SUPPLEMENTARY DATA ©2018 Ame©2018 SSC-H SSC-A SSC-A Murine GLU and GCG+ (D) Human from GLU Murine Neurod1 of Sort FACS (E,F,G) Emission

SUPPLEMENTARY DATA ©2018 Ame©2018 SSC-H SSC-A SSC-A murine GLU and GCG+ (D) human from GLU murine NeuroD1 of sort FACS (E,F,G) emission. 660/ excitation, 633nm by detected (FSC scatter forward (SSC), scatter 1 Figure Supplementary Cells E A I Cells FSC-A FSC-A FSC-H - Cre/ROSA26 rican Diabetesrican Association.Published online at http://diabetes.diabetesjournals.org/lo Cells - F cls sn gtn a lble. DHL Aiet Bioa Agilent (D,H,L) labelled. as gating using cells YFP

FSC-W GCG GCG+, human of sorting FACS (A,B,C) . FSC-W FSC-W B J Singlets F Singlets FSC-A FSC-A FSC-A Singlets -

) n ple it (FSC width pulse and A) 20nm emission; GLP emission; 20nm - Venus (H) and NeuroD1cell (L) (H) and populations. Venus - eu cls sn gtn a lble. IJK FC sr o murine of sort FACS (I,J,K) labelled. as gating using cells Venus

561 582/15-A 710/50-A 710/50-A 2 3 4 5 101 102 103 104 101 102 103 104 10 10 10 10 G K C 10 2 633 660/20-A 10 - 10 10 1 staining was detected by 561nm excitation, 582/15nm 582/15nm excitation, 561nm by detected was staining 1 Negative 1 1 10 10 YFP YFP 3 10 - W) to collect single cells. CHGA/SCG2 staining was was staining CHGA/SCG2 cells. single collect to W) GLU-Venus 2 2 10 10 4 ND1 10 GCG+ GCG- 3 3 10 10 -

and negative cells. Cells were sorted by side side by sorted were Cells cells. negative and 5 4 4 nalyzer 2100 trace of RNA quality quality RNA of trace 2100 nalyzer H D L okup/suppl/doi:10.2337/db18 25200500 100020004000 (nt) 25200500100020004000(nt) 25200500100020004000(nt)

- 0883 / - /DC1

H SUPPLEMENTARY DATA

Supplementary Figure 2. Comparison between EECs from human jejunum and ileum

(a) Principal component analysis (PCA) plot of matched human jejunum vs ileum samples, labelled by cell population, anatomical location and patient (n=3 cell populations from both anatomical regions from each of 2 participants) of the 500 top variable genes when using a DESEQ2 model. PCA separated EECs from negative cells on the first component and the anatomical location of GCG+ and GCG- cells on the second component. (b) PCA plot for the same samples as in (a), but excluding the non- enteroendocrine (Neg) samples from the DESEQ2 model and the PCA analysis. PCA showed separation of ileal from jejunal samples on the first component and GCG+ from GCG- on the second component. Samples from the different participants are depicted by different sized symbols.

(c,d) Enrichment vs expression plots for GCG+ (c), and GCG- (d), cell populations. Enrichment is presented as the log2 fold difference between the cell populations indicated, and expression is presented as the log2 base mean normalised expression extracted from the DESEQ2 model. Red – enriched (adjusted P<=.1 in DESEQ2 model) in jejunum; Green – enriched in ileum.

GIP, CCK, SST, MLN, SCT, GHRH, ASIC5, and TRPA1 were notably higher in jejunal than ileal EECs (GCG+ or GCG-), whereas GCG, NTS and TAC1 were higher in ileum than jejunum.

©2018 American Diabetes Association. Published online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db18-0883/-/DC1 SUPPLEMENTARY DATA Supplementary Figure 3. Differential vs mean expression plots for human jejunum EEC populations. Differential expression is presented as the log(2) fold difference between the cell populations indicated, and expression is presented as the log(2) base mean normalised expression extracted from the DESEQ2 model. (A,D) GCG+ vs negative, (B,E) GCG- vs negative (C,F), GCG+ vs GCG-. Transcription factors are shown in A,B,C; ion channels are shown in D,E,F. Red – enriched (i.e. adjusted P<0.1) in GCG+; Blue – enriched in GCG-; Green – enriched in negative cells. Key genes are labelled

A D 10 ASIC5 ISL1 INSM1 RFX6 10 PAX6 CACNA1A 8 NKX2-2ARX NEUROD1 BEX1 FEV KCNG3 KCNK3 MYT1 ONECUT3 8 KCNB2 e e KCNH7 KCNJ6 c c HCN4 6 FOXO6 CITED4 ETV1 CACNA1C n EGR4 FOXA2 n 6 TRPA1 e MAK ZNF483 ARNT2 GCG+ e

r DDN r FOXJ1 PTPRN TSHZ2 ST18 GCG+ e 4 RFX2 e KCNJ11 f LMX1B APBB1 L3MBTL4 f f f 4 i DACT2 i SCN8A CACNA1H DUSP26 HLF PLAGL1 SCN3A d d 2 CACNA1D d d

l l 2 o o

f HHEX f 0 2 2 0 g g o o

L -2 L -2 Negative Negative -4 -4

-6 -6 0 5 10 0 5 10 Log2 base mean expression Log2 base mean expression B E 10 FEV ONECUT3 NKX2-2 NEUROD1 10 ISL1 INSM1 8 BEX1 RFX6 CACNA1A EGR4 PAX4 MYT1 KCNB2 ARX PAX6 8 ASIC5 KCNK3 e ETV1 e SCN3A DACT2 FOXA2 TRPC7 KCNK16 c NEUROG3 PTPRN c KCNG3 6 FOXQ1 KCNK17 KCNJ6 CACNA1C n FOXO6 n KCNH7 AGT NPAS4 HHEX SMAD9ARNT2 6 TRPA1 CACNA1H e GCG- e

r ACTL6B CITED4 ETV4RFX2 ST18 r DLX5 DLL1 SCN8A GCG- e 4 LMX1B CUX2 MNX1 e f f

f DUSP26 APBB1 WFS1 f

i i 4 KCNJ11 HR CACNA1D d TSHZ2 NR4A3 d 2 d d

l l 2 o o f f 0 PLAGL1 2 2 0 g g o o

L -2 L -2 Negative Negative -4 -4

-6 -6 0 5 10 0 5 10 Log2 base mean expression Log2 base mean expression C F 10 10 8 8 e e

c 6 c n n 6

e GCG+ e r r GCG+

e 4 e f f

f f 4 i L3MBTL4 i GABRA2 TRIM29MAK TCF4 GPX3 d HLF PLAGL1 d ASIC5 MYCN POU5F1 2 FLI1 KIT TSHZ2 d ISL1 PAX6 d l l 2 ARX INSM1 HSPA1B

o RFX6 o CACNA1A

f DDX5 f 0 BEX1 CACNA1C 2 NR4A1 2 0

g g TRPA1 FEV NEUROD1 o ATOH8 DACT2FOXQ1 HR ONECUT3 o CACNA1D L -2 CUX2 NR4A3 L -2 CACNA1H GATA2 PRKCHMNX1 NEUROG3 PAX4 GCG- GRIA4 SCN3A GCG- HHEX -4 -4 TRPC7

-6 -6 0 5 10 0 5 10 Log2 base mean expression Log2 base mean expression

Peptidomics is a relatively new and under-developed field within proteomics, despite the varied and significant roles of peptides as signalling molecules in many organisms. Peptides, due to their small size in comparison to proteins, have specific properties that can be used to extract and analyse them intact by mass-spectrometry, giving information about their exact processing and post-translational modifications. Several methods to extract peptides from different tissues or fluids have been published, but none was developed to study intestinal tissues and the large variety of hormone peptides they produce.

Peptide extraction and purification for LC-MS/MS studies initially requires the complete disruption of the tissue of interest while inhibiting all protease activity. This is to avoid contamination with degradation products of larger proteins as well as degradation of the peptides of interest. The low- molecular weight proteins and peptides must then be separated from the rest of the proteins and other molecules. We compared several methods adapted from the literature to develop a robust and reliable method to quantify peptide levels from intestinal tissues. Methods tested for tissue lysis and protein extraction included tissue homogenization in 0.25% acetic acid, 8M urea, 90% methanol with 1% acetic acid, guanidine hydrochloride or 80% acetonitrile, or using Trizol to extract both proteins and peptides from the same sample1–4 Peptides were separated from proteins using either 10kDa molecular weight cut-off filters or 80% acetonitrile. Samples were finally purified using SPE columns, dried and reduced alkylated before being analysed by mass-spectrometry.

The different methods allowed the detection of many unique peptides by LC-MS/MS with an important variability between the different methods, from 1778 unique peptides for the extraction using 8M urea to 25 with 90% methanol/1% acetic acid. Many corresponded to fragments of longer peptides or proteins and could represent degradation products, coming either from the protein turn- over of the tissue or degradation happening during sample preparation. However, most of the peptides were unique to the extraction method (SupFig 4a), due to their different chemical properties and also due to different degradation levels. Focusing on the analysis of the classical gut peptide hormones significantly reduced the number of hits, and clearly showed that most of the expected intact peptides were detected using the guanidine HCl extraction method (SupFig 4b). In addition to improving the number of detected hormone peptides, the guanidine HCl extraction method showed the highest levels of these peptides when quantified (SupFig 4c). Urea and Trizol extraction were the second best methods for peptide levels and variety of peptide hormones extracted, and tissue homogenization using Trizol can be a reliable technique to extract both RNA and peptides from a same sample. We also compared peptide extraction from the proteins using either protein precipitation with 80% ACN or using 10kDa filters, the latter resulting in the loss of many peptides and decreased levels (Supp Fig 4c).

©2018 American Diabetes Association. Published online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db18-0883/-/DC1 SUPPLEMENTARY DATA We then confirmed the reproducibility of the guanidine HCl +80% ACN extraction method, as the %CV from 5 replicates repeated twice from different mice was around 20% for most of the peptides, with higher %CV for peptides detected with lower abundance (Supp Fig 4d), and we could also show that peptide quantification was linear with the amount of starting material in a range from 2 to 40 mg (Supp Fig 4e), indicating that tissue weight can be a satisfactory normalisation factor to compare different samples.

Tissue homogenization in guanidine HCl followed by protein precipitation with 80% acetonitrile is therefore a promising method to extract peptides from intestinal tissue for mass-spectrometry analysis, allowing the detection a large range of peptides and their robust quantification.

Homogenization optimization

Small pieces of mouse jejunal mucosa of similar size (~40mg) were homogenized in different conditions in duplicates with lysing Matrix D using a FastPrep-24 for 4 x 40s at 6ms-1, and peptide extracted before being dried on a centrifugal concentrator. Samples were resuspended in 500L 0.1% aqueous formic acid and purified by solid-phase extraction using a Waters HLB μElution solid‐phase extraction (SPE) plate as described. Samples were analysed using high-flow rate based LC-MS method as described before.

For the acetic acid, guanidine HCl and urea extractions, tissues were directly incubated in 300L of the lysing solution (0.25% acetic acid aqueous, guanidine HCl 6M or urea 8M) and homogenized. 200L of the homogenate was collected into a new tube after centrifugation to remove tissue debris and lysing matrix (5min, 2000g, 4˚C). 80% acetonitrile (800L) was added to precipitate the proteins and peptides in solution were collected following centrifugation (10min, 12,000g, 4˚C). For the guanidine HCl homogenates, addition of 80% acetonitrile induced a separation between an organic phase (top) and aqueous phase (bottom) in addition to the protein precipitation, with the peptides in the aqueous phase (data not shown). Peptides from guanidine HCl homogenization were also extracted using a Vivaspin 500 MCWO 10kDa (GE-healthcare) instead of using 80% ACN, collecting the flow-through after a 30 minutes spin at 16,000g at 4˚C. Flow-through was then acidified adding 10% formic acid solution and extracted by SPE.

For the 80% ACN and methanol/acetic acid extraction, tissues were directly homogenized in 80% acetonitrile aqueous or 90% methanol 1% aqueous acetic acid, 200L of homogenate was recovered from the lysing D matrix and spun at 12,000g at 4˚C for 10 minutes to separate the precipitated proteins. Peptides in supernatant were transferred to a new tube and dried before SPE extraction as previously.

©2018 American Diabetes Association. Published online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db18-0883/-/DC1 SUPPLEMENTARY DATA For the Trizol extraction, tissue was homogenized in 500L Trizol and the homogenates transferred

to a new tube. 100L chloroform was added and mixed vigorously and sample centrifuged after 10min incubation at RT (15min, 12,000g, 4˚C). The upper phase (aqueous) was then transferred to a new tube (from which RNA can be extracted using standard methods), 800uL 80% ACN was added to precipitate any remaining proteins and sample centrifuged (10min, 12,000g, 4˚C) and supernatant dried before SPE extraction. The lower (phenol) phase was treated with 100% ethanol (1:3 ethanol:phenol phase by volume) to precipitate DNA and the mixture centrifuged (10min , 2,000g, 4˚C) then transferred to a new tube. Proteins and peptides were precipitated by addition of 750L cold acetone and centrifugation at 7500g for 5min at 4˚C. Pellets were washed once with 1mL 0.3M guanidine HCl in ethanol and air dried before being resuspended in 200uL 8M urea. Finally, proteins were precipitated with 80% ACN, centrifuged (10min, 12,000g, 4˚C) and the supernatant (containing the peptides) was transferred to a fresh tube, dried and SPE extracted as previously.

1. Basak, O. et al. Induced Quiescence of Lgr5+ Stem Cells in Intestinal Organoids Enables Differentiation of Hormone-Producing Enteroendocrine Cells. Cell Stem Cell 0, R106-188 (2016).

2. Van Camp, K. A., Baggerman, G., Blust, R. & Husson, S. J. Peptidomics of the zebrafish Danio rerio: In search for neuropeptides. J. Proteomics 150, 290–296 (2017).

3. Van Dijck, A. et al. Comparison of extraction methods for peptidomics analysis of mouse brain tissue. J. Neurosci. Methods 197, 231–237 (2011).

4. Ziros, P. G., Chartoumpekis, D. V. & Sykiotis, G. P. A Simple Protocol for High Efficiency Protein Isolation After RNA Isolation from Mouse Thyroid and Other Very Small Tissue Samples. in 383– 393 (Humana Press, New York, NY, 2016). doi:10.1007/978-1-4939-3756-1_25

a and b: Venn diagrams showing the total number of unique peptides (a) and the number of unique peptides matching classical hormone peptides (b) found in the different methods for peptide extraction. c: Quantification of different gut hormone peptide using different extraction methods, normalised by tissue weight. Data is represented as mean +-sd from 2 replicates. d: Quantification variability for different peptides measured from 10 homogenates from 2 different mice. Data is the mean value +- sd and values indicated above are the %CV. e: Raw peak area of peptide quantification for different samples spanning different weights from 2 to 40 mg plotted against the tissue weight. Linear regression is plotted, showing a good correlation between tissue weight and peptide quantification.

SUPPLEMENTARY DATA

Supplementary Table 1. Demographics of human participants from transcriptomic study. RYGB = Roux-en-Y Gastric Bypass; donor = organ transplant donor; TG = Total gastrectomy for prevention / treatment of gastric cancer

ID Source Age Gender BMI

R3 France – RYGB N/A N/A >40

R4 France – RYGB 24 F 47

R5 France – RYGB 31 F 43

R6 France – RYGB 46 M 41

D1 France – donor 27 M 24

D2 France – donor 26 F 21

HGH006 UK – TG 57 M 34

HGH007 UK – TG 22 M 29

HGH008 UK – TG 23 M 28

HGH010 UK – TG 64 M 27

HGH018 UK – TG 61 M 37