bioRxiv preprint doi: https://doi.org/10.1101/261529; this version posted February 7, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license. Spatial reconstruction of single enterocytes uncovers broad zonation along the intestinal villus axis Andreas E. Moor 1, Yotam Harnik 1, Shani Ben-Moshe 1, Efi E. Massasa 1, Keren Bahar Halpern 1 and Shalev Itzkovitz 1 + 1 Department of Molecular Cell Biology, Weizmann Institute of Science, Rehovot, Israel. +To whom correspondence should be addressed. E-mail: [email protected] The intestinal epithelium is a highly structured tissue composed of we established a comprehensive panel of landmark genes, characterized repeating crypt-villus units1,2. Enterocytes, which constitute the most by RNAseq of laser-capture-microdissected epithelial samples originating abundant cell type, perform the diverse tasks of absorbing a wide range from differential villus zones. We used these to reconstruct the spatial of nutrients while protecting the body from the harsh bacterial-rich tissue coordinates of enterocytes in scRNAseq data and uncovered vast environment. It is unknown if these tasks are equally performed by all enterocytes or whether they are spatially zonated along the villus axis3. heterogeneity and sub-specialization along the villus axis. This spatial Here, we performed whole-transcriptome measurements of laser-cap- division of labor of mature enterocytes could be optimal for achieving the ture-microdissected villus segments to extract a large panel of landmark Figure 1 a genes, expressed in a zonated manner. We used these genes to localize Generalist absorption single sequenced enterocytes along the villus axis, thus reconstructing top cells B a global spatial expression map. We found that most enterocyte genes Specialized absorption middle cells were zonated. Enterocytes at villi bottoms expressed an anti-bacterial tissue dissociation single cell RNA sequencing Stable differentiation Reg gene program in a microbiome-dependent manner, potentially re- 5 bottom cells small intestinal villus ducing the crypt pathogen exposure. Translation, splicing and respiration 4 3 genes steadily decreased in expression towards the villi tops, whereas C spatial tissue villusDynamic epithelium subtypes distinct mid-top villus zones sub-specialized in the absorption of carbo- 2 reconstruction 1 Amino-acid rich meal 15 1 5 hydrates, peptides and fat. Enterocytes at the villi tips exhibited a unique Carbohydrate rich meal laser capture & landmark genes gene-expression signature consisting of Klf4, Egfr, Neat1, Malat1, cell bulk RNA sequencing adhesion and purine metabolism genes. Our study exposes broad spatial b c d heterogeneity of enterocytes, which could be important for achieving bottom landmark genes top landmark genes 1 1 2010107E04Rik 2010109I03Rik 2210407C18Rik Atp5a1 Acad9 Atp5e Ada their diverse tasks. Atp5g3 Ap1g1 Atp5h Atp5j 0.9 Apoa4 0.9 Atp5j2 Atp5l Apoc3 Atp5o Cep57 Atpif1 Cfap20 Ccl25 Chchd10 0.8 Chchd1 0.8 Ckmt1 Chek2 Cox4i1 Cox5a Clca4a Cox5b Cldn7 Cox6b1 Fgd4 Cox6c 0.7 0.7 Main Text Gcn1l1 Cox7a1 Cox7a2 Glipr1 Cox7b Fabp1 Gm10680 5 Gpx2 Gm20594 Gsta1 0.6 0.6 villus top Lgals4 Ifrd1 Lypd8 Krt20 4 Minos1 The intestinal tract is responsible for nutrient digestion and absorption, Ndufa1 Lgals3 3 Ndufa4 Lrrc41 Ndufa5 Ndufb6 0.5 Mrpl48 0.5 Genes 2 Genes Ndufb8 Myo7a Ndufc1 Olfr1385 secretion of mucus and hormones, interactions with commensal micro- Plac8 Reg3b Olfr46 1 villus bottom Reg3g Pam Rpl18 0.4 0.4 Rpl35a Pkib 1,2 Rpl38 Pmp22 crypt Rpl39 biota and protection of the organism from pathogenic microbes . This Rpl41 Psma7 Rplp1 Rab34 Rps12 0.3 0.3 Rps14 S100a10 Rps18 S100a6 Rps2 Serpinb1a wide array of tasks requires the presence of different cell types that are Rps27 Rps27l Slc17a5 Rps28 0.2 Slc25a22 0.2 Rps29 Rps8 Slc28a2 Sis Sprr2a2 specialized for their respective functions. Enterocytes, which represent Spink4 Ssbp2 Tm4sf5 Tma7 Tbk1 0.1 Txn1 0.1 Uba52 Tlr1 Uqcr10 Tmsb4x Uqcr11 Ythdc2 the majority of cells in the epithelial layer, constantly migrate along Uqcrh Uqcrq Zfp280d 1 2 3 4 5 1 2 3 4 5 the villi walls until they are shed off from their tips 3-5 days after their Zones Zones emergence from crypts. The positions of enterocytes along the villus axis e f Mki67 Alpi 3 1 4 correlate with their age , exposure to morphogen gradients and hypoxia , Spatial signature yet the positional effects on enterocyte function are largely unknown. Pre- 20 20 tSNE 2 0 tSNE 2 0 tSNE_2 vious work investigated transcriptomic changes along the small intestinal tSNE_2 20 crypt-villus axis with bulk samples and DNA microarray-based expres- −20 −20 Villus zone 5,6 7 −20 0 20 −20 0 20 6 5 sion profiles in mouse and human tissue . This body of work revealed tSNE_1tSNE 1 tSNE_1tSNE 1 4 0 3 tSNE 2 2 Muc2 Cck 1 some broad compositional differences of the crypt and the villus, yet its Crypt low spatial resolution (comparing bulk crypts to bulk villi), uncontrolled 20 20 −20 mixes of different cell types and the low sensitivity of microarray-based 0 0 tSNE 2 tSNE 2 tSNE_2 tSNE_2 transcriptomics precluded the detection of spatial expression changes and −20 −20 −20 −10 0 10 20 tSNE 1 heterogeneity of enterocytes along the villus. −20 0 20 −20 0 20 tSNE_1tSNE 1 tSNE_1tSNE 1 Single-cell RNA sequencing (scRNAseq) has revolutionized our ability to Figure 1: Spatial reconstruction of villus enterocytes. characterize individual cells in-depth8; it was recently utilized in the intes- a, Scheme of the experimental approach. The villus epithelium is dissociated into single cells; these cells are profiled by scRNAseq. In parallel, spatial landmark genes 9 10 tine to identify cell types and sub-populations of intestinal stem cells , are retrieved by bulk RNAseq of villus quintiles obtained using laser capture micro- tuft cells11,12 and enteroendocrine cells9,12–14. However, spatial heterogene- dissection (LCM). The original position of the sequenced single cells is then inferred ity and specialization along the villus axis within the enterocytes, the larg- based on their expression levels of the landmark genes. b, Laser capture microdis- section of villus epithelium quintiles. Scale bar 50µm. c, LCM-RNAseq expression of est cell compartment, has not been addressed. Relating such heterogeneity the villus-bottom landmark genes. d, LCM-RNAseq expression of the villus-top land- to tissue coordinate is challenging, as the spatial origin of individual cells mark genes. Profiles in c-d are normalized to the maximum expression level for each is lost when the tissue is dissociated for scRNAseq. We and others have gene. e, tSNE plots of intestinal marker gene expression in single Lgr5-eGFP negative cells13. Mki67 is expressed in transient amplifying cells in the crypt, Alpi in villus en- developed approaches to spatially reconstruct scRNAseq data by making terocytes, Muc2 in goblet cells and Cck in enteroendocrine cells. Depicted analyses use of known expression profiles of landmark genes characterized by RNA are based on raw data from NCBI GEO datasets GSM2644349 and GSM264435013. in-situ hybridization15–18 (Fig. 1a). This approach is infeasible, however, f, tSNE plot of the enterocyte and progenitor populations in the sorted Lgr5 negative when no prior knowledge exists regarding zonated landmark genes. Here dataset. Each cell is colored according to its inferred villus zone. Moor et al. | biorXiv | February 8th, 2018 | 1 bioRxiv preprint doi: https://doi.org/10.1101/261529; this version posted February 7, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license. diverse tasks of this tissue3. on the ratio between the summed expression of the top and bottom land- mark genes. For each cell, x correlated with its position along the villus To extract a panel of enterocyte landmark genes, we used laser capture axis (Methods, Supp. Fig. 2). By computing the x values of the five laser microdissection (LCM) to isolate epithelial cells from five equally spaced captured areas we were able to assign each cell to one of 6 zones from the compartments between the bottom and tops of villi in the mouse Jeju- bottom to the top of the villus (Fig. 1f). We averaged, for every gene, the num (Fig. 1a,b). Bulk RNA sequencing (RNAseq) of these isolated villus expression of single cells in each of these zones, to obtain a comprehensive quintiles revealed genes with decreasing (Fig. 1c) and increasing (Fig. 1d) spatial map of gene expression along the intestinal villus (Fig. 2a,b). expression gradients. We defined a set of 62 villus-bottom landmark genes and 43 villus-top landmark genes to be used for spatial reconstruction of Our spatial reconstruction included more than 9,832 enterocyte expressed scRNAseq data (Supp. Fig. 1). genes, 8,126 of which (83%) were significantly zonated (Methods, q-val- ue < 0.05). Thus, differentiated enterocytes exhibit ubiquitous spatial We used our LCM-RNAseq reference to identify a scRNAseq dataset13, heterogeneity with only a small minority of genes invariably expressed which included enterocytes that spanned the entire villus axis (Supp. Fig. from the bottom to the top of the villi. We used single molecule Fluores- 2). Mature and progenitor enterocytes were clearly demarcated by the cence in-situ Hybridzation17 (smFISH) to validate our predicted zonated expression of Alpi and Mki67 respectively (Fig. 1e). We assigned each se- expression profiles for 15 enterocyte genes, demonstrating the accuracy of quenced mature enterocyte a unit-less spatial coordinate x that was based reconstruction (Fig.