Fig. 2: Proteomics and single-cell transcriptomics identify novel disease signatures of Scz patient organoids.

a Schematic of proteome mapping pipeline using tandem mass tag (TMT) chemistry and liquid chromatography-mass spectrometry (LC-MS). Briefly, whole organoids were subjected to tryptic digestion to isolate proteins from each sample. Proteins were next isobarically barcoded with TMT reagents, and then condensed into a single multiplexed suspension. This allowed samples to be run concurrently, eliminating batch-specific technical variance, and a low-noise assessment of both organoid variability (using a 4x2 design) and the identification of translated disease factors. Multiplexed suspensions were subsequently subjected to quantitative TMT-LC-MS for peptide detection, deconvolution, and quantification, and then computationally analyzed. b Scz organoids recapitulate the developmental proteome. Given numerous cellular signatures of Scz that separated cases from Ctrls (Fig. 1), we sought to determine the proteome diversity of Ctrl and Scz 3D organoids. The Scz organoid proteome was 99.95% identical in pooled protein diversity relative to Ctrls. Thus, Scz proteome differences are likely to be explained not in the induction of differential factors at the posttranslational level, but rather their total molecular quantity. c Individual proteins exhibit low variability in expression between groups. Given the high degree of conservation in peptide diversity between both Ctrl and Scz organoids, we next examined dispersion of individual proteins by generating their coefficient of variation (CVs) for each group. In both Ctrl and Scz samples, over 3000 detectable peptides for each group exhibited low variability (<20% CVs; blue dots). Only 3 and 10 proteins exhibited “high” variability in Ctrls and Scz organoids, respectively (red dots). In these panels, the black lines/box segregate proteins with median expression differences within groups. This reflects that Ctrl and Scz organoids exhibited similar patterns of proteome reproducibility in this particular analysis. d Subtle proteome differences in Scz patient-derived organoids. A heat-map of differentially expressed proteins shows differences in LC-MS expression intensities in Ctrl versus Scz organoids. Specifically, Scz organoids are defined by the differential expression of just 222 peptides, or ~5.9% of the detected proteome (heat-map). This heat-map thus provides a visualization of the individual peptide factors identified via LC-MS, and the consistency of expression differences at the single-protein level. Thus, consistent with our hypothesis, Scz organoid samples principally differed in their quantity, rather than diversity, of developmental factors. e Disease factors are detected and differentially expressed in Scz organoids. A heat-map of selected differentially expressed proteins in Scz organoids (Log2 key from panel d still applies). This included POU-domain peptide fragments that putatively mapped to the forebrain neuronal-development factor POU3F2 (BRN2), and disease factors with known and/or likely involvement in disease pathophysiology (e.g., COMT/PLCL1). In addition, factors with putative genetic risk but otherwise unestablished disease biology (e.g., PTN) were also differentially expressed in Scz organoids. f Schematic of pipeline for live single-cell RNA sequencing of organoids and clustering of 26,335 transcriptomes recapitulates fetal brain cellular identities. Briefly, organoids from 4 Ctrl and 3 Scz lines were generated concurrently, pooled by-line, and dissociated to a single-cell suspension. We rapidly conducted survival-based high-throughput FACS to purify samples to 2000 live cells/ µm per line. For robustness, post-FACS live cell viability was also cross-confirmed using Countess-II. Live cell suspensions were next rapidly loaded into 10x chromium microfluidic devices to produce barcoded single-cell nanodroplet emulsions. This emulsion was later broken, barcoded samples were amplified, libraries prepared, subjected to Illumina sequencing, and then parsed through a suite of unbiased computational analyses. Clustering of marker genes for cell-type specific clusters was conducted via pairwise comparisons of the normalized expression values for cells of a given cluster vs. the cells of all other clusters. Thus yielded unbiased gene sets, which were defined by the top 10 gene markers for each cluster that met a high-pass FDR threshold of 1% and >15,000 total read counts. Many of these prototypic markers defined cell-types consistent with human fetal tissue (see below). Here we present UMAP coordinates for 26,335 transcriptomes split by Ctrl and Scz cases, presenting the cell-type clusters identified in our unbiased clustering analysis. Cell-type labels were determined via a variety of approaches including marker gene-expression, automated annotation, and, namely, comparison with human fetal samples (see Methods for analysis pipeline). Bar chart (right) depicts cell-type proportions, illustrating that ~93% of Ctrl scRNA-Seq transcriptomes were identified as neural progenitors, proliferating cells, or terminal cortical cell-types (e.g., neurons and glia). Compared to this, only ~75% of Scz cell-types exhibited a similar conservation of identity. Compared to remaining cell types in Ctrl organoids (~7% cells), the remaining ~25% of Scz scRNA-Seq transcriptomes reflected enrichment for brain-related cell-types including putative neuroendothelial cells, structural markers, developing vasculature, retinal, and choroid plexus markers in Scz organoids. This analysis therefore revealed the cell-types produced at the expense of neurons in Scz organoids alluded to in pulse-chase experimentation depicted in Fig. 1e. Of note, all cell-types exhibited reproducible proportions across individual iPSC donors within respective groups (i.e., all Scz organoids exhibited similarly reproducible alterations in cell-type diversity, which was defined by an overarching loss of neurons). g Confirmation of progenitor and neuronal depletion in Scz organoids. Scz organoids exhibited a striking depletion of progenitors (SOX2+ and PAX6+) and pan-neuronal markers (MAP2+, DCX+, and STMN2+). These expression differences reflect both abundance and magnitude. Thus, replicating our prior results (see Fig. 1), scRNA-Seq analysis confirmed that progenitors and neurons are depleted in Scz organoids. Representative UMAPs for SOX2 and MAP2 are provided given that these are the same markers shown for progenitor and neuronal depletion in Fig. 1c, d. For Fig. 2f–j, total n = 26, 335 transcriptomes, n = 20,844 genes from 7 iPSC lines; Ctrl n = 15,089 transcriptomes from 4 Ctrl iPSC lines, and Scz n = 11,246 transcriptomes from 3 Scz iPSC lines. Ctrl: Control, Scz: Schizophrenia.