RESULTS
104Transcriptomic diversity of hippocampal and entorhinal cells 105 To survey the transcriptomic diversity and functional specification of the mesial temporal 106 cortex, we used snRNA-seq to profile five subregions of the hippocampal-entorhinal system 107 6 collected from fresh frozen postmortem brains of clinically unremarkable human donors. These 108 specimens were selected from a larger pool of postmortem human brains based on the quality 109 of isolated nuclei and RNA. Taking into consideration dramatic cytoarchitectonic variations, 110 we microdissected the hippocampal formation (DG, CA2-4, CA1, and subiculum) and EC for 111 a total of five subregions (Fig. 1A). 112 Unbiased isolation of nuclei using our previously described protocol (Li et al., 2018; 113 Zhu et al., 2018) followed by snRNA barcoding, cDNA sequencing and quality filtering yielded 114 108,315 high-quality single-nucleus profiles from all five subregions ( Fig. 1A, S1A-D). 115 Analysis of the expression of genes known to be enriched in major cell subpopulations 116 suggested these included 44,697 neurons, of which 35,768 (80.02%) were glutamatergic 117 excitatory neurons (expressing the gene encoding the vesicular glutamate transporter SLC17A7) 118 and 8,929 (19.98%) were GABAergic inhibitory neurons (expressing the gene encoding the 119 GABA synthesis enzyme GAD1), reflecting the expected 80:20 ratio of these populations. In 120 addition, we identified 63,618 (58.73% of the total population) non-neuronal cells. 121 We next analyzed the transcriptomes of those nuclei on the Uniform Manifold 122 Approximation and Projection (UMAP) layout representing their similarities at cellular 123 granularity ( Fig. 1B-D). Iterative clustering defined 69 transcriptomically distinct cell clusters 124 representing presumptive cell types across all individuals (donors). These transcriptomically 125 diverse subpopulations were organized into a dendrogramatic taxonomy reflecting their gene 126 expression patterns and were subsequently assigned identities commensurate with predicted cell 127 types. This allowed us to identify 26 subtypes of excitatory neurons (Fig. 1E and S2A-B), 23 128 inhibitory neuron subtypes ( Fig. 1E and S2C-D), and 20 non-neuronal cell types and subtypes 129 ( Fig. 1E and S2E-F). Similar single nucleus approaches applied to human neocortical samples 130 7 yielded comparable numbers and distributions of cell populations in medial temporal gyrus 131 (MTG) (Hodge et al., 2019)and dorso-lateral prefrontal cortex (dlPFC) (Li et al., 2018)(Fig. 132
S1E-F).
133Within excitatory neuron subtypes, we found marked transcriptional diversity that 134 reflects differences in the cytoarchitectonic organization among the subregions of the HIP (DG, 135 CA2-4, CA1, and subiculum) and EC (Fig. 1E). For example, in addition to ADCYAP1-136 expressing mossy cells in DG, we found three distinct subclusters of PROX1-expressing granule 137 cells. We also identified excitatory neurons in CA1 and CA2-4 that could be deconstructe...