and behavior in normal and pathological conditions10. Among the multiple roles, gut microbiota represents a constitutive source of numerous microbial antigens physiologically interacting with the host immune system, which is altered in AN11. Caseinolytic protease B (ClpB), produced by Escherichia coli (E. coli), was earlier identified as an antigen mimetic of α-MSH12. This identification was achieved by applying a targeted proteomic approach to E. coli K12, a laboratory strain of the common Enterobacteriaceae family of gut bacteria. However, it is still unknown if ClpB can be detected by proteomics in human gut microbiota and if other α-MSH-like bacterial proteins can be present in microbiota of both healthy subjects and patients with AN. Thus, in the present exploratory study, involving a small number of study subjects, we applied a targeted proteomic approach to detect and identify α-MSH-like bacterial proteins in fecal microbiota samples of adolescent patients with AN and in age-matched healthy controls (HCs), both females. By studying adolescents, we expected to reveal early putative changes of their gut microbiota, which should be less influenced by adaptation to starvation and medications in adult patients. A small number of study participants was sufficient to answer our question if α-MSH-like bacterial proteins can be detected by targeted proteomics in human gut microbiota. For this purpose, total bacterial protein was extracted from individually cultured fecal samples of 6 patients with AN and 6 HCs, separated using 2-dimensional gel electrophoresis (2-DGE) and transferred to a nitrocellulose membrane, which was revealed using polyclonal anti-α-MSH antibodies. This procedure resulted in the detection of several α-MSH-like immunopositive protein spots in all individual fecal protein samples of both study groups (Fig. 1). By comparing these results between the study groups, it was noticed that both patients with AN and HCs display a similar pattern of α-MSH-like immunoreactivity with protein spots present at about 70, 25, and 15 kDa (Fig. 1). However, in 3 out of 6 samples of patients with AN, an additional spot at around 100 kDa was detected, which was not present in any of the control samples (Fig. 1). Immunoblots of fecal bacterial proteins separated by 2-DGE and revealed with α-MSH antibodies show the presence of α-MSH-immunopositive protein spots. Left 2 columns (a–g) show individual samples from all 6 studied HCs, and right 2 columns (g–l) from all 6 studied patients with AN. Note the consistent presence of α-MSH-positive spots at about 70, 25, and 15 kDa in both HCs and patients with AN. A specific to patients with AN α-MSH-positive spot of about 100 kDa was detected in microbiota samples of 3 patients (surrounded in red in h, j, l). Non-adjusted images were cropped on the right side, which did not contain any signal. Uncropped images are shown in the supplementary Fig. 1. For the identification of α-MSH-like bacterial proteins, we selected 3 samples from HCs (a, b, c in Fig. 2) and 3 samples from patients with AN (h, j, l in Fig. 2) displaying common to both AN and HCs spots which are marked as spot #2 ( ~ 70 kDa), #3 (~25 kDa) and #4 (~15 kDa) surrounded in yellow in Fig. 2a, b as well as the AN-specific spot #1 (~100 kDa) surrounded in red in Fig. 2b. The areas of the polyacrylamide gels corresponding to these α-MSH-positive spots (Fig. 2c, d) were dissected and used for the mass-spectrometry (MS) identification of their protein content. α Mass spectrometry identification of α-MSH immunopositive spots. Immunoblots revealed with α-MSH antibodies (upper row) and corresponding silver-stained 2-DGE gels (lower row) from a healthy control (a, c) and a patient with AN (b, d) were used for mass spectrometry identification of the consistently detected α-MSH-immunopositive spots, which were numbered: #1,2,3, and 4. The common spots between HC and AN were surrounded by yellow, and the AN-specific spot by red. No adjustments to the blots and gels were made. Non-labeled images are shown in the Supplementary Fig. 2. The MS revealed the presence of several proteins in each analyzed spot. Table 1 illustrates the most probable target proteins characterized by the best peptide coverage, high score for protein identification, as well as molecular weight matching. The full list of identified proteins is available in the supplementary data file labeled with the same letter for a study subject as in Fig. 1 and for the protein spot # as in Fig. 2. Table 1Occurrence of predicted bacterial proteins in α-MSH-positive protein spots identified by MS with a reference to protein identification score, coverage and molecular weight (MW) in fecal samples from HCs and patients with ANSpot #Protein nameHighest score-10 log(P)Highest coverage (%)Protein MW (kDa)Occurrence in controls (n = 6)Occurrence in patients (n = 6)#1Chaperone protein ClpB475.718095.50/63/6#2Chaperone protein DnaK512.049869.16/66/6#3Triosephosphate isomerase, chaperone DnaK fragment333.786426.95/66/6#4Autonomous glycyl radical cofactor, chaperone DnaK fragment280.849114.35/65/6 Occurrence of predicted bacterial proteins in α-MSH-positive protein spots identified by MS with a reference to protein identification score, coverage and molecular weight (MW) in fecal samples from HCs and patients with AN The MS analysis revealed that the most probable α-MSH-like bacterial protein common to both HCs and patients with AN appearing in the ~70 kDa spot (#2) is the 69.1 kDa Enterobacteriaceae chaperone protein DnaK, which was consistently identified in all tested samples. The amino acid sequence alignment between E. coli DnaK and α-MSH revealed the presence of a fragment with the discontinuous sequence homology (Supplementary Fig. 3). We also noticed that this spot contained, albeit with a low % coverage, the Enterobacteriaceae elongation factor G, a 77.6 kDa protein which was previously identified to contain a melanocortin-like peptide sequence having anti-inflammatory properties13. In the ~25 kDa spot (#3), the highest score was assigned to 26.9 kDa triosephosphate isomerase from E. coli; however, all tested samples of this spot also contained DnaK, likely due to its shorter fragment. For the 15 kDa spot (#4), several high-scoring proteins were found, ex. 14.3 kDa autonomous glycyl radical cofactor. However, this protein was not a consistent finding among different samples. Instead, DnaK fragments were consistently detected in the 15 kDa spot. In all 3 samples from patients with AN, a specific to the AN group, α-MSH-positive spot of ~100 kDa (#1) was found to contain the 95.5 kDa chaperone ClpB as the most likely target protein. Thus, using a targeted proteomic approach, our study revealed that gut bacteria in both healthy subjects and in patients with AN produce several proteins characterized by immunoreactivity for α-MSH. We can call such proteins as α-MSH-like, assuming that this connotation reflects their structural homology with α-MSH. The protein candidates have been identified using mass spectrometry, i.e., our technological approach was able to generate molecular targets for further functional validation in both in vitro and preclinical animal studies, as well as in larger cohorts of patients with AN, including both adolescents and adults as potential biotherapeutics and/or disease biomarkers. Such an approach for the gut microbiota analysis has not yet been reported and may contribute to the field of the ‘omics’ technologies applied to various pathological conditions. It can be considered as a subtype of the metaproteomic analysis of gut microbiota, which provides a large spectrum of data characterizing the composition and metabolic functions of the total bacterial proteome14. Instead, the targeted approach, applied here, indicates the presence of “moonlight” properties of proteins distinct from their functions in bacteria. Here, a moonlight property is based on the structural homology between bacterial proteins and a host regulatory peptide. In the present study, which involves patients with an eating disorder, we focused on α-MSH, a neuropeptide playing a key role in appetite regulation. The present data, by showing a reproducible α-MSH-like pattern among all study subjects, indicate a constitutive production of α-MSH-like bacterial proteins in human gut microbiota. Indeed, both HC and patients with AN displayed 3 main α-MSH-immunopositive protein spots, which contained the chaperone DnaK or its shorter fragments. Thus, Enterobacteriaceae DnaK was identified as a molecular target for its future validation as an α-MSH-like protein. DnaK might likely play a role as an antigen mimetic of α-MSH, stimulating α-MSH cross-reactive IgG, but it will not directly activate MC4 receptor because its α-MSH-like fragment (Supplementary Fig. 1) does not contain the α-MSH pharmacophore amino acid sequence (Arg-Trp), necessary for the receptor activation15. On the contrary, the α-MSH -like fragment in the Enterobacteriaceae ClpB protein contains such a sequence16. Indeed, the full-size ClpB protein is able to activate MC4R with a similar to α-MSH potency (non-published). MC4R, the key satietogenic MC receptor, is expressed in the gut enteroendocrine cells producing peptide YY (PYY), which can transmit the enhancing effects of ClpB on the intestinal satiety pathways mediated by this satietogenic hormone17. Increased blood level of PYY is consistently found in patients with AN18. The identification of ClpB as an α-MSH-like protein specific to patients with AN gives new experimental evidence of its relevance to the AN pathophysiology16. Nevertheless, it is surprising that in spite of the ubiquitous presence of Enterobacteriaceae in the human gut and, therefore, production of ClpB protein by these bacteria, it was not detected in all tested fecal samples, including the control group. It is presently unclear if this finding reflects genuine biological specificity, technical sensitivity, or normal inter-individual variation in gut microbiota, including the effect of diet. ClpB was previously detected in the plasma of all studied subjects at variable levels with no links to present results. This finding points to an increased α-MSH-like immunogenicity of the ClpB protein in some patients with AN. The origin of this phenomenon is presently unknown and can be related to the presence in AN microbiota of a particular Enterobacteriaceae species producing ClpB with enhanced α-MSH-like properties. Further studies in larger cohorts combining targeted proteomics with shotgun metagenomics should be able to clarify this question. Of interest, in a longitudinal study of gut microbiota composition in patients with AN, increased abundance of Enterobacteriaceae was associated with a worse disease progression19. A recent study in the same cohort of adolescents explored their plasma levels of ClpB, showing its high individual variability among both patients and controls without significant group differences20. Nevertheless, increasing dynamics of plasma ClpB were associated with disease relapse, while low level of anti-ClpB antibodies was detected in the acute stage of the disease, suggesting that a deficient immune control of ClpB may contribute to its increased anorexigenic effect20. Our study has several limitations. A relatively small number of patients with AN may underlie either over- or under-estimation of the occurrence of the α-MSH-like ClpB protein. Also, some weakly stained α-MSH-like spots were not analyzed. Finally, only functional validation of the newly identified α-MSH-like bacterial proteins, such as DnaK, may conclude on their putative α-MSH mimetic properties. It is, nevertheless, noteworthy that both DnaK and ClpB play a key role in the bacterial response to stress by forming a functional triad with small heat shock proteins, reversing protein aggregation21. Since α-MSH and other MC peptides are key messengers in stress-response22, a possibility of a coordinated response to stress between the microbiota and host is intriguing. In conclusion, a targeted proteomic approach to human gut microbiota was employed in this exploratory study, revealing the presence of several α-MSH-like bacterial proteins. Chaperon DnaK was identified as a putative α-MSH-like protein common to both healthy subjects and patients with AN. Nevertheless, the functional properties of DnaK as an α-MSH-like protein should be tested by its ability to activate MC receptors and/or to stimulate α-MSH-cross-reactive IgG, which in turn should be tested for their capacity to modulate α-MSH signaling as a peptide carrier protein. α-MSH-immunoreactive ClpB protein was detected in half of the group of patients with AN. Since Enterobacterial ClpB is known as an antigen-mimetic of α-MSH having its own anorexigenic properties, these results further support its relevance to the pathophysiology of AN, which should be further verified in larger independent cohorts. Finally, the targeted proteomics of human fecal microbiota appears as a new application of the proteomic analysis for the identification of new molecular targets in the microbiota-brain communication. Methods Study subjects and fecal sample collection The study included female adolescents with a restrictive form of AN (mean ± SEM: age 14.8 ± 0.8 years, body weight 45.5 ± 3.4 kg, body mass index (BMI) - standard deviation score (SDS) −2.33 ± 0.8 kg/m2, illness duration 14.8 ± 2.9 months, n = 6) diagnosed according to the DSM-523. Patients with AN were admitted to the Department for Child and Adolescent Psychiatry of the Rheinisch-Westfälische Technische Hochschule (RWTH) Aachen University Hospital. Female healthy-control (HC) adolescents with normal body weight [>20th and <80th age-adjusted percentile of BMI-SDS] were enrolled using newspaper advertisements (mean ± SEM: age 15.8 ± 0.3 years, body weight 26.8 ± 2.9 kg, BMI-SDS -0.4 ± 0.2 kg/m2, n = 6). The use of BMI-SDS is more sensitive for the evaluation of its significance as a biomarker as compared to BMI, since it reflects the difference from the average standard BMI known for each age and sex. This is particularly sensitive during the growth in children and adolescents. Written informed consent was obtained from the study participants and their legal guardians. The study, which was conducted under the Declaration of Helsinki, was approved by the ethics committee of the RWTH Aachen University Hospital (approval number EK 148-16). Fecal samples from patients with AN were taken at hospital admission and from HCs during the hospital visit as previously described, without previous fasting, in a standard procedure used for microbiome analysis19. For the proteomic analysis, fecal samples were put into a dry plastic container and stored at -80°C until the assay. Gut protein extraction A 250 mL culture containing Luria-Bertani broth (LB, Difco) was inoculated at an OD595 ± 0.01 (corresponding to 106 CFU/mL) from a 50 mL preculture made from 150 mg of stool previously ground with a potter. After 24 h at 37 °C at 140 revolutions per minute (rpm), bacteria were recovered by centrifugation (5000 × g, 30 min, 4 °C, 5430/5430 R centrifuge, Eppendorf). The pallet was washed three times with 10 ml of Dulbecco’s phosphate buffer saline (DPBS, Sigma-Aldrich). The pellet was freeze/thawed (−80 °C/37 °C) and resuspended in a denaturing buffer named R2D2 (7 M urea, 2 M thiourea, 2 mM tri-N-butylphosphine (TBP), 20 mM dithiothreitol (DTT), 0.5% (w/v) 3-(4-heptyl)phenyl-3- hydroxypropyl)dimethylammoniopropanesulfonate (C7BzO), 2% (w/v) 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate hydrate (CHAPS)). The bacterial suspension was sonicated twice on ice (3 s ON/1 s OFF, amplitude 24%, 6 min), then the lysate was centrifuged (12,000 × g, 30 min, 4 °C) and the supernatant was ultra-centrifuged (40,000 × g, 45 min, 4°C). Protein concentration was evaluated on the final supernatant using the Bradford assay (Bio-Rad). Samples were stored in aliquots at −20 °C until further use. Two-dimensional gel electrophoresis proteins separation For first dimension, proteins were migrated according to their isoelectric point using Ettan IPGphor3 isoelectric focussing system (GE Healthcare Life Sciences). A total of 50 µg was injected per immobilized pH gradient (IPG) strip (ReadyStrip IPG Strips 7 cm pH 3-10 nonlinear, Bio-Rad). Two strips were realized for each individual (control and AN patient). The rehydration solution was made of 2 µl bromophenol blue, 2.5 µl ampholytes (IPG buffer pH 3-10, GE Healthcare), and a sufficient volume of denaturing buffer to reach a 250 µl final volume. The strips were overlaid with mineral oil (Bio-Rad) and rehydrated overnight. Proteins were focused by an increasing electric current reaching a maximum of 7000 Vh (volt/hours). The focused strips were stored at −20 °C until the second dimension. Thawed strips were immersed for 15 min in reductive buffer (4 M DTT, 6 M urea, 1 M Tris-HCl [pH 6.8], 2% (w/v) SDS, 30% (v/v) glycerol in deionized water) followed by an alkylated buffer for 30 min in the dark (4 M iodoacetamide, 6 M urea, 1 M Tris-HCl [pH 6.8], 2% (w/v) SDS, 30% (v/v) glycerol in deionized water). The strips were loaded onto two 12.5% polyacrylamide gels (Acrylamide/Bis-Acrylamide 30% [29:1], Sigma-Aldrich). Electrophoresis was carried out in a Criterion Cell (Bio-Rad) using a constant current of 20 mA/gel. Six microliters of molecular weight marker (PageRuler Prestained protein ladder, Thermo Scientific, Waltham, MA, USA) were used. After migration, one gel was used for nitrate staining and protein identification by mass-spectrometry (MS), and the second gel was used for immunodetection by western blotting. Immunodetection by Western blotting For blotting the ready kit Trans-Blot Turbo RTA Transfert (Bio-Rad) was used. According to the manufacturer’s manual, the gel and the 0.22 µm nitrocellulose membrane were placed between sponge-pads soaked in transfer buffer (Bio-Rad) and blotted for 15 min with 2.5 A and 25 V using the Trans-Blot Turbo transfer system (Bio-Rad). The membrane was first blocked with 5% (w/v) milk (Skim milk powder, Millipore, Sigma-Aldrich) in Tris-buffered saline supplemented with Tween (TBS-Tween; 20 mM Tris-Base, 0.5 M NaCl, 0.05% Tween 20; pH 7.4) for 30 min at room temperature (RT) on a roller mixer. The membrane was next incubated overnight on a roller mixer at 4 °C with the anti α-MSH rabbit primary antibody (custom-produced by DelphiGenetics, Gosselies, Belgium, for TargEDys SA, France, which provided it for the Inserm UMR1239 laboratory) diluted at 1:1000 in 5% (w/v) milk in TBS-Tween. The membrane was washed five times with TBS-Tween (5 × 5 min with 5 mL at RT on a roller mixer), then incubated 1 h with the secondary antibody Goat anti-rabbit IgG/HRP (Invitrogen) diluted at 1:2500 in 5% (w/v) milk in TBS-Tween. Subsequently, five washes were realized with TBS-Tween (5 × 5 min with 5 mL at RT on a roller mixer). The immunodetection was performed with the Clarity Western ECL substrate kit (Bio-Rad) as described by the manufacturer’s manual and using the ChemiDoc XRS+ imaging system (Bio-Rad) coupled with the Image Lab software (version 6.0.1; Bio-Rad). Following immunodetection by chemiluminescence, the photograph of the membrane in the exact same position was also taken with the imaging system with the colorimetric option. Finally, immunodetection by chemiluminescence and colorimetric photographs was merged. Protein spot identification by mass spectrometry The second gel was used for nitrate staining, followed by spot excising and protein identification by MS. The following steps described above were realized on the HeRacLeS-PISSARO proteomics platform (Rouen, France, https://www.ibisa.net/plateformes/pissaro-4.html). After electrophoresis migration, proteins were fixed overnight (10% (v/v) acetic acid, 30% (v/v) ethanol, 60% (v/v) deionized water) followed by 3 washing steps (5 min with deionized water). After sensibilisation for 1 min (stock solution of 0.2% (w/v) of sodium thiosulfate diluted at 1:10 in deionized water), gel was incubated for 30 min in the dark with the nitrate staining solution (0.1% (w/v) of silver nitrate dissolved in deionized water). All incubation steps were carried out on a horizontal rotary shaker. The revealed protein spots chosen according the image analysis were excised and firstly immersed in reductive buffer for 15 min (5 mM DTT in deionized water) and then in an alkylated buffer for 30 min in the dark (20 mM iodoacetamide in deionized water). Gel bands were washed with 50% acetonitrile (ACN) (Fisher, Hampton, NH)/50% 10 mM ammonium bicarbonate (Sigma-Aldrich) and dried 3 times with 100% ACN. Proteins were digested with trypsin (0.01 µg of trypsin/spot, Promega, Madison, WI) dissolved in 10 mM ammonium bicarbonate, overnight at 37 °C. The peptides were extracted from the gel with several washing steps (deionized water 10 min, 100% ACN 10 min, 0.1% (v/v) of trifluoroacetic in 10 mM ammonium bicarbonate 10 min, 100% ACN 3 × 10 min). Finally, peptides were dried by vacuum centrifugation (SpeedVac concentrator, ThermoFisher Scientific, USA) at 30 °C and stored at −20 °C until MS analysis. For MS analysis, dried peptides were dissolved in 0.1% (v/v) of formic acid (FA) in deionized water and analysed on a Q-Exactive Plus OrbitrapTM mass spectrometer (Thermo Scientific, Waltham, MA, USA) coupled with an Easy-nLC 1200 (Thermo Scientific). One microliter of sample (0.2 µg) was injected and first eluted toward an enrichment column (C18 PepMap100, Thermo Scientific). The separation was then performed with an analytical column needle (NTCC-360/internal diameter: 100 µm; particle size: 5 µm; length: 153 mm, NikkyoTechnos, Tokyo, Japan). The mobile phases consisted of H2O/0.1% formic acid (FA) (buffer A) and CH3CN 80%/H20 20% with 0.1% FA (buffer B). The tryptic peptides were eluted at a flow rate of 300 nL/min using a three-step linear gradient: from 2 to 55% of buffer B over 106 min, from 55 to 100% of buffer B over 4 min, and 100% of buffer B for 15 min. The mass spectrometer was operated in the positive ionization mode with the capillary voltage and the source temperature set at 1.8 kV and 275 °C, respectively. The samples were analysed using the Higher-Energy Collisional Dissociation (HCD) method. The first scan (MS spectra) was recorded using the Orbitrap analyser (R = 70,000) in the mass range of m/z 400–1800. AGC target and maximum injection time were set to 3.106 and 60 ms, respectively. Then, the 10 most intense ions were selected for tandem mass spectrometry (MS2) experiments, excluding singly charged species. The dynamic exclusion of the already fragmented precursor ions was carried out for 20 s. Fragmentation occurred in the HCD cell analyzer at a normalized collision energy of 28. The MS2 spectra were also recorded using the Orbitrap analyser at a lower resolution (R = 17,500). After MS analysis, raw data were imported into Peaks® Studio X+ (Bioinformatics Solutions Inc.). The searches were performed with the following specific parameters: charge between 2 and 5, abundance threshold of 1.0E-7, enzyme specificity, trypsin; three missed cleavages permitted; variable modifications, carbamidomethylation (C), oxidation (M), pyro-glu from E and Q, and acetylation of the N-terminal; and MS/MS scan mode, Orbitrap. The mass tolerances for the precursor and fragment ions were set at 10 ppm and 0.5 Da, respectively. Spot identification was performed against all the bacterial taxa of the Uniprot/Swiss-Prot database. Only significant hits with a false discovery rate (FDR ≤ 1) for peptide and protein cutoff (−logP ≥ 20 and number of unique peptides ≥2) were considered. Supplementary information Supplementary information Supplementary data Acknowledgements We thank TargEDys SA, France, for providing α-MSH antibodies and the proteomic platform PISSARO, University of Rouen Normandie, France, for the mass spectrometry analysis. This study was supported by the ERANET Neuron MiGBAN consortium (01EW1906A-01EW1906B) “Microbiome Gut Brain Axis in Anorexia Nervosa,” funded by the European Union (EU), the German Ministry for Education and Research (BMBF), and the French Agence Nationale de la Recherche (ANR). S.O.F. research was also supported by the PTM2 program, Inserm, France. Author contributions L.B., L.W., and E.L. performed the fecal protein extraction and 2DGE, and immunodetection. L.B. and A.B.M. generated the mass spectrometry data. L.B., L.W., and S.O.F. analyzed the data. J.S. and B.H.D. performed clinical characterization of the cohorts and provided patient samples and clinical input. B.H.D., J.S., and S.O.F. obtained funding. S.O.F. conceptualized the study and wrote the manuscript. L.B. and L.W. revised sections of the manuscript. Data Availability The full raw dataset of proteomic data is provided in a supplementary file. Competing interests Authors declare no conflict of interest in relation to this study. SOF was a co-founder of TargEDys SA and is an inventor on the patent protecting the use of anorexigenic properties of ClpB (WO2015082633A1). The patent is licensed to BIOCODEX, France, which manufactures and commercializes ClpB-producing Hafnia alvei probiotic for the control of body weight. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s41522-025-00850-3. References Associated Data Supplementary Materials Supplementary information Supplementary data Data Availability Statement The full raw dataset of proteomic data is provided in a supplementary file.