FAM3D: A gut secreted protein and its potential in the regulation of glucose metabolism

The number of diabetic patients is rising globally and concomitantly so do the diabetes associated complications. The gut secretes a variety of proteins to control blood glucose levels and/or food intake. As the drug class of GLP-1 agonists is based on a gut secreted peptide and the positive metabolic effects of bariatric surgery are at least partially mediated by gut peptides, we were interested in other gut secreted proteins which have yet to be explored. In this respect we identified the gut secreted protein FAM3D by analyzing sequencing data from L- and epithelial cells of VSG and sham operated as well as chow and HFD fed mice. FAM3D was overexpressed in diet induced obese mice via an adeno-associated virus (AAV), which resulted in a significant improvement of fasting blood glucose levels, glucose tolerance and insulin sensitivity. The liver lipid deposition was reduced, and the steatosis morphology was improved. Hyperinsulinemic clamps indicated that FAM3D is a global insulin sensitizer and increases glucose uptake into various tissues. In conclusion, the current study demonstrated that FAM3D controls blood glucose levels by acting as an insulin sensitizing protein and improves hepatic lipid deposition.


Introduction
The global burden of diabetes is tremendous, currently being one of the leading causes of death in the world. According to the latest estimation of the International Diabetes Federation from 2021, approximately 537 million adults were living with diabetes and 541 million people were at increased risk of developing type 2 diabetes (T2D), the most common form [1]. Once insulin resistance has developed the associated aberrant high blood glucose levels can lead to severe health problems, including cardiovascular complications, stroke, vision loss, nerve damage or kidney failure [2]. The complications associated with T2D pose a significant burden to the patients, but also to the health care system. Thus, it is essential to lower and control blood glucose levels in diabetic patients before the concomitant health problems of diabetes manifest. Various treatment options are available for T2D but, as lifestyle interventions are often not sufficient, pharmaceutical therapies are commonly prescribed. Since the underlying cause of T2D is insulin resistance, restoring insulin sensitivity poses an effective strategy for the prevention or treatment of T2D. Today, Metformin, which is the most prescribed first in line therapy for T2D, is ameliorating insulin resistance. However, Metformin alone is often not sufficient for complete blood glucose level control, hence many patients receive a combinatorial therapy with a second or third anti-diabetes drug [3]. Besides Metformin, sulfonylureas, sodium-glucose co-transporter-2 (SGLT2) inhibitors, insulin and stable analogues of the gut secreted protein glucagon-like peptide 1 (GLP-1) or Dipeptidyl peptidase-4 (DPP4) inhibitors which prevent the rapid degradation of gut peptides are currently very successful diabetes medications [4,5]. Ultimately, a very effective treatment for T2D is bariatric surgery [6]. After bariatric surgery an increased secretion of gut hormones such as GLP-1, oxyntomodulin and peptide YY (PYY) can be observed, which may mediate at least a part of the positive metabolic effects [6,7]. Thus, other gut secreted peptides might be interesting pharmaceutical targets and provide new alternative therapeutic options for the treatment of T2D.
The FAM3 family is a cytokine like family containing four members: FAM3A, FAM3B (PANDER), FAM3C and FAM3D. The first three members of the family are fairly well characterized and described to be involved in the pathophysiology of diabetes and non-alcoholic fatty liver disease (NAFLD) [8][9][10][11]. In contrast, the physiological role of the fourth member, FAM3D, remains poorly understood. It is primarily expressed in the digestive system [12], and the intestinal expression of FAM3D was shown to be induced by ingestion of fats and reduced by fasting in mice [13]. In humans, FAM3D plasma levels were reported to be increased following a high fat meal [13]. In addition FAM3D was discussed as chemotactic agonist for the G-protein coupled receptors, formyl peptide receptors 1 & 2 (FRP1 & FRP2), which are primarily expressed on neutrophils and monocytes, suggesting pro-inflammatory properties of FAM3D [14]. In contrast, data obtained in the FAM3D knock out mouse model suggested that this protein has an essential role in the maintenance of colon homeostasis, protects against inflammation associated cancer and supports a normal microbiota composition [12].
In the presented study we identified FAM3D as a gut secreted protein, of which circulating levels were increased in patients after bariatric surgery. By using an adeno associated virus (AAV) approach to overexpress FAM3D in diet induced obese mice, we investigated the role of FAM3D in glucose metabolism. FAM3D not only displayed insulin sensitizing potential, but also ameliorated symptoms of NAFLD. As our data rather excluded signaling via FRP1/2 the so far only known receptors of FAM3D we performed a kinase activity screen in adipose tissue to reveal novel signaling routes of FAM3D. This screen indicated receptor tyrosine kinase signaling as potential pathway mediating the insulin-sensitizing effects of FAM3D.

Materials
All chemicals were purchased from Sigma Aldrich unless otherwise specified.

Human FAM3D plasma levels
FAM3D levels were measured in serum samples of patients who underwent open abdominal surgery for Roux-en-Y bypass or sleeve gastrectomy. The age range of the patients was 21-61 years and 19 females and 10 males were analyzed Pre-surgery blood samples were taken 14 days before Roux-en-Y gastric bypass surgery to avoid any confounding effects of pre-operative diets. All investigations have been approved by the ethics committee of the University of Leipzig (363-10-13122010 and 017-12-230112) and were carried out in accordance with the Declaration of Helsinki. Study participants provided witnessed written informed consent before entering the study. FAM3D levels were analyzed by the Ab-Match ASSEMBLY Human FAM3D kit (Labforce). The ELISA kit was validated by quantification of absorbance in conditioned media from HEK293-A cells overexpressing either human FAM3D-GFP or GFP as control. A distinct FAM3D specific signal was detected in the ELISA in conditioned media taken from FAM3D overexpressing cells but not in conditioned media taken from GFP overexpressing cells. A western blot done in parallel with the same samples showed a FAM3D specific band in cell lysate as well as supernatant from FAM3D transfected cells but no bands in the respective samples from GFP transfected cells (data not shown). The human serum samples were diluted 1:5 for measurements. The assay was performed according to the manufacturer's instructions.

In vivo experimental procedures
All animal experiments were approved by the Veterinary office of the Canton Zürich. Animals were housed in individually ventilated cages at standard housing conditions (22 • C, 12 h reversed light/dark cycle, dark phase starting at 7 am). The animals had access to food and water ad libitum. Chow diet contained 18% proteins, 4.5% fibers, 4.5% fat, 6.3% ashes (Provimi Kliba SA).

Dietary intervention.
A high-fat diet (23.9% proteins, 4.9% fibers, 35% fat, 5.0% ashes, Provimi Kliba SA) in which approximately 60% of the calories are derived from fat was used to induce obesity and insulin resistance in mice. The feeding regimen was initiated when the animals were eight weeks old and maintained until all measurements had been finished.

Global inducible FAM3D knockout mouse.
The FAM3D fl/fl mice were generated by the ETH Phenomics Center Zurich (epic.ethz.ch). LoxP sites flanking exon six of FAM3D were introduced using the newly developed technique of EASI-CRISPR. Here, specific CRISPR/Cas9 complexes for the insertion sites of the loxP sites were injected directly into mouse oocytes together with a single-stranded template for homologous recombination that contains 150 bases long homology arms and the two loxP-sites flanking exon 6. Following this procedure, the embryos were transferred into foster mothers. Offspring with successful insertion of the loxP sites were once backcrossed to C57BL6/N mice. Subsequently the animals were mated to Rosa26-creERT2 animals (Taconic). At eight weeks of age recombination of the floxed allele was induced by oral tamoxifen gavage (2 mg/mouse in sunflower oil) and the animals were changed to HFD. To sustain the knock-out tamoxifen was reapplied every four weeks. Following eight weeks of HFD the metabolic phenotyping was started. The HFD was maintained until all measurements had been performed. Male and female mice were carefully phenotyped.

Injection of AAV.
Male C57BL/6N animals received HFD for 10 weeks before AAV injection. After the injection the HFD regimen was continued until the end of all measurements. Animals were assigned to the experimental groups based on bodyweight and fasting blood glucose levels. 1 × 10 11 Virus genomes per mouse were diluted in PBS (Gibco) to a volume of 50 µl and injected intravenously at the tail vein. Metabolic phenotyping was started two weeks post virus injection.

Fasting blood glucose.
For measurement of fasting blood glucose food was withdrawn for six hours during the dark phase. Blood was collected from a little incision at the tail vein. Blood glucose levels were determined using a standard glucometer (ACCU-CHEK Aviva, Roche).

Intraperitoneal glucose tolerance test (ipGTT).
Animals were weighted and after a 6 h fasting period during the dark phase fasting blood glucose levels were determined. D-glucose in saline (Braun) was injected intraperitoneally at a dose of 1 g/kg body weight. Blood glucose levels were measured 15 min, 30 min, 60 min, 90 min and 120 min post glucose injection using a standard glucometer (ACCU-CHEK Aviva, Roche).

Oral glucose tolerance test (oral GTT).
For the oral GTT the same procedure as for the ipGTT was applied. D-glucose in saline (Braun) was orally gavaged at a dose of 1 g/kg body weight.

Insulin tolerance test (ITT).
For the ITT the same procedure as for the ipGTT was applied. Insulin (Actrapid, Novo Nordisk) was injected intraperitoneally (ip.) at a dose of 0.75 U/kg bodyweight.

Pyruvate tolerance test.
For the Pyruvate tolerance test the same procedure as for the ipGTT and the ITT was applied. Pyruvate was dissolved in saline (Braun) and injected ip. at a dose of 2.0 g/kg.

Blood collection for the determination of plasma parameters.
Blood was collected submandibular in EDTA coated tubes from either random fed or 6 h fasted animals. Immediately after blood collection 3 µl of Aprotinin (25000 KIU/ml) (Roth) were added. To obtain plasma whole blood was centrifuged at 2000 g for 20 min at 4 • C.

2.2.2.10
. Urine collection. The animals were placed over a collection tube and urination was facilitated by light strokes over the abdomen of the animals. Urine was collected either in random fed or 6 h fasted animals.

2.2.2.11
. Body composition. The body composition was determined by magnetic resonance imaging (EchoMRI 130, Echo Medical Systems). Fat and lean mass were analyzed using the EchoMRI 14 software.

Indirect calorimetry.
Indirect calorimetry measurements were performed in the Phenomaster (TSE Systems) according to the manufacturer's guidelines. O 2 and CO 2 levels were measured continuously every 13 min for 60 s. Energy expenditure was determined according to the manufacturer's guidelines. The respiratory quotient was calculated by determining the ratio of CO 2 production over O 2 consumption. Activity as well as food and water intake were monitored continuously.

Hyperinsulinemic euglycemic clamps.
After eight weeks on HFD regimen the animals were injected with AAV. Two weeks later the animals underwent surgery, and a catheter was implanted into the vena jugularis. After 5-7 days of recovery, hyperinsulinemic euglycemic clamps were performed only on animals which regained more than 90% of their preoperative weight. For the clamps the animal were fasted for 6 h, in the last 80 min of the fasting period Glucose-D-[3-3 H] (0.05 µCi/ min) (Perkin Elmer) was infused to determine the basal glucose uptake and glucose production. Blood was sampled at the end of the 80 min basal Glucose-D- [3-3 H] infusion. Subsequently Glucose-D-[3-3 H] (0.1 µCi/ml), Insulin (12 mU/kg/min) (Actrapid, Novo Nordisk) and 20% (w/v) glucose (Braun) as needed were infused until the animal reached a steady state at 6 mmol/l blood glucose. Upon reaching the steady state a bolus of Deoxy-D-glucose 2-[ 14 C(U)] (10 µCi) (Perkin Elmer) was administered to determine the glucose uptake into the individual organs. Blood was sampled when the steady state was reached as well as 2 min, 15 min, 25 min, 35 min after the Deoxy-D-glucose 2-[ 14 C] bolus. Subsequently, the animals were euthanized, and tissues were harvested. For the analysis plasma samples were deproteinized with Ba(OH) 2 and ZnSO 4 as described in Kim et al., 2009 [15]. Tissues were homogenized, boiled and the phosphorylated Deoxy-D-glucose 2-[ 14 C] was separated from the non-phosphorylated Deoxy-D-glucose 2-[ 14 C] by ion exchange columns (chloride form) (BioRad). The glucose turnover is calculated by dividing the rate of Glucose-D-[3-3 H] infusion with the Glucose-D-[3-3 H] specific activity in the plasma [15]. The endogenous glucose production was determined by subtraction of the glucose infusion rate from the glucose disposal rate [15]. The insulin stimulated glucose disposal rate is determined by subtracting the basal glucose disposal rate from the glucose disposal rate during the clamp [15]. The organ specific glucose uptake was calculated by determining the organ specific accumulation of phosphorylated Deoxy-D-glucose 2-[ 14 C] and the decay of Deoxy-D-glucose 2-[ 14 C] in the plasma. Two rounds of hyperinsulinemic clamps were performed and if possible, the data were pooled.

Recombinant protein injection.
Recombinant protein was diluted in saline (Braun) and injected intravenously into the tail vein at a dose of 0.4 mg/kg bodyweight. Thereafter the animals were fasted and 4,5 h later they either underwent a glucose tolerance test or they were thoroughly perfused with ice cold PBS (Gibco) and tissues were harvested for further analysis.

Tissue harvest.
For tissue harvest the animals were euthanized by carbon dioxide asphyxiation. The tissues were carefully dissected and snap frozen in liquid nitrogen. The popliteal lymph node was removed from inguinal white adipose tissue (ingWAT).

Cell suspension preparation.
For the collection of peritoneal immune cells, the animals were euthanized, and the abdomen was washed twice with PBS (Gibco) containing 2 mM EDTA through a syringe inserted into the peritoneal space. Cells were collected by centrifugation. The livers were minced and digested in IMDM (Gibco) with 3% FBS (Gibco), 2 mg/ml of type IV collagenase and 0.125 mg/ml Dnase I for 30 min and passed through a 40-μm-cell strainer (Becton Dickinson). The liver cells were first centrifuged for 5 min at 20 g and then separated by density centrifugation with 30% Percoll at 2000 rpm, 20 min, low acceleration and no brake. The pellet in the bottom was collected for further analysis. Erythrocytes were lysed with ACK buffer.

Analysis of metabolic parameter in plasma and urine
Triglycerides in plasma were determined with the COBAS TRIGB kit (Roche/Hitachi). Free fatty acids were determined via a NEFA-HR(2) assay (Wako Chemicals). Glucose in urine was determined using a colorimetric glucose-assay-kit. Insulin, Glucagon, Leptin and Adiponectin levels were all measured by mouse ELISA kits (Crystal Chem). VEGF A levels were determined with the Mouse VEGF Quantikine ELISA Kit (R&D). ALT activity was determined by a kinetic colorimetric assay and T4 was determined with a Thyroxine (T 4 ) competitive ELISA kit (Invitrogen). All assays were performed according to the manufacturer's instructions.

Liver histology, liver triglycerides and glycogen
One lobe of the liver was fixed for 24 h in 4% PFA and then processed by STP120 Spin Tissue Processor (ThermoScientific) according to the manufacturer's protocol. Subsequently, the samples were paraffin embedded and cut to 5 µm sections. Staining with hematoxylin eosin was performed. The tissue sections were imaged by light microscopy using an AxioPhot microscope equipped with an AxioCam MR (Zeiss). Liver steatosis was analyzed as described in [17], for the analysis of liver micro steatosis the pipeline presented in [17] was modified and trained for micro steatosis recognition. The other lobes of the liver were snap frozen in liquid nitrogen for further analysis. Liver triglycerides were extracted using a chloroform: methanol (2:1) mixture and normalized to tissue weight. Liver glycogen was extracted by homogenization of the tissue in ice cold water. For enzyme inactivation the samples were boiled for 10 min and cleared by centrifugation. The supernatant was analyzed with the colorimetric based Glycogen Assay kit (Biovision) according to the manufacturer's instructions.

FAM3D tissue accumulation
To determine the accumulation of FAM3D in tissues, tissues were lysed in RIPA buffer (Cell Signaling) supplemented with protease inhibitor cocktail (Complete, Roche). Protein concentration was determined by Pierce™ BCA Protein Assay Kit (ThermoFisher) and the lysates of the different organs were all diluted to 0.04 mg/ml total protein. The presence of FAM3D in the tissues was analyzed using the high throughput Western blot system Sally Sue (ProteinSimple). Mouse FAM3D primary antibody (RD systems) at a dilution of 1:50 was used.

Kinase activity screen
To determine the kinase activity profile in white adipose tissue, the frozen tissues were grinded with a handheld mortar and lysed in MPER buffer (ThermoFisher) supplemented with Halt protease and phosphatase inhibitor cocktail (ThermoFisher) for 20 min at 4 • C on a rotor. Then lysates were cleared by two to three centrifugation steps (12,000 rpm, 4 • C) and aliquots were prepared and snap frozen for further analysis. For the whole process all tubes were always pre-chilled on ice. The protein concentration was determined with the Pierce™ BCA Protein Assay Kit (ThermoFisher).
The kinase activity profile in the lysates was analyzed with the PAMGene technology. Therefore, 1 µg of protein was loaded onto phosphotyrosine kinase (PTK) and phosphoserine/threonine kinase (STK) chip arrays. Each array contained peptides (196 PTK and 144 STK) with tyrosine or serine/threonine phosphorylation sites. Upon phosphorylation FITC-labelled anti-phosphotyrosine or anti-phosphoserine/ threonine antibodies were used to monitor peptide phosphorylation. The kinetics of phosphorylation were recorded by a charge coupled device (CCD) camera in combination with the Evolve software v1.2 (PamGene). The PAMGene experiments were performed by the Functional Genomics Center Zürich. For the analysis AAV-FAM3D animals were compared with AAV-Stuffer animals. The Bionavigator software v. 6.2 (PamGene) was used for data analysis und upstream kinase activity predictions. The computational analysis was performed by PamGene.

Protein extraction and Western blot
Tissues were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 2 mM EDTA, 0.5% sodium deoxycholate, 1.0% Triton X100, 10% glycerol) supplemented with protease inhibitor cocktail (Complete, Roche) and Halt phosphatase inhibitor cocktail (ThermoFisher). Tissues were homogenized in RIPA buffer by metal beads in the TissueLyser LT (Qiagen). The lysates were centrifuged for 5-10 min at 12,000 rpm and the cleared supernatant was collected. Protein concentration was determined with the Pierce™ BCA Protein Assay Kit (Thermo Fisher) and equal amounts of protein (10-30 µg) were loaded onto SDS-Polyacrylamide gels for size separation. The proteins were transferred to nitrocellulose membranes, which were blocked for 1 h at RT in 5% BSA or 5% milk prior to incubation with the primary antibody at 4 • C overnight. Unless otherwise stated antibodies were purchased from Cell Signaling and diluted 1:1000: AKT (9272 S), pAKT Ser437 (4060 S), ERK 1/2 (4695 S), JAK1 (3344), HSP90 (4877 S), FAM3D (AF3027, RD systems). Next the membranes were incubated with a secondary HRPconjugated antibodies from rabbit (1:10 000, Merck) or goat (1:10 000, Sigma) for 60 min at room temperature. Following a short incubation with Luminol, luminescence was determined with the Image Quant system (GE Healthcare Life Sciences).

Virus production
Murine FAM3D/Oit1 (UniProt P97805) was codon optimized for the expression in mice and cloned into a pFB vector under a LP1 promotor58 to allow liver-specific protein expression. A pFB-Stuffer vector59 was used as control. AAV8 was produced by transfection of human embryonic kidney (HEK)− 293 T cells with the expression plamids (pFB_LP1-mFAM3D or pFB_Stuffer) plus the plasmids pDP8 (Plasmid Factory) and the pHelper plasmid (ThermoFisher). The purification and titration was performed as described previously [18].

Recombinant mouse FAM3D-Fc production
Recombinant FAM3D was produced in HEK 293-E6 suspension cells cultured in F17 + 0,1% Pluronic F68 + L-Glutamin/Glutamax + 25 µg/ ml G418 (all ThermoFisher). For transfection, plasmids harboring mFAM3D-Fc under the CMV promotor were diluted in Opti MEM I mixed 1:1 with in Opti MEM I diluted 293fectin (both ThermoFisher). 24 h after transfection Tryptone N1 at an end concentration of 0,5% was added. At day five post transfection the cell suspension was cleared by centrifugation. The supernatant was mixed with Mab-Select beads (GE Healthcare) and incubated overnight. With the help of Econo-Pac® Chromatography Columns (Biorad), the beads were collected and washed prior to elution of the purified protein.

Quantification and Statistical analysis
For in vivo studies, littermates were used for all experiments. Unless otherwise indicated, all results are expressed as mean ± standard error of the mean (SEM). When two groups were compared a two-tailed, unpaired student's t-test (significance cut off: p < 0.05) was used to assess statistical significance. If more than two groups were compared a twoway ANOVA with Turkey multiple comparison testing was applied to test for statistical significance. Paired Student's t test was used to analyze the differences in paired samples. Possible outliers were removed if they differed more than the standard deviation multiplied times two from the mean. All graphs and statistical analyses were performed using GraphPad Prism (Version 8). Statistical significance is indicated as following: * = p ≤ 0.05; * * = p ≤ 0.01; * ** = p ≤ 0.001.

FAM3D is a gut-secreted protein associated with the beneficial metabolic effects of bariatric surgery
Here we aimed to identify additional secreted gut peptides besides the usual suspects which have positive effects on metabolic parameters such as body weight, blood glucose levels, triglycerides or insulin sensitivity. In our previous study, we characterized the transcriptome of mouse ileal and colonic L-and epithelial cells following high fat diet (HFD) and bariatric surgery [19]. An ileum vs. colon comparison in L-cells was made separately for each metabolic condition (HFD, chow, VSG and sham) as well as in epithelial cells for VSG and sham operated mice. FAM3D was colonically enriched in all comparisons. After additional parameters such as intestinal tissue-specificity, high expression level, secretion, expression in cell lines and available literature were considered, FAM3D was chosen for further analysis because a) it performed particularly well in all selection criteria mentioned above and b) other members of the FAM3 family had been described to play a role in diabetes and NAFLD.
The expression of FAM3D was significantly higher in the colon compared to the small intestine (SI) in both, L-cells and enterocytes. Given the expression profile and the size of 25 kD, FAM3D compared to GLP-1 is not a classical gut peptide per se, as it is much bigger and expressed not only in enteroendocrine L-cells but also in the surrounding enterocytes [ Fig. 1A]. Based on the hypothesis, that gut peptides play a role in the remission of insulin resistance and normalization of blood (caption on next page) C. Moser et al. glucose levels as well as weight loss following bariatric surgery, we measured FAM3D serum levels in human patients before and after bariatric surgery. We detected a significant increase in FAM3D 12 months post-surgery compared to pre-surgery baseline levels [ Fig. 1B]. This led us to hypothesize that FAM3D is a gut secreted protein, which might be involved in the regulation of energy metabolism and glucose lowering effects of bariatric surgery.

The overexpression of FAM3D evoked a positive glucometabolic phenotype
Encouraged by this finding, we decided to assess the metabolic effects of FAM3D in more detail in an obese and insulin resistant C57BL/ 6N mouse model. An adeno-associated virus (AAV) expressing FAM3D or Stuffer as control was injected intravenously into mice that were fed a diet in which approximately 60% of calories are derived from fat (HFD) for 10 weeks. The HFD regimen was continued after AAV application. This led to a distinct and long-lasting (up to 16 weeks) hepatic overexpression resulting in increased FAM3D levels in plasma [Suppl. Fig. 1A; B]. The overexpression of FAM3D caused a significant improvement of the glucometabolic phenotype in HFD fed mice. Fasting and random fed blood glucose levels [ Fig. 1C; F], as well as glucose and insulin tolerance [ Fig. 1D; E] were significantly improved upon FAM3D overexpression. The analysis of plasma from fasted animals demonstrated a significant reduction of triglycerides [ Fig. 1G], while free fatty acids, insulin and glucagon levels were unchanged [ Fig. 1H -J]. Other hormones which are known to improve blood glucose levels, such as Leptin, Adiponectin and Thyroxine (T4) were unchanged in both the fasted and the fed state [Suppl. Fig. 1C -E]. Random fed vascular endothelial growth factor A (VEGF-A) and free fatty acids did also not differ between the groups [Suppl. Fig. 1F; G]. Additionally, no increase in glucose excretion via the urine was observed [Suppl. Fig. 1H]. Furthermore, FAM3D overexpression reduced fasting blood glucose levels [Suppl. Fig. 1J] and improved glucose tolerance [Suppl. Fig. 1I] also in lean, chow fed mice. All these effects were independent of bodyweight [ Fig. 1K], as well as lean or fat mass [ Fig. 1L]. In line with the body weight data, FAM3D overexpression did not affect energy expenditure (EE), the respiratory quotient (RER), food/water intake and activity [Suppl. Fig. 1K -O]. Since FAM3D was reported to be a chemotactic agonist for neutrophils and monocytes [14], we investigated liver and peritoneum 2 weeks after initiation of FAM3D overexpression for the presence of inflammatory markers. No increased infiltration of macrophages or neutrophils was observed upon overexpression of FAM3D [Suppl. Fig. 2A; B]. Further, the alanine transferase (ALT) activity in plasma did not differ between the two groups. Hence, there was no indication that the AAV mediated overexpression was affecting liver health in a negative manner [Suppl. Fig. 2C]. In addition to the FAM3D overexpression model, an inducible global FAM3D knockout mouse (Rosa26-creERT2 x FAM3D fl/fl) [Suppl. Fig. 2D] was generated. Following tamoxifen induced knockout of FAM3D surprisingly no metabolic alterations were detected under obesogenic conditions. Neither fasting blood glucose levels, nor glucose tolerance (intraperitoneal (ip.) and oral) or body weight was changed upon FAM3D knockout in HFD fed mice [Suppl. Fig. 2E -H].
In summary, increased secretion of FAM3D upon overexpression led to improved glucose metabolism in obese and insulin resistant mice by improving insulin sensitivity without eliciting pro-inflammatory effects.

FAM3D ameliorated hepatic lipid deposition and steatosis
Since the FAM3 family is discussed to play a role in the pathophysiology of NAFLD [8], we next studied the effect of FAM3D on liver steatosis in mice. The animals were on HFD 10 weeks prior to AVV injection and continued receiving HFD afterwards. We observed a significant reduction in the deposition of liver triglycerides following AAV-mediated overexpression of FAM3D [ Fig. 2A], a finding which is corroborated by histology [ Fig. 2D; E]. A trend towards a reduction in liver steatosis area (p = 0.0597) and a significant reduction in liver micro steatosis area were observed upon FAM3D overexpression in mice on HFD [ Fig. 2B; C]. Similar to the ipGTT and the ITT [ Fig. 1E; F], FAM3D overexpressing animals performed significantly better in a sodium pyruvate tolerance test [ Fig. 2F], suggesting reduced gluconeogenesis in liver, possibly due to improved insulin sensitivity. Liver glycogen levels, which are primarily regulated by insulin, were unchanged [ Fig. 2G]. In contrast to FAM3A and FAM3C, which are both activating the AKT signaling pathway in liver when chronically overexpressed in HFD fed mice, FAM3D did not increase AKT and pAKT Ser437 levels in liver [ Fig. 2H -J] [9,10].
In conclusion, FAM3D overexpression led to reduced hepatic lipid deposition and steatosis but unlike the other family members FAM3A and FAM3C, FAM3D did not affect the AKT signaling cascade in the liver.

FAM3D acted as systemic insulin sensitizer
Hyperinsulinemic euglycemic clamps are widely considered as the gold standard to assess the action of insulin, in vivo [15,20]. Therefore, we performed hyperinsulinemic euglycemic clamps [ Fig. 3A] to further investigate the improvement in systemic insulin sensitivity and to determine which organs contribute to the observed improved glucose uptake. Mice were put on HFD for eight weeks prior to injection of the AAV. HFD regimen was continued afterwards. Two weeks after AAV injection the animals underwent surgery and three weeks after virus administration the hyperinsulinemic clamps were performed. In agreement with the metabolic data presented in Fig. 1, FAM3D overexpressing mice required a significantly higher glucose infusion rate (GIR) to maintain euglycemia [ Fig. 3B]. Glucose turnover at basal conditions was unchanged, but it was significantly increased in AAV-FAM3D HFD animals under insulin stimulated conditions [ Fig. 3C]. The endogenous glucose production (EGP) did not differ between the groups at basal and at insulin stimulated conditions [ Fig. 3D]. When determining glucose uptake into individual organs, a significant increase in glucose uptake was observed in interscapular brown adipose tissue (iBAT), inguinal white adipose tissue (ingWAT), liver, heart, soleus and tibialis anterior (TA) muscle [ Fig. 3E -J]. Only in the two Fig. 1. : FAM3D in the context of bariatric surgery in humans and its positive metabolic phenotype in mice upon overexpression (A) Transcriptomics of Lcells and enterocytes from colon and small intestine (SI). Fragments per kilobase of transcript per million mapped reads (fpkm) are presented (n = 7 per group). (B) FAM3D levels were measured in serum of patients who underwent open abdominal surgery for Roux-en-Y bypass or sleeve gastrectomy 14 days pre-and 12 months post the intervention (n = 30 per group). Mice received HFD for 10 weeks before they were injected with an adeno-associated virus (AAV) expressing FAM3D or Stuffer as control. After AAV injection the dietary intervention with HFD was continued. (C) Fasting blood glucose was measured two weeks post virus injection (n = 8 (Stuffer) and n = 9 (FAM3D)). (D) Intraperitoneal glucose tolerance test (ipGTT) performed three weeks post virus injection (n = 8 (Stuffer) and n = 9 (FAM3D)). 6 h fasting blood glucose levels were determined prior to intraperitoneal injection of 1 g/kg bodyweight D-glucose in saline. Blood glucose was measured 15 min, 30 min, 60 min, 90 min and 120 min post glucose injection. (E) Insulin tolerance test (ITT) performed five weeks post virus injection in analogy to the ipGTT except that 0.75 U/kg bodyweight insulin instead of glucose was injected (n = 8 (Stuffer) and n = 9 (FAM3D)). (F) Random fed blood glucose measured three weeks post virus injection (n = 8 (Stuffer) and n = 9 (FAM3D)). (G) Triglycerides (n = 10 per group), (H) free fatty acids (n = 10 per group), (I) insulin (n = 12 per group) and (J) glucagon levels (n = 10 per group); all measured in plasma samples from 6 h fasted animals two weeks post virus injection. (K) Bodyweight was monitored weekly in random fed animals following virus injection (n = 7 per group). (L) Lean and fat mass were determined by EchoMRI 16 weeks post virus injection (n = 7 per group). All data presented are mean ± SEM, Statistical analysis was performed by ordinary one-way ANOVA and multiple comparison testing (A), by paired Student's t-test (B) and by unpaired Students t test (C-L), Significance is indicated as * p < 0.05, * * p < 0.01 and * ** p < 0.001. skeletal muscles, gastrocnemius, and extensor digitorum longus (EDL) glucose uptake was unchanged [ Fig. 3K; L].
In summary, the increased glucose uptake into various tissues under hyperinsulinemic euglycemic clamp conditions confirmed that FAM3D acts as systemic insulin sensitizer and indicated that the positive metabolic phenotype of FAM3D is not mediated by a single organ.

Effects of FAM3D were mediated by receptor tyrosine kinase signaling
Due to the global insulin sensitizing effect of FAM3D, we moved on to determine its main target tissue(s). To do so, we quantified the amount of recombinant Fc-tagged FAM3D protein taken up into various tissues after a single intravenous (iv.) injection of recombinant FAM3D in vivo. Injecting recombinant FAM3D into animals receiving HFD for 10 weeks prior to injection to induced obesity and insulin resistance resulted in a significant improvement in glucose tolerance four hours post FAM3D administration suggesting that the recombinant protein is functional [ Fig. 4A; B]. Thus, to study FAM3D tissue accumulation chow fed lean animals were injected iv. with recombinant FAM3D four hours prior to perfusion and harvest of various organs. Whole organs were then lysed and FAM3D tissue accumulation determined by high throughput Western blotting. The amount of FAM3D was particularly high in inguinal and epididymal white adipose tissue [ Fig. 4C]. As no lipophilic structures were predicted for FAM3D, it is very likely that the increase of FAM3D in white adipose tissue was driven by protein function [21]. Surprisingly, the amount of FAM3D in BAT and the two skeletal muscles TA and Soleus was comparably low, even though these tissues displayed an increased glucose uptake in the clamps.
As our previous data indicated that the insulin-sensitizing effects of FAM3D were not mediated by its only so far known receptors FRP1/2 [Suppl. Fig. 2A-C] we moved on to elucidate FAM3D's molecular mode of action by performing a kinase activity profiling in its main target tissue WAT. Two weeks after initiation of AAV mediated overexpression of FAM3D in HFD fed animals whole inguinal adipose tissue was isolated and lysed from FAM3D overexpressing and control animals in the fasted state. The kinase activity profile was then analyzed with a PAMStation by dispensing the whole tissue lysate on microarray chips harboring various peptide fragments with kinase specific phosphorylation sites for either tyrosine (PTK) or serine and threonine residues (STK). Based on the peptide phosphorylation pattern kinase activity was computationally predicted. An overall reduction in peptide phosphorylation was observed upon long-term AAV mediated overexpression of FAM3D [Suppl Fig. 3A; B]. Kinase activity predicted by the PTK phosphorylation . Quantification of (I) AKT and (J) pAKT Ser437 levels normalized to HSP90 of the Western blot displayed in (H). All data presented are mean ± SEM, possible outliers were removed if they differed more than the standard deviation multiplied times two from the mean. Statistical analysis was performed by unpaired Students t test. Significance is indicated as * p < 0.05, * * p < 0.01 and * ** p < 0.001.
(caption on next page) C. Moser et al. Fig. 3. : FAM3D acted as a systemic insulin sensitizer Mice received HFD for 8 weeks before they were injected with an AAV expressing FAM3D or Stuffer as control. After AAV injection the dietary intervention with HFD was continued. Two weeks after AAV injection the animals underwent surgery and three weeks after AAV injection the hyperinsulinemic euglycemic clamps were performed. (A) Schematic description of the experimental set-up for the hyperinsulinemic euglycemic clamps. (B) Glucose infusion rate (GIR) at the steady state at a level of 6 mmol/l blood glucose, data of the two clamp studies were pooled (n = 12 (Stuffer) and n = 16 (FAM3D)). (C) Glucose turnover at basal and insulin stimulated levels (n = 8 per group). (D) Endogenous glucose production (EGP) at basal and insulin stimulated levels (n = 8 per group). Tissue specific glucose uptake in (E) Brown adipose tissue (BAT), data of the two clamp studies were pooled (n = 12 per group), (F) inguinal white adipose tissue (ingWAT), data of the two clamp studies were pooled (n = 11 (Stuffer) and n = 14 (FAM3D)), (G) Liver, data of the two clamp studies were pooled (n = 8 (Stuffer) and n = 9 (FAM3D)), (H) Heart (n = 7 (Stuffer) and n = 8 (FAM3D)), (I) Soleus muscle, pooled data of the two clamp studies (n = 11 (Stuffer) and n = 14 (FAM3D)), (J) Tibialis anterior muscle (n = 8 (Stuffer) and n = 7 (FAM3D)), (K) Extensor digitorum longus muscle (EDL) (n = 8 (Stuffer) and n = 7 (FAM3D)), (L) Gastrocnemius muscle (n = 8 (Stuffer) and n = 7 (FAM3D)). Two rounds of hyperinsulinemic clamps were performed and when possible, the data were pooled. All data presented are mean ± SEM, possible outliers were removed if they differed more than the standard deviation multiplied times two from the mean. Statistical analysis was performed by unpaired Students t test. Significance is indicated as * p < 0.05, * * p < 0.01 and * ** p < 0.001. For the kinase activity screen mice were fed HFD for 10 weeks before they were injected with an AAV expressing FAM3D or Stuffer as control. After AAV injection the dietary intervention with HFD was continued. Two weeks post virus injection the animals were sacrificed after a 6 h fasting period and inguinal white adipose tissue was harvested. (D) List of the top ranked regulated Kinases predicted by the phosphorylation pattern on the chip containing peptides phosphorylated by phosphotyrosine kinases (PTK) or (E) serine-threonine kinases (STK) respectively, scored by the mean kinase score and the mean specificity score (n = 10 per group). (F) Western blots of ingWAT lysates from the kinase activity screen for one of the top PTK (JAK1) and one of the top STK (ERK1/2) candidates (n = 6 (Stuffer) and n = 7 (FAM3D)). Quantification of (G) ERK1/2 and (H) JAK1 levels normalized to HSP90 of the Western blots displayed in (F). All data presented are mean ± SEM. Statistical analysis was performed by unpaired Students t test (A; B; G; H). Significance is indicated as * p < 0.05, * * p < 0.01 and * ** p < 0.001. pattern indicated that increased FAM3D levels resulted in altered signaling of receptor tyrosine kinases [ Fig. 4D] [22]. The STK phosphorylation pattern induced by FAM3D overexpression pointed towards an alteration in cell cycle regulation [ Fig. 4E] [23,24]. The data of the kinase activity screen were validated for ERK 1/2 [ Fig. 4F; G] and Janus kinase 1 (JAK1) (p = 0.0959) [ Fig. 4F; H], targets which were among the top five regulated kinases predicted by the PTK or STK chip respectively.
In summary, we identified white adipose tissue as the main target tissue of FAM3D and changes in receptor tyrosine kinase signaling and cell cycle regulation as potential pathways mediating the effects of FAM3D.

Discussion
The potential of gut secreted peptides for the treatment of obesity and diabetes is well known. GLP-1 synthetic peptide agonists with a prolonged half live compared to the naïve protein or DPP4-inhibitors which stop the rapid degradation of peptides aim to increase gut peptide levels in the blood thereby prolonging their positive metabolic actions [5]. Here, we showed that the gut secreted protein FAM3D has potential for the development as a new therapeutic treatment option for insulin resistance.
While FAM3A and FAM3C levels were decreased in livers of diabetic mice and humans, levels of FAM3B (PANDER) in serum, pancreatic islets, and liver were increased. This differential regulation of the individual FAM3 family members in a diabetic state is discussed to overall reduce AKT-and AMPK-activity thereby increasing hepatic gluconeogenesis and lipogenesis and ultimately leading to the development of NAFLD [8]. If FAM3A or FAM3C were overexpressed in obese and diabetic mice an attenuation of hepatic lipid deposition, hepatic gluconeogenesis and an improvement of global insulin sensitivity can be observed [9,10]. Both proteins were reported to contribute to a healthier metabolic phenotype by activation of the AKT-pathway in the liver independently of insulin [9,10]. Based on our data, we can conclude that FAM3D globally and specifically in the liver induced a similar beneficial metabolic phenotype as FAM3A and FAM3C although no effect on AKT signaling in the liver could be observed upon chronic FAM3D overexpression. A recent publication showed that FAM3D can inhibit gluconeogenesis determined in a CCK8 assay in high glucose induced HepG2 cells [25]. In accordance the sodium pyruvate tolerance test suggested reduced gluconeogenesis in liver upon overexpression of FAM3D. However, the gold standard for measuring endogenous glucose production the hyperinsulinemic euglycemic clamp study in mice overexpressing FAM3D did not indicate any effects on EGP but showed an insulin dependent increase in glucose turnover and uptake into various tissues. In context of the clamp data the results of the pyruvate tolerance test should rather be interpreted as consequence of improved insulin sensitivity leading to better tissue clearance of glucose than as reduced gluconeogenesis in liver.
We did not observe changes in bodyweight or glucose homeostasis upon FAM3D knockout. In light of existing data on other gut peptides, this finding was not surprising. For example, knockout of a single incretin receptor, e.g. GLP-1R -/-or GIPR -/-, led to only a mild phenotype, while a double incretin receptor knockout (DIRKO) displayed a much stronger phenotype [26][27][28]. Genetic ablation of the corresponding receptor is an often applied strategy to study the role of a peptide on a molecular level. In this respect it would be very interesting to identify the receptor(s) of FAM3D and to study the role the ligand-receptor pair(s) play(s) in the regulation of energy metabolism. Peng et al. identified FAM3D as a chemotactic agonist for the formyl peptide receptors FPR1 and FPR2, primarily expressed on monocytes and neutrophils [14]. In the current study no attraction of immune cells upon FAM3D overexpression was observed. Also as T2D is characterized by a low grade inflammation, a pro-inflammatory protein would rather worsen the diabetic phenotype, which contradicts the improved glucose metabolism found here [29]. Hence, we propose the existence of alternative receptor(s) mediating the positive metabolic phenotype of FAM3D. Future studies are needed to find this/these FAM3D receptor(s).
We identified receptor tyrosine kinase signaling as a possible molecular mode of action of FAM3D. Receptor tyrosine kinases play an essential role in the regulation of cell survival and metabolism, proliferation and differentiation as well as cell migration and cell cycle control [30]. Thus, the observed reduction in receptor tyrosine kinase signaling is in accordance with the observed reduced activity of cell cycle regulating kinases. The downregulation of overall kinase activity could be explained by our model. Upon ligand binding many receptor tyrosine kinases undergo endocytosis. From the endosomes the receptors are then either recycled to the cell surface or they are degraded by lysosomes [31]. With our long-term chronic overexpression model, it is likely that we established a constantly active negative feedback loop resulting in downregulation of kinase activity. Receptor tyrosine kinases are involved in several cancer types and various receptor tyrosine kinase inhibitors are approved for malignant therapy. In the context of these therapies, it became evident that tyrosine kinase inhibitors also have a positive effect on T2D and even type 1 diabetes (T1D). Several patient reports, in vivo and in vitro experiments, which are reviewed here [32][33][34], displayed great potential of tyrosine kinase inhibitors in the improvement of glucose metabolism. This is in accordance with the observed downregulation of receptor tyrosine kinase signaling upon chronic FAM3D overexpression and the observed positive glucometabolic phenotype. Potential underlying molecular mediators would be, amongst others, the platelet derived growth factor receptor (PDGFR) and vascular endothelial growth factor receptor (VEGFR) families. Both appeared within the top regulated kinases upon FAM3D overexpression. Following inhibition of PDGFR an increase in insulin sensitizing adiponectin can be observed [35] while the reduced vascular remodeling after VEGFR inhibition is causing an amelioration of T2D associated inflammation [36]. However, neither adiponectin, nor VEGFA levels were changed upon FAM3D overexpression, suggesting a different mechanism of action for FAM3D. Another possible mediator of the glucometabolic phenotype might be the MAPK/ERK pathway. ERK1/2, was significantly down-regulated upon chronic FAM3D treatment and it was shown that MEK inhibitors greatly improve insulin sensitivity in diet induced obese mice [37,38]. A reduction in the ERK-mediated negative feedback loop signaling to insulin receptor substrate 1 (IRS1) is discussed as one possible molecular mechanisms for this effect amongst others [39]. Alternatively, the inhibition of MAPK/ERK was described to reduce lipolysis in adipose tissue, ameliorating the release of insulin resistance driving free fatty acids [40]. However, no evidence for either of these two options could be observed following FAM3D overexpression.
Various other molecular mechanisms for the anti-diabetic effects of tyrosine kinase inhibitors are also discussed, but additional studies are needed to further elucidate them. Studies aimed at identifying the receptor(s) and the underlying signaling mechanism of FAM3D are required as well. Both should help to unravel the pathways regulating the positive metabolic effects of FAM3D.

Conclusion
In summary FAM3D was identified as gut secreted protein which improves glucose metabolism in diet induced obese mice. Upon the overexpression of FAM3D a global insulin sensitizing effect could be observed. This resulted not only in improved fasting blood glucose levels, glucose and insulin tolerance, but also in an increased glucose uptake into various tissues during hyperinsulinemic euglycemic clamps. Further, an amelioration of hepatic lipid deposition and steatosis was observed. Thus, FAM3D might be a potential candidate for the treatment of insulin resistance and hepatic lipid deposition. While we could identify a downregulation of receptor tyrosine kinase-signaling as a possible underlying molecular mechanism further studies are needed to fully elucidate the pathway and receptor(s) mediating the positive metabolic effects of FAM3D.

Funding
FL und KMO were supported by SNF 310030B_182829.
CRediT authorship contribution statement CM, BH, HN, SW and CW designed the study; CM, KG, MB, CH, LB, PK, EH performed the experiments; KMO, FL supported the flow cytometry experiments and their analysis, MB provided the human plasma samples, BS did the liver steatosis analysis, TL produced the AAV and the recombinant protein, HK performed the analysis of the sequencing data. SW, CM, MB and CW wrote the paper. All authors reviewed and edited the manuscript.

Data Availability
Data will be made available on request.