a) Reproducibility of detected genes was evaluated by comparing all possible pairwise comparisons within replicates of each group (control RNA extraction and 120, 50, 30, 10, 5, 2 and 1 LCM-seq samples), shown as mean with 90% confidence interval of reproducible ratios. LCM-seq improved reproducibility of detection of low and medium level expressed genes compared to the RNA extraction protocol (RNA extraction) (mean±s.e.m., n4). ( b) Standard deviation of gene expression within replicates binned according to gene expression levels. LCM-seq samples of 120–30 cells showed reduced technical variation for low and medium level expressed genes compared with the RNA extraction protocol (RNA extraction) (mean±s.e.m., n4). ( c) Mean number of genes detected in the different sample groups (mean±s.e.m.). A larger number of genes were detected in the 120 cell samples subjected to LCM-seq compared to the group subjected to RNA extraction before sequencing (0.1 RPKM as cutoff, P=0.04, Student's t-test). ( d) Gene expression correlation was high for 120, 50, 30 and 10 cell samples, while 5, 2 and 1 cell samples showed the heterogeneity among spinal MNs (Spearman’s correlation, genes expressed 1 RPKM in at least one sample were used). ( e) PCA for all groups based on top 500 variable genes in expression confirmed the high similarity between 120 and the RNA extraction group as well as with 50, 30 and 10 cell LCM-seq samples.