Prohormone convertase 1/3 deficiency causes obesity due to impaired proinsulin processing

Defective insulin processing is associated with obesity and diabetes. Prohormone convertase 1/3 (PC1/3) is an endopeptidase required for the processing of neurotransmitters and hormones. PC1/3 deficiency and genome-wide association studies relate PC1/3 with early onset obesity. Here, we find that deletion of PC1/3 in obesity-related neuronal cells expressing proopiomelanocortin mildly and transiently change body weight and fail to produce a phenotype when targeted to Agouti-related peptide- or nestin-expressing tissues. In contrast, pancreatic β cell-specific PC1/3 ablation induces hyperphagia with consecutive obesity despite uncontrolled diabetes with glucosuria. Obesity develops not due to impaired pro-islet amyloid polypeptide processing but due to impaired insulin maturation. Proinsulin crosses the blood-brain-barrier but does not induce central satiety. Accordingly, insulin therapy prevents hyperphagia. Further, islet PC1/3 expression levels negatively correlate with body mass index in humans. In this work, we show that impaired PC1/3-mediated proinsulin processing, as observed in human prediabetes, promotes hyperphagic obesity.

body weight gain in baboons and rodents 13,14 . In contrast, inhibition of central insulin signaling increases food intake and body weight gain 15 . In humans, intranasal insulin application induces satiety and reduces body weight gain 16,17 . Furthermore, brain-specific insulin receptor ablation increases food intake in rodents 18,19 . Moreover, obesity is associated with central insulin resistance 20 or reduced insulin transport into the brain 21 . Since less than 1% of peripheral insulin is taken up by the brain 22 via a saturable receptor-mediated process 23,24 and proinsulin has a 100-fold reduced affinity for the insulin receptor 25 , it is conceivable that impaired proinsulin processing could lead to reduced central insulin action and thus hyperphagic obesity.
Here, we developed genetic models allowing spatiotemporal ablation of Pcsk1 in mice. We show that central PC1/3 deficiency leads to endocrinopathies, but only mildly affects body weight regulation. In contrast, pancreatic β cell-specific Pcsk1 ablation leads to profound hyperphagic obesity mediated by the lack of mature insulin. Furthermore, PCSK1 expression in human islets negatively correlated with body mass index. Our results suggest that PC1/3 deficiency leads to obesity due to the absence of insulin-targeted anorexic pathways.

Results
Induced whole-body PC1/3 deletion leads to pronounced hyperphagic obesity Human PC1/3 deficiency leads to severe obesity while Pcsk1 null mice are runted and have a high mortality rate due to developmental defects 8 . Consistent with these data, mice on a C57BL6/N genetic background carrying a knockout-first conditional-ready allele that disrupts Pcsk1 did not produce homozygous knockout mice when intercrossed (34 pups from 3 different breeding pairs, expected: 8 homozygous mice, obtained: 0). To circumvent embryonic lethality, we excised the FRT-flanked cassette and generated tamoxifeninducible whole-body PC1/3 knockout mice (Pcsk1 fl/fl UBC-Cre ERT2 ). Four weeks after induction, Pcsk1 expression in the hypothalamus and pancreatic islets was efficiently ablated, and PC1/3 protein was absent in the hypothalamus (Fig. 1A, B). In agreement with the clinical presentation of PC1/3 deficiency in humans, food intake increased by 16.9 ± 6.7% and mice of both sexes rapidly became obese (Fig. 1C, D). Glucose tolerance was impaired, likely due to a lack of mature insulin (Fig. 1E, F). Thus, Pcsk1 ablation in early adulthood in rodents phenocopies hyperphagic obesity as seen in human PC1/3-obesity 3,4 .

POMC-specific PC1/3 ablation induces mild and transient overweight
To test whether the lack of processed central satiety hormones mediates the obesity associated with PC1/3 deficiency, we broadly ablated PC1/3 in the brain. Since Nes-Cre 1Kln driver mice show reduced body weight compared to wild-type controls, we generated littermate mice that all express Cre. Surprisingly, neither heterozygous Pcsk1 fl/wt Nes-Cre nor homozygous Pcsk1 fl/fl Nes-Cre mice of either sex showed altered body weight development ( Fig. 2A). Challenging these mice with a high-fat diet also did not change body weight development or glycemia between genotypes ( Supplementary Fig. S1A, B). Similarly, Pcsk1 fl/fl Chat-Cre of both sexes had similar body weights compared to controls (Fig. 2B). Within the central nervous system, PC1/3 processes and thereby activates several anorexigenic propeptides including POMC to pro-adrenocorticotropic hormone (proACTH). POMC ablation is associated with obesity in humans and rodents 10,11 , highlighting impaired POMC processing as a prime potential mediator of PC1/3related obesity. Compared to controls, Pcsk1 fl/fl Pomc-Cre mice showed mild overweight which normalized at around 6 and 12 months of age in Fig. 1 | Induced whole-body PC1/3 deletion leads to pronounced hyperphagic obesity. A Normalized Pcsk1 expression in the hypothalamus and pancreatic islets isolated from 12-week-old mice 4 weeks after knockout induction. B PC1/3 protein expression in hypothalami isolated from 14-week-old mice 5 weeks after knockout induction. Note: signal at 55 kDa is unrelated to PC1/3 and known to appear in this batch of antibody. C Food intake pre (from 7 to 8 weeks of age) and post (10-11 weeks of age) induction at 8 weeks of age. Each dot represents the average of a cage with 2-3 mice. D Body weight development. E Circulating glucose concentration after injection of 2 g kg −1 glucose at 12 weeks of age 4 weeks post males and females, respectively (Fig. 2C). Western blotting of pituitary extracts showed that Pcsk1 fl/fl Pomc-Cre mice can process POMC to proACTH ( Supplementary Fig. S1C), suggesting that PC1/3 is not essential to process POMC. However, ACTH-stimulation testing revealed that Pcsk1 fl/fl Pomc-Cre mice suffer from adrenal insufficiency (Fig. 2D). Supporting this observation, serum glucocorticoid levels (corticosterone and 11-dehydrocorticosterone, corresponding to cortisol and cortisone in human, respectively) were moderately reduced in Pcsk1 fl/fl Pomc-Cre mice while mineralocorticoids (aldosterone, 11deoxycorticosterone and progesterone) did not differ from control mice ( Fig. 2E and Supplementary Fig. S1D). The androgens testosterone and androstenedione were reduced in Pcsk1 fl/fl Pomc-Cre mice, suggesting decreased gonadal steroidogenesis ( Supplementary  Fig. S1E). To check whether the unexpectedly mild gain in body weight in Pomc-specific PC1/3 knockout mice was due to compensatory mechanisms during development, we next deleted PC1/3 in POMCexpressing cells in adulthood. Male mice developed mild late-onset obesity, while this did not reach statistical significance in females (Fig. 2F).
The orexigenic hormone AgRP is processed by PC1/3 26 and ablation of AgRP itself or AgRP/NPY neurons in adult mice leads to cessation of feeding [27][28][29] . However, Pcsk1 fl/fl AgRP-Cre mice had a normal body weight development (Fig. 2G). Importantly, Cre-only control mice of all Cre driver lines used so far did not show a body weight phenotype ( Supplementary Fig. S1F-J). Taken together, these data suggest that in rodents whole-body but not neuronal ablation of PC1/3 leads to profoundly increased body weight development. PC1/3 deficiency-related obesity must therefore be mediated by other, potentially peripheral, PC1/3 targets.
β cell-specific Pcsk1 ablation blocks proinsulin processing and induces hyperglycemia but not overt diabetes One cell type that could mediate the obesity related to PC1/3 deficiency is the pancreatic β cell. Islet β cells secrete insulin and islet amyloid polypeptide (IAPP), both of which are dependent on PC1/3 processing and are regulators of satiety 13,30 . Therefore, we generated Pcsk1 fl/fl Pdx-Cre ERT mice. Nearly all (98%) β cells expressed Cre recombinase-mediated reporter fluorescence ( Supplementary Fig. S2A). Pcsk1 gene expression was strongly reduced while Pcsk2 and Cpe (encoding the insulin processing proteases PC2 and CPE, respectively) were upregulated and the prohormone convertase inhibitors Pcsk1N and Scg5 were unchanged in fluorescence-activated cell sorting (FACS)-sorted β cells ( Fig. 3A and Supplementary Fig. S2B). PC1/3 protein was efficiently ablated in isolated pancreatic islets and pancreatic sections (Fig. 3B, C). This effect was long-lasting, as islet Pcsk1 was still absent 75 weeks post knockout induction ( Supplementary  Fig. S2C). During development, Pdx is expressed in the pancreas, in certain brain regions, as well as in the gastrointestinal L cells (where PC1/3 activates the satiety hormone GLP-1) but is limited to the β cell and acinar cells in adulthood [31][32][33] . We, therefore, induced mice at 8 weeks of age. Circulating concentrations of active GLP-1, intestinal GLP-1 content, as well as Pcsk1 gene expression in the hypothalamus, pituitary, olfactory bulb, amygdala, and adrenal gland were unchanged in Pcsk1 fl/fl Pdx-Cre ERT mice ( Supplementary  Fig. S2D-F). These data demonstrate that Pcsk1 fl/fl Pdx-Cre ERT mice are an efficient model for β cell-specific Pcsk1 ablation. Pcsk1 fl/fl Pdx-Cre ERT mice were unable to produce mature insulin, whereas proinsulin-related peptides were markedly increased in the islets as well as in the circulation, and further increased over time, likely to compensate for the lack of mature insulin (Fig. 3D, E). Insulin gene expression, islet number, and size as well as pancreas weight were increased, resulting in doubled β-cell mass (Fig. 3F-H and Supplementary Fig. S2G-I). Furthermore, β cells had a degranulated appearance, suggesting that they were metabolically stressed ( Supplementary Fig. S2J). Protein kinase B (AKT) was still phosphorylated in response to glucose in fat and liver, albeit to a lesser extent than in control mice, likely due to immature insulin's bioactivity 25 (Fig. 3I-K). Mice of both sexes were hyperglycemic but the survival frequency remained comparable to control littermates (Fig. 3L). Glucose tolerance was impaired and further deteriorated with increased age and high-fat diet feeding, yet mice of both sexes did not become catabolic ( Fig. 3M and Supplementary Fig. S2K, L). Pcsk1 fl/fl Pdx-Cre ERT mice prior to induction, heterozygous knockout mice, and Cre controls did not show glucose intolerance ( Fig. 3M and Supplementary Fig. S2M). These data show that Pcsk1 fl/fl Pdx-Cre ERT mice exhibit features of diabetes but counteract hyperglycemia by increasing the production and secretion of insulin precursors.

Deficiency in proinsulin processing mediates hyperphagic obesity
Similar to the inducible whole-body PC1/3 knockout mice (Fig. 1D), Pcsk1 fl/fl Pdx-Cre ERT mice of both sexes showed a marked increase in body weight following knockout induction, despite a significant loss of calories due to hyperglycemia with subsequent glucosuria (Fig. 4A, B). Uninduced mice and Cre-only controls did not show altered body weight development ( Supplementary Fig. S3A, B). The increase in body weight was driven by a 12% enhanced caloric intake ( Fig. 4C and Supplementary Fig. S3C, D) and led to gradually enlarged fat pads, but not to a generally increased body size (Fig. 4D-F). Physical activity, energy expenditure, body core temperature, respiratory exchange ratio, and UCP1 expression in brown adipose tissue were not changed ( Fig. 4G-K and Supplementary  Fig. S3E-H), suggesting that energy output is not affected by β cellspecific ablation of PC1/3. Next, we investigated whether processing deficits in IAPP mediate the observed obesity phenotype. Surprisingly, Pcsk1 fl/fl Pdx-Cre ERT mice showed circulating levels of fully processed IAPP and the concentration increased steadily with age ( Fig. 4L). Western blotting with antibodies that react to all species of IAPP ( Fig. 4M) or only to fully processed IAPP (Fig. 4N) confirmed that Pcsk1 fl/fl Pdx-Cre ERT mice process proIAPP in the absence of β cell PC1/3. IAPP and insulin are co-secreted from β cells 34 and normalizing proinsulin secretion by implanting Pcsk1 fl/fl Pdx-Cre ERT mice with subcutaneous insulin pellets prevented the increase in circulating IAPP (Fig. 4O), suggesting that the increase in circulating IAPP is secondary to increased proinsulin secretion. Exogenous recombinant IAPP reduced food intake in control but not in Pcsk1 fl/fl Pdx-Cre ERT mice ( Supplementary Fig. S3I), suggesting that the constant exposure to high circulating levels of IAPP might induce resistance to central IAPP action. To test this hypothesis, we made Pcsk1 fl/fl Pdx-Cre ERT mice on an Iapp null background and found that the body weight phenotype does not depend on the presence of IAPP (Fig. 4P). IAPP also slows gastric emptying but despite high peripheral levels of IAPP, the rate of gastric emptying was unchanged ( Supplementary Fig. S3J). Furthermore, circulating satiety hormones derived from other organs (FGF21, ghrelin, leptin, PYY) were unchanged ( Supplementary Fig. S3K). Finally, ghrelin expression in the stomach of Pcsk1 fl/fl Pdx-Cre ERT mice remained unchanged (Supplementary Fig. S3L). These results indicate that obesity in β cellspecific PC1/3 knockout mice is mediated by increased food intake and that this effect is independent of IAPP. Incompletely processed proinsulin crosses the blood-brain barrier but does not induce central satiety Next, we investigated whether impaired proinsulin processing may mechanistically underlie hyperphagic obesity. Pcsk1 fl/fl Pdx-Cre ERT were implanted subcutaneously with insulin pellets which prevented hyperglycemia and hyperphagia (Fig. 5A, B). Since insulin crosses the blood-brain barrier in a receptor-mediated manner 23,24 and proinsulin has a low affinity for the insulin receptor 25 , we hypothesized that proinsulin is unable to reach the brain and induce central signaling. Therefore, we assessed plasma and cerebrospinal fluid (csf) and found proinsulin-related peptide levels to be undetectable in csf from control mice but readily measurable in csf from Pcsk1 fl/fl Pdx-Cre ERT mice, on average corresponding to 2.5% of circulating levels (Fig. 5C). Intracerebroventricular injection of insulin into the third ventricle reduced food intake but this was not observed for an equimolar dose of proinsulin (Fig. 5D). These data imply that immature insulin can cross the blood-brain barrier but is unable to induce central satiety.

Islet PCSK1 expression correlates with body mass index in humans
In humans, inactivation mutations of PC1/3 lead to severe obesity 3 . However, the contribution of tissue-specific PCSK1 expression is unknown. Therefore, we next assessed RNA expression levels in a published RNA-seq data set of human islets 35 . Body mass index negatively correlated with PCSK1 and CPE expression but not with PCSK2 or SCG5, suggesting that islet PC1/3 could be an important determinant of body weight regulation in humans ( Fig. 6A-C).

Discussion
Our work shows that impaired proinsulin processing is an important contributor to PC1/3-related hyperphagic obesity. While central PC1/3 deficiency only mildly and transiently increased body weight development, β cell-specific ablation of PC1/3 induced pronounced obesity in our preclinical mouse models and insulin therapy prevented hyperphagia.
Peripheral signals that influence energy intake are processed by central circuits and in this process, hypothalamic POMC neurons are a major integration site. Activation of POMC suppresses appetite 36 , while POMC inhibition and mutations that reduce POMC activity induce obesity in rodents, dogs, and humans 10,11,36,37 . Since PC1/3 cleaves POMC to the anorectic α-MSH, the canonical view is that PC1/3-related obesity could be regulated by reduced processing of POMC. However, there seems to be a disconnect between POMC processing and body weight development. Hypothalamic extracts isolated from the obese Pcsk1 N222D mutant mouse still have 55% α-MSH content 9 but PC2 null mice who produce no α-MSH 38 are not obese 39 . Furthermore, pituitaries isolated from whole-body PC1/3 knockout mice still have 50% of α-MSH content 8 and ablation of PCSK1 in human embryonic stem cellderived hypothalamic neurons had no effect on α-MSH content 40 . Thus, PC1/3 is dispensable for POMC processing and the latter does not necessarily regulate body weight development. Supporting this notion, deletion of PC1/3 specifically in POMC neurons in adulthood did not change body weight development for up to 9 weeks after knockout induction 41 . We confirmed these findings but in addition observed mild late-onset obesity in aged mice. Furthermore, we found life-long Pcsk1 knockout in POMC neurons to induce obesity which resolved after 6 and 12 months of age in male and female mice, respectively. These mice had detectable levels of ACTH but an insufficient amount for normal adrenal function. Gonadal androgens were reduced in these mice but the breeding performance was normal. Similarly, humans with PC1/3 deficiency suffer from hypogonadism 1,3 and our data suggests that this is partly mediated by defective PC1/3 activity in POMC neurons.   There are considerable phenotypic differences in the presentation of PC1/3 deficiency between humans and rodents. Pcsk1 null mice are runted or suffer from preimplantation lethality 8,42 and a Cmv-Cre version of our mouse model is also not viable. This apparent contrast to patients with PC1/3 deficiency could arise from the fact that levels of adrenal androgens are high during early development in humans but mice solely depend on gonadal androgen synthesis. Thus, mice might display a more severe pathology, as a lack of gonadal androgens in humans may be partly compensated for.
In contrast to our whole-body PC1/3 knockout model, broad or targeted central knockout lines unexpectedly had only mild transient body weight phenotypes. Therefore, we reasoned that there might be an important, potentially peripheral, PC1/3 target that regulates body weight development. The pancreatic β cell seemed to be a logical candidate, as its major products, insulin and IAPP are processed by PC1/3 and both activate central satiety signaling pathways. Indeed, Pcsk1 fl/fl Pdx-Cre ERT mice with knockout induction in adulthood quickly became hyperphagic and obese. Conceptually, these mice represent a model of severe type 2 diabetes, as they increase the production and secretion of less-active and immature insulin to keep glycemia under control. Caloric intake was elevated and despite the loss of calories due to glucosuria they gained body weight in excess. They phosphorylated AKT and did not develop life-threatening hyperglycemia, nor did they become catabolic even though these mice lacked fully processed insulin. In contrast, a lack of both mature and immature insulin, as observed in the context of type 1 diabetes, leads to severe catabolism. Noteworthy, defective islet prohormone processing and increased proinsulin levels were also described in individuals prior to and after the onset of type 1 diabetes (reviewed: 43 ). However, it is currently unknown whether an islet intrinsic prohormone processing defect predisposes individuals to develop autoimmune diabetes or whether increased secretory demand due to autoimmune β-cell destruction leads to a relative increase in proinsulin to insulin ratio.
The investigation of the β cell-derived satiety hormone IAPP in Pcsk1 fl/fl Pdx-Cre ERT mice yielded unexpected results. First, in contrast to proinsulin, proIAPP was processed in the absence of PC1/3, possibly by PC2. Indeed, β cell expression of PC2 and CPE, two other proteases involved in proinsulin/proIAPP processing, were increased. Second, circulating levels of IAPP were increased and this was prevented by insulin therapy, confirming that insulin and IAPP are co-secreted 34 . Despite very high circulating levels of endogenous IAPP, these mice were still obese, suggesting that IAPP did not induce satiety. Third, Pcsk1 Iapp double knockout mice had the same body weight phenotype as Pcsk1 single knockout mice expressing Iapp. This shows that the obesity phenotype was not due to IAPP resistance induced by constant high circulating levels of IAPP. In this context, it is interesting that in contrast to proinsulin, proIAPP was not found to be elevated in individuals with type 2 diabetes [44][45][46][47] . These data prove that in our mouse model IAPP did not regulate obesity.
On the other hand, impaired proinsulin processing mediated the obesity phenotype in Pcsk1 fl/fl Pdx-Cre ERT mice, and insulin replacement by implanting subcutaneous insulin pellets prevented hyperphagia. The anorectic effect of central insulin 13,14,16,17,48 and the orexigenic effect of central insulin deficiency 15,18-20 is well described (reviewed: 49 ). Since insulin uptake into the brain 22 is a saturable receptor-mediated process 23,24 , we hypothesized that Pcsk1 fl/fl Pdx-Cre ERT mice become obese because incompletely processed insulin does not efficiently induce central satiety. Surprisingly, we readily detected proinsulinrelated peptides in the csf of Pcsk1 fl/fl Pdx-Cre ERT mice, suggesting that these peptides can cross the blood-brain barrier despite a 100-fold reduced affinity for the insulin receptor 25 . However, in support of our hypothesis, central injection of proinsulin did not induce satiety while injection of insulin did, despite the much longer half-life of proinsulin 50 . While a lack of central insulin action is the most straightforward explanation for hyperphagia in these mice, we can currently not rule out that hyperproinsulinemia per se induces hyperphagia and that insulin therapy prevented this phenotype indirectly by reducing βcell secretory activity and thus circulating proinsulin levels. An argument against this hypothesis is the observation that AKT phosphorylation in insulin target tissues was reduced and not increased in Pcsk1 fl/fl Pdx-Cre ERT mice.
Although Pdx expression in adulthood is largely limited to β cells 32,33 , Cre recombination-mediated reporter gene expression in certain Pdx-Cre driver mice was also detected in the duodenum and in hypothalamic neurons 51,52 , raising the possibility that the observed hyperphagia could be mediated by non-β cell targets. Nevertheless, the levels of the intestinal PC1/3 target GLP-1 were not altered and we, as well as others 53 , have found the targeted gene not to be reduced in the hypothalamus when Pdx-Cre ERT(Dam) driver mice were studied. Furthermore, none of the neuron-specific PC1/3 knockout models recapitulated the pronounced obese phenotype observed in whole-body or Pcsk1 fl/fl Pdx-Cre ERT mice and insulin replacement therapy prevented hyperphagia. These data strongly suggest that Pcsk1 deficiency in pancreatic β cells mediated the observed obesity phenotype. Noteworthy, both constitutive and adult-onset deletion of PC1/3 in POMCexpressing tissues induced obesity, although only mildly and transiently, showing that POMC processing regulates body weight development.
The complete absence of PC1/3 in humans [3][4][5]54 and in our inducible whole-body knockout mouse model induces hyperphagic obesity. However, also a partial reduction in Pcsk1 expression seems to regulate body weight development. Mouse models with reduced PC1/3 activity develop obesity 9,55 and in human nonsynonymous polymorphisms in PCSK1 confer risk of obesity 6,7 . Interestingly, prediabetes and type 2 diabetes, which are associated with obesity, are characterized by reduced PCSK1 expression in pancreatic islets 35 and impaired proinsulin processing, leading to an increased proinsulin to insulin ratio [45][46][47] . Insulin knockout mice are not viable 56 but Ins1 or Ins2 null mice have normal circulating insulin levels and do not develop hyperglycemia or obesity 57 . Thus, reducing insulin dosage per se is not sufficient to alter body weight development. The increased insulin secretion observed in prediabetes may be the consequence of an increased insulin demand due to obesity-associated insulin resistance. Still, the existence of an intrinsic insulin secretion defect was proposed and is supported by GWAS studies 58,59 . Of particular interest are the above-mentioned polymorphisms in PCSK1 that are associated with obesity 6,7 . It is conceivable that genetic predisposition to altered insulin expression and processing may promote obesity and not (only) be the consequence thereof 60 . Furthermore, Prader-Willi Syndrome, a genetic disease characterized by strong hyperphagia is associated with reduced levels of PC1/3 in islets 61 . Supporting a role for β cell-derived PC1/3 in human body weight regulation, we found body mass index to be negatively associated with islet PCSK1 expression. Of note, the β cell endopeptidase CPE but not PCSK2 also negatively correlated with body mass index. Interestingly, Cpe null mice are heavily obese and lack mature insulin 62 , while PC2 null mice are not obese 63 .
Taken together, our data reveal that the major mechanism for PC1/3 deficiency-associated obesity is not a defect in central processing and we suggest that the pancreatic β-cell-derived insulin is a major player in this pathology. Therefore, a judicious exogenous insulin therapy provided in a near-physiological dose may prevent obesity in affected individuals by inducing satiety without promoting exaggerated fat deposition.

Mice
All animal experiments were performed in accordance with the federal laws of Switzerland and approved by the cantonal veterinary office of Basel (approval numbers 2511 and 3045). Whole-body Pcsk1 knockout mice were derived from embryonic stem cells carrying a transgene Article https://doi.org/10.1038/s41467-022-32509-4 encoding an FRT-flanked lacZ and neo cassette with loxP sites around exon 5 of the Pcsk1 gene (EUCOMM program) and these mice were subsequently crossed to FLP-transgenic mice 64

Body weight development and food intake
Body weight was measured weekly using the dynamic weighing feature that averages measurements taken over a 5 s period (Scout STX, Ohaus, Nänikon, Switzerland). To increase readability, not all body weight data are depicted in the figures. Food intake was assessed by weekly measurement of the food given/left in the rack feeder of the cage.

Glycemia and glucose tolerance
Blood glucose was measured with a glucometer (FreeStyle Freedom Lite, Abbott, Baar, Switzerland) from a drop of blood obtained from a small incision into the tail vein. Fasting blood glucose was obtained in animals deprived of food from 8:00 to 14:00. Blood was collected into heparinized capillaries by tail-tip bleeding and transferred to tubes containing EDTA, stored on ice, followed by centrifugation at 10,000 × g for 5 min at 4°C and storage at −20 or −80°C. Terminal blood was collected by cardiac puncture through the skin. Glucose tolerance was assessed by intra-peritoneal or oral application of 2 g kg −1 glucose after fasting from 8:00 to 14:00. Blood glucose was assessed prior to and 15, 30, 60, 90, and 120 min after glucose administration. Baseline blood glucose was measured once, all other timepoints twice. When the concentration of blood glucose was above the measurable range, 2 ul blood was mixed with 2 ul diluent 100 (R50AA-4, Meso Scale Diagnostics, Rockville, MD, USA) and measured by a glucometer. If the obtained values differed by more than 2 mM, a third measurement was obtained and the value furthest away was excluded. Initially, plasma insulin was assessed at 0, 15 and 30 min of the glucose tolerance test (K152BZC, Meso Scale Diagnostics) but this assay seemed to be highly cross-reactive with split proinsulins and therefore the data obtained from UBC-Cre ERT2 Pcsk1 fl/fl and Pdx-Cre ERT Pcsk1 fl/fl could not be compared to their Pcsk1 fl/fl controls, making a useful interpretation obsolete.

ACTH stimulation test
Mice were handled daily for at least 2 weeks to minimize stress responses. A blood sample was collected by tail-tip bleeding at 9:00, followed by intra-peritoneal injection of 10 μg kg −1 synthetic human ACTH (Synacthen). A second blood sample was collected 1 h later and plasma corticosterone was assessed (ADI-900-097, Enzo Life Sciences, Lausen, Switzerland).

Gastric emptying test
Mice were deprived of food from 8:00 to 14:00, followed by oral gavage of 2 g kg −1 glucose and 100 mg kg −1 acetaminophen (A5000, Sigma-Aldrich). Plasma levels of acetaminophen were assessed using the Paracetamol Three Reagent System (K8002, Cambridge Life Sciences).

RNA isolation and real-time quantitative PCR
Total RNA of various tissues was extracted using the NucleoSpin RNA Kit (740955, Macherey Nagel, Oensingen, Switzerland) and reverse transcribed using random hexamers and GoScript reverse transcriptase (A2801, Promega, Dübendorf, Switzerland). Quantitative real-time PCR was performed using GoTaq qPCR (A6002, Promega) and primer sets designed by Primer-BLAST (NCBI, Bethesda, MD, USA) manufactured by Microsynth (Balgach, Switzerland), run on an ABI 7500 Fast Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA). Whenever possible, primers were designed in an exon-exon spanning manner, and the melt curve, product size, and purity, as well as a serial dilution, were checked to ensure specificity. Primer sequences are listed in Supplementary Table S1. Gene expression was analyzed using the comparative 2 ΔΔCT method using actin, gapdh, and 18S as housekeeping genes.

Protein isolation from tissue and western blotting
Tissues were dissected and dissolved in protein lysis buffer (20 mM Tris pH = 7.5, 1% Triton X-100, 150 mM NaCl, 10% glycerol, 1 mM Na3VO4, 10 mM NaF, 5 mM PMSF, 1 mM EDTA pH = 8.0) containing protease inhibitor cocktail (11836145001, Sigma-Aldrich), incubated for 15 min on ice, sonicated for 2 min (Elmasonic X-tra 150 H) and centrifuged for 10 min at 10,000 × g at 4°C. The supernatant was stored at −80°C. Protein content was assessed by a BCA assay (23225, Thermo Fisher Scientific). Then, 5-20 ug of the lysates were mixed with sample buffer (NP0007, Thermo Scientific Fisher) and sample reducing agent (NP0009, Thermo Scientific Fisher), loaded onto a gel (4-12% NP0336 or 12% NP034, Thermo Scientific Fisher) and separated and blotted onto a nitrocellulose membrane (9004700, BIO-RAD, Cressier, Switzerland) with the Invitrogen XCell Surelock Mini-Cell Electrophoresis System. The membrane was blocked in 5% milk in TBST, thereafter it was stained with the primary antibody at 4°C overnight. On the next day, a secondary antibody coupled to horseradish peroxidase was added and incubated for 2 h at room temperature. SuperSignal™ West Pico PLUS Chemiluminescent Substrate was added and the membrane was imaged with the Fusion FX system (VILBER, France).

Immunohistochemistry
Pancreata were dissected, weighed, and incubated in 4% buffered formalin (Formafix AG, Hittnau, Switzerland) overnight at 4°C, thoroughly washed with PBS, processed on a TPC 15 DUO, and embedded in paraffin on a TES Valida (both: Medite, Dietikon, Switzerland) and sectioned into 5 μm slices on an HM340E microtome (Thermo Fisher Scientific). Tissue sections were deparaffinized, rehydrated, and incubated with antibodies raised against insulin, CD45 (to visualize lymph nodes), PC1/3, and DAPI (to visualize nuclei). Detailed information about the antibodies used is given in Supplementary Table S2. Slides were mounted using a fluorescence mounting medium (S3023 Agilent Dako, Basel, Switzerland) and analyzed.

Islet isolation
Mice were euthanized and injected with collagenase (4189, Worthington, Lakewood, NJ, USA) through the bile duct. Pancreata were excised and incubated at 37°C for 30 min with occasional shaking. The reaction was quenched by the addition of an HBSS medium containing 0.5 w/v BSA. The solution was filtered and islets resuspended in RPMI-1640 medium containing 11.1 mM glucose, 100 units ml −1 penicillin, 100 μg ml −1 streptomycin, 2 mM glutamax, 50 μg ml −1 gentamycin, 10 μg ml −1 fungizone and 10% FCS. Islets were hand-picked, lysed in RNA lysis buffer, and stored at −80°C until RNA was isolated as described above.

β-cell mass
Tissue sections from three to four different depths of the pancreas (at least 100 μm apart) were prepared as described above. Pancreas sections were imaged using an automated Eclipse Ni microscope (Nikon, Egg, Switzerland) with an Orca Flash 4.0 camera (Hamamatsu Photonics, Solothurn, Switzerland), a 4× Objective NA 0.2 (Nikon), and a PL-200 slide loading robot (PRIOR Scientific, Fulbourn, UK) using NIS Elements software v5. Ilastik software 75 was trained to segment the pancreas area of stitched multichannel images, including channels for nuclei (DAPI), autofluorescence (GFP), lymph nodes (Alexa-555), and insulin (Alexa-647). The simple segmentation outputs of ilastik and corresponding original multichannel images were subsequently processed using Fiji 76 . In a first step, the ilastik simple segmentation recognition of the pancreas was checked and corrected manually if needed. In a second step, the β-cell identification on the corresponding channel was obtained by setting all pixels that do not overlap with the pancreas area to 0, thresholding using the Otsu method and performing a connected component analysis on objects bigger than 900 µm 2 and having a circularity value bigger than 0.2. Insulin object recognition was adjusted manually if needed. All measurements (pancreas area and individual insulin areas) were saved to a spreadsheet. The Fiji analysis workflow was facilitated using a custom ImageJ script, which is available upon request from the corresponding author. The script was run on a remote-access server system (University of Basel). β-cell mass was calculated as pancreas weight (mg) × sum (insulin + area)/sum (pancreas area). All samples were handled in a blinded fashion by the same researcher.

Flow cytometry and fluorescence-activated cell sorting (FACS)
Following pancreatic islet isolation, islets were washed and resuspended in 900 µl PBS + 0.5 mM EDTA. 200 µl digestion solution (0.01% trypsin + 0.5 mM EDTA) was added and incubated at 37°C for 90 s with manual pipetting every 30 s. Nine milliliter cold FACS buffer (PBS + 0.5% BSA + 5 mM EDTA) was added to stop the digestion and the solution was centrifuged 350 × g for 5 min. The supernatant was removed and the cells were resuspended in 200 µl FACS buffer. Fc receptors were blocked by the addition of 2 µl anti-CD16/CD32 antibody and incubated for 15 min on ice. Immune cells were labeled by the addition of 2.5 µl CD45-APC antibody and incubation for 30 min. Cells were washed and resuspended in 100 ul FACS buffer and analyzed on a CytoFLEX flow cytometer (Beckman Coulter, Nyon, Switzerland) or a BD Fortessa (BD Biosciences, San Jose, USA) using FlowJo (v10.6.1, FlowJo LLC). If cells had to be sorted, a BD SORPAria 3, a BD FACSAria 3, or a BD Influx cell sorter (BD Biosciences) was used. See Supplementary Table S2 for more information about antibodies used.

Hormone measurements
Proinsulin-related peptides were assessed by a proinsulin ELISA (10-1232-01, Mercodia, Uppsala, Sweden). According to the manufacturer's instructions, this assay does not cross-react with mouse mature insulin or C-peptide but our data suggest that it also detects split proinsulins. IAPP was measured by ELISA (human mature IAPP, EZHA-52K, Millipore, St. Charles, MO, USA). This assay employs a monoclonal capture antibody (F002; binds all molecular forms of proIAPP) and a monoclonal detection antibody (F025; binds amidated IAPP at its C terminus) 77 and is 100% cross-reactive to rodent IAPP. To determine active GLP-1, mice were injected with 25 mg kg −1 sitagliptin (sc-364620, Santa Cruz Biotechnology, Dallas, USA) 30 min prior to oral glucose loading (2 g kg −1 glucose) and blood was collected in tubes containing Diprotin A (H-38250050, Bachem, Bubendorf, Switzerland). Active GLP-1 was assessed by the V-PLEX assay (K1503OD-1, Meso Scale Diagnostics) and serum levels of ghrelin (active), leptin, and PYY (total) were analyzed using the U-PLEX metabolic kit (K152ACL-1, Meso Scale Diagnostics). To determine the ileal content of active GLP-1, 3 cm of the terminal ileum was excised, rinsed with PBS, extracted in 0.18 N HCl in 70% EtOH, and assessed with the kit described above.

Glucosuria
To assess glucose concentration in the urine, mice were restrained and urine was collected with a pipette and frozen at −20°C. Samples were treated with glucose oxidase (G7141, Sigma-Aldrich) and peroxidase (P8250, Sigma-Aldrich), and the resulting H 2 O 2 concentration was quantified by a colorimetric assay using glucose as standard and employing 4-aminoantipyrine (A4382, Sigma-Aldrich) and sodium 3,5dichloro-2-hydroxybenzenesulfonate (D4645, Sigma-Aldrich) as coloring reagents. The plate was read at 520 nm on a Synergy H1 Hybrid Multi-Mode Reader (BioTek, Winooski, VT, USA).

Insulin therapy
One week after the first tamoxifen administration, mice were briefly anesthetized with isoflurane, shaved, and wiped with betadine (Mundipharma, Basel, Switzerland). An 18G needle was used to pinch the skin and a trocar was inserted subcutaneously. Afterward, three insulin pellets (or blanks for controls) were briefly immersed into diluted betadine and transplanted into the flank of the mouse via the previously inserted trocar (LinShin, Toronto, Canada). Pilot studies in streptozotocin-induced C57BL/6N hyperglycemic mice showed that three insulin pellets are needed to normalize blood glucose and that the achieved normoglycemia persisted for about 21 days.

IAPP sensitivity testing
Mice were single-caged at 5-6 weeks of age and acclimatized to eating from removable custom-made food hoppers. At 10 weeks of age, mice were fasted overnight and intra-peritoneally injected with 20 μg kg −1 recombinant rodent IAPP (a gift from Amylin Pharmaceuticals, San Diego, CA, USA), or saline at 9:00. Food hoppers were weighed and returned to the cage 15 min after injection and weighed again 1 and 2 h later. The studies were done in a cross-over design with 1 week washout period between tests.

Implantation of telemetric sensors and indirect calorimetry
Mice were induced at 8 weeks of age and TA-F10 sensors for temperature and activity measurements (Data Sciences International, St. Paul, MN, USA) were implanted intra-peritoneally under isoflurane anesthesia at 10 weeks of age. Thereafter, mice were single-housed. At 11 weeks of age, mice were transported from the University of Basel to the University of Zürich (Tierspital, laboratory of Thomas A. Lutz) and placed in a PhenoMaster open circuit indirect calorimetry system (TSE Systems, Bad Homburg, Germany). The system was calibrated using precision calibration gases prior to use. Air flow rate was 0.41 l min −1 . Mice were acclimatized to the new environment for 7 days and then oxygen consumption and carbon dioxide production rates were sampled at 20-min intervals for 3 days. From these data, energy expenditure and respiratory exchange ratio were calculated. Physical activity and body temperature were measured every 5 min and analyzed using Dataquest A.R.T. 3.1 software (Data Sciences International).

Cannula surgeries, intracerebroventricular injections, and refeeding experiments
12-14-week-old female wild-type C57Bl/6N mice were anesthetized by injection of 65 mg kg −1 body weight ketamine (Labatec Pharma SA) and 12 mg kg −1 body weight xylazine (Streuli Pharma AG) in saline. Viscotears (Lacrinorm, Bausch & Lomb Swiss AG) were applied to the eyes, and the hairs on the top of the skull were shaved. Animals were then affixed to a stereotaxic frame (Kopf). The skin was incised to expose the skull and the periostium was removed by application of acetone. A burr hole was placed at coordinates (x: −0.7 mm; y: −1.3 mm) from bregma. The guide cannula (C315GS-5/SP, Bilaney consultants) was implanted (x: −0.7 mm; y: −1.3 mm; z: −1.85 mm 78 ) to reach the lateral ventricle. To fix the guide cannula in place, cyanoacrylate glue was used and once cured cannula was further fixed using dental acrylic cement (Paladur, Kulzer). Immediately after surgery, postoperative analgesia was performed by i.p. injection of 0.1 mg kg −1 body weight Buprenorphine (Bupaq, Streuli Pharma AG). Afterward, 0.3 mg ml −1 Buprenorphin was provided via the drinking water for 3 days. Mice remained single-caged and were allowed to recover for at least 2 weeks after surgery. For refeeding studies, mice received 1 µl intracerebroventricular infusions of either saline, 0.4 µU human insulin (Actrapid, Novo Nordisk), or an equimolar amount of human proinsulin (R&D Systems) at 14:00. The mice remained fasted thereafter until food (pre-weighed on a precision balance (K2-20, G&G, Kaarst, Germany)) was once again presented at the onset of the dark cycle (18:00). Any remaining food was weighed 1, 4, 12, 24, and 48 h after refeeding. This was done in a cross-over fashion, although not all mice underwent all treatments as in some occasions guide cannulas were blocked and mice could not be injected.

Further software
The manuscript was prepared using Microsoft Office version 16.

Statistical analyses
Data are presented as mean ± SEM. Two groups were compared using the Mann-Whitney U test, three or more groups by one-way ANOVA with a Holm-Sidak post hoc test, and data comparing several timepoints by two-way ANOVA with a Holm-Sidak post hoc test. Statistical analyses were conducted with Prism v8.4 (GraphPad Software, San Diego, CA, USA).

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Human islet data were extracted from GSE50244. Source Data are provided with this paper.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.