Comment on "Single cell RNA-seq reveals new types of human blood dendritic cells, monocytes and progenitors"

Abstract
Single-cell RNA-sequencing is a challenging technology to generate insight into individual cell phenotypes and conditions. However, technological limitations have to be considered to avoid misleading interpretation. Here we demonstrate how existing microarray transcriptomes of defined cell populations can help to correctly classify such new data and to distinguish cell- cell interaction, transitional or activation stages from new cell phenotypes.

Technical Comment
Villani et al. describe new subtypes of monocytes and dendritic cells (DC) based on single-cell RNA-seq and thereby propose the existence of 4 monocyte and 6 DC subsets in the blood (1). This profound analysis provides insight into the complexity of mononuclear phagocyte but also raises questions when compared with existing knowledge. Still in its infancy, limitations of this cutting edge technology as well as the biological function of phagocytes in immunity have to be considered for interpretation.
Interestingly, most cell profiles cluster within stable and discriminative patterns. Only few mismatches between the supposed gated population and the transcriptional phenotype may occur (Villani et al., table S13). However, compared to population profiling, there is extreme variability or failure of gene specific transcript detection per cell even for highly expressed and phenotype specific genes (CD14-transcripts in classical monocytes: 4.68 to 13,090.83 TPM/cell). Of ≈5326 different genes per cell, only ≈150 genes reveal ≥1‰ of transcripts per cell and only ≈ 60% are identified in common between two cells of the same population, thus demanding for pattern analysis and matching with existing reference transcriptomes. A challenge of flow cytometry is reliable collection of single cells. This limitation is explained by orientation of doublets when passing the laser beam (2). Antigen presentation or scavenging of apoptotic or necrotic cells predispose monocytes for cell-cell interactions and aberrant transcript patterns.
Against this background, comparing transcriptional relatedness of the new phenotype patterns with microarray transcriptomes of neutrophils, B-, NK-, CD4+ and CD8+ T-cells (3-5), monocyte (6-8) or DC subsets (9-12) identifies strong overlap of monocyte type-4 markers with NK- and T-lymphocytes (figure 1).

Figure 1: Mapping of single cell RNA-seq data to reference transcriptomes of defined cell populations generated by microarray hybridisation: The published marker candidates for DC and monocyte subclasses (1) are ranked in each marker panel by AUC from top to bottom and grouped separately for the characterization of monocyte (417 markers) and DC populations (947 markers). Annotations are given on the right side of the figure. The heat maps on the left side are derived from the published single cell RNA-seq data (1) and on the right side from population based microarray transcriptomes ((1) GSE38351; (2) GSE58173; (3) GSE18565; (4) GSE16836; (5) GSE66936; (6) GSE35340; (7) GSE23618; (8) GSE37750; (9) E-TABM-34).

The leading candidates PRF1, GNLY and KLRC4 are about 240/90/16-, 200/50/9- and 300/350/3-fold higher expressed in NK/CD8+/CD4+ cells than monocytes, respectively. Many markers belong to the upper 5%-quantile of highest transcripts in CD4+, CD8+ and NK-cells and thus are preferred candidates of detection in monocyte-lymphocyte interactions. While statistical selection of CD8+ T-cell specific markers compared to monocytes identifies in the top 500 candidates 11% overlap with the 149 type-4 monocyte markers, additional restriction to the 5%-quantile of highest expression in lymphocytes increases the fraction to 28%. Cell interaction instead of a new phenotype is supported by markers so far specific for T- and NK-cells (TCRBV3S1, TRGC2, TRDC, NKG7, IL2RB, LCK and ZAP70). Furthermore, the leading monocyte specific transcript CD14 drops in average by ≈50% in type-4 (2449 TPM) compared to type-1 (4356 TPM) monocytes, corresponding to a 1:1 “dilution” of the monocyte transcriptome. Although many type-4 markers are highest expressed in NK-cells, patterns and frequency of NK-, CD8+ and CD4+ T- cell-specific transcripts in individual type-4 monocytes suggest that different lymphocyte interactions are possible.
Similarly, type-3 markers are leading transcripts in neutrophils. Besides a novel phenotype, doublet formation or phagocytosis of apoptotic neutrophils with incomplete degradation of transcript remnants are possible explanations (figure 1). Despite partially shared expression between phagocytes, the 132 type-3 markers are up to 110-fold higher in neutrophils than monocytes and restriction to the 5%-quantile of highest expression in neutrophils raises the overlap within the statistically top 500 neutrophil specific genes compared to monocytes from 10% to 27%.
The interpretation as cell-cell interactions is supported by the fact that the statistically leading lymphocyte or neutrophil specific markers within the top 5% quantile of highest expression compared to monocytes are able to identify type-4 and type-3 monocytes by clustering, respectively (figure 2A).

Figure 2: Pattern recognition in monocytes using population based transcriptome data: A) Independently selected genes dominant in CD4+, CD8+ or NK-cells compared to monocytes were tested against the 149 type-4 markers suggested by Villani et al. and applied to the single-cell RNA-seq data of monocytes. All marker sets identify the group of type-4 monocytes as presented by hierarchical clustering. Green squares identify the monocytes type-4 samples and the overlap of type-4 with lymphocyte population markers, which demonstrates the imprint with CD4+, CD8+ or NK cell specific genes. The rainbow heat map next to each cluster indicates microarray signal intensities of the individual genes in three transcriptomes of CD14+ monocyte compared to three transcriptomes of each, CD4+, CD8+ or CD56+ (NK) cells. Signal and corresponding colour intensity are indicated for each cluster. B) All genes differentially expressed between NK-cells and monocytes (top) or neutrophils and monocytes (bottom) were investigated for increased or decreased expression in monocytes type-4 (top) or type-3 (bottom) when compared to monocytes type-1. Genes are sorted by intensity of differential expression as indicated on the left side by fold increase. The green labels identify the type-3 and type-4 monocyte samples and the overlap with the type-3 and type-4 marker genes. The ratio between mean values of gene specific TPM in monocytes type-1 and type-4 (top) or type- 3 (bottom) are presented for each gene as a color scale on the right of each cluster ranging from blue (=0) to red (≥2). This indicates how the decreased expression of NK-cell or neutrophil specific transcripts in monocytes type-1 is gradually disappearing with decreasing magnitude of differential expression in the lymphocyte or neutrophil microarray reference transcriptomes. Inversely, with increasing dominance of monocyte specific transcripts, the populations of type-4 and type-3 monocytes reveal reduced expression. The mean values of the ratios for the 100 genes most differentially expressed at each side of the scale are indicated. C) A new pattern of activated monocytes (black frame) was identified when mapping single-cell RNA-seq data to transcriptomes of LPS, TNF and IFN stimulated monocyte. This activation is frequently independent from the type-3 and type-4 monocyte phenotypes but also coincides with few cell interaction states. This suggests biologically relevant cell interactions and not technical artifacts. Reference transcriptomes for stimulated monocytes were derived from GSE38351.

Furthermore, lymphocyte and neutrophil related patterns seem not to be restricted to the suggested markers but to correspond to the whole profile of differential expression (figure 2B). Interestingly, the overall pattern of type-3 and type-4 phenotypes seems to occur on the background of type-1 monocytes (figure 1).
In contrast to these patterns of cell-cell interactions, which contain genes highly specific for these few cells, we identified a set of markers, which are less exclusive and correspond to a defined pattern of monocyte activation. This pattern cumulated in few type-1 (n=18), type-2 (n=5), type-3 (n=1) and type-4 (n=2) monocytes and was identified by comparing the single cell transcriptomes with microarray profiles of monocytes stimulated with LPS, TNF and IFN (figure 2C).
Dendritic cell phenotypes DC1, DC2 and DC6 correspond to the established microarray profiles of BDCA3+, BDCA1+ (equal with CD1c+ and mDC) and plasmacytoid (CD123+) DC, respectively. However, the DC4 phenotype overlaps with non-classical monocytes as already described by the authors. Indeed, three independent sets of microarray transcriptomes match DC4 patterns to non-classical monocytes (figure 1). While exclusion of CD16+ cells seems to eliminate non-classical monocytes from the DC pool (13), the CD16+ DC described by Lindstedt et al. (E-TABM-34) (12) also present like non-classical monocytes (figure 1) and induced a critical revision of the classification (14, 15).
Taken together, existence of six DC and four monocyte subsets should be reconsidered. There is not enough evidence that type-3 and type-4 monocytes as well as DC4 are undoubtedly new entities. Although the current nomenclature considers 3 monocyte subsets in blood, CD14++CD16- (classical), CD14++CD16+ (intermediate) and CD14+CD16++ (non-classical), the described type-3 markers consist of typical neutrophil transcripts and do not overlap with microarray transcriptomes generated from intermediate monocytes. Nevertheless, the investigations by Villani et al. are very important. Association of predominantly intermediate and non-classical monocytes with lymphocytic or granulocytic patterns and identification of activation sheds new light on the fate of monocytes and their differentiation. The activation pattern we extracted, was not only restricted to single cells, suggesting gain of exceptional knowledge also from individual cell-cell interactions. This new sequencing technology illustrates the extent of biological divergence, which is so far unrecognized by population profiling. However, microarray transcriptomes of thoroughly defined cell types and conditions provide stable and characteristic signatures and are indispensable references for classification of transcription patterns. This extraordinary treasure of already existing knowledge in public repositories is so far insufficiently exploited and integrated in the analysis of these evolving sequencing data.


References

1. Villani AC, Satija R, Reynolds G, Sarkizova S, Shekhar K, Fletcher J, Griesbeck M, Butler A, Zheng S, Lazo S, Jardine L, Dixon D, Stephenson E, Nilsson E, Grundberg I, McDonald D, Filby A, Li W, De Jager PL, Rozenblatt-Rosen O, Lane AA, Haniffa M, Regev A, Hacohen N. Single-cell RNA-seq reveals new types of human blood dendritic cells, monocytes, and progenitors. Science (2017) PMID: 28428369
2. Kudernatsch RF, Letsch A, Stachelscheid H, Volk HD, Scheibenbogen C. Doublets pretending to be CD34+ T cells despite doublet exclusion. Cytometry A. (2013) PMID: 23281028
3. Smiljanovic B, Grün JR, Biesen R, Schulte-Wrede U, Baumgrass R, Stuhlmüller B, Maslinski W, Hiepe F, Burmester GR, Radbruch A, Häupl T, Grützkau A. The multifaceted balance of TNFα and type I/II interferon responses in SLE and RA: how monocytes manage the impact of cytokines. J Mol Med. (2012) PMID: 22610275
   
4. Ghannam K, Martinez-Gamboa L, Spengler L, Krause S, Smiljanovic B, Bonin M, Bhattarai S, Grützkau A, Burmester GR, Häupl T, Feist E. Upregulation of Immunoproteasome Subunits in Myositis Indicates Active Inflammation with Involvement of Antigen Presenting Cells, CD8 T-Cells and IFNγ. PLoS One (2014) PMID: 25098831
5. Stuhlmüller B, Mans K, Tandon N, Bonin-Andresen M, Smiljanovic B, Sörensen T, Schendel P, Martus P, Listing J, Detert J, Backhaus M, Neumann T, Winchester RJ, Burmester GR, Häupl T. Genomic stratification by expression of HLA-DRB4 alleles identifies differential innate and adaptive immune transcriptional patterns - A strategy to detect predictors of methotrexate response in early rheumatoid arthritis. Clin Immunol. (2016) PMID: 27570220
6. Ingersoll MA, Spanbroek R, Lottaz C, Gautier EL, Frankenberger M, Hoffmann R, Lang R, Haniffa M, Collin M, Tacke F, Habenicht AJ, Ziegler-Heitbrock L, Randolph GJ. Comparison of gene expression profiles between human and mouse monocyte subsets. Blood (2010) PMID: 19965649
7. Ancuta P, Liu KY, Misra V, Wacleche VS, Gosselin A, Zhou X, Gabuzda D. Transcriptional profiling reveals developmental relationship and distinct biological functions of CD16+ and CD16- monocyte subsets. BMC Genomics (2009) PMID: 19712453
8. Liu B, Dhanda A, Hirani S, Williams EL, Sen HN, Martinez Estrada F, Ling D, Thompson I, Casady M, Li Z, Si H, Tucker W, Wei L, Jawad S, Sura A, Dailey J, Hannes S, Chen P, Chien JL, Gordon S, Lee RW, Nussenblatt RB. CD14++CD16+ Monocytes Are Enriched by Glucocorticoid Treatment and Are Functionally Attenuated in Driving Effector T Cell Responses. J Immunol. (2015) PMID: 25911752
9. Hutter C, Kauer M, Simonitsch-Klupp I, Jug G, Schwentner R, Leitner J, Bock P, Steinberger P, Bauer W, Carlesso N, Minkov M, Gadner H, Stingl G, Kovar H, Kriehuber E. Notch is active in Langerhans cell histiocytosis and confers pathognomonic features on dendritic cells. Blood (2012) PMID: 23074278
10. Széles L, Póliska S, Nagy G, Szatmari I, Szanto A, Pap A, Lindstedt M, Santegoets SJ, Rühl R, Dezsö B, Nagy L. Research resource: transcriptome profiling of genes regulated by RXR and its permissive and nonpermissive partners in differentiating monocyte-derived dendritic cells. Mol Endocrinol. (2010) PMID: 20861222
11. Aung LL, Brooks A, Greenberg SA, Rosenberg ML, Dhib-Jalbut S, Balashov KE. Multiple sclerosis-linked and interferon-beta-regulated gene expression in plasmacytoid dendritic cells. J Neuroimmunol. (2012) PMID: 22688425
12. Lindstedt M, Lundberg K, Borrebaeck CA. Gene family clustering identifies functionally associated subsets of human in vivo blood and tonsillar dendritic cells. J Immunol. (2005) PMID: 16210585
13. See P, Dutertre CA, Chen J, Günther P, McGovern N, Irac SE, Gunawan M, Beyer M, Händler K, Duan K, Sumatoh HRB, Ruffin N, Jouve M, Gea-Mallorquí E, Hennekam RCM, Lim T, Yip CC, Wen M, Malleret B, Low I, Shadan NB, Fen CFS, Tay A, Lum J, Zolezzi F, Larbi A, Poidinger M, Chan JKY, Chen Q, Rénia L, Haniffa M, Benaroch P, Schlitzer A, Schultze JL, Newell EW, Ginhoux F. Mapping the human DC lineage through the integration of high-dimensional techniques. Science (2017) PMID: 28473638
14. Robbins SH, Walzer T, Dembélé D, Thibault C, Defays A, Bessou G, Xu H, Vivier E, Sellars M, Pierre P, Sharp FR, Chan S, Kastner P, Dalod M. Novel insights into the relationships between dendritic cell subsets in human and mouse revealed by genome-wide expression profiling. Genome Biol. (2008) PMID: 18218067
15. Worah K, Mathan TS, Vu Manh TP, Keerthikumar S, Schreibelt G, Tel J, Duiveman-de Boer T, Sköld AE, van Spriel AB, de Vries IJ, Huynen MA, Wessels HJ, Gloerich J, Dalod M, Lasonder E, Figdor CG, Buschow SI. Proteomics of Human Dendritic Cell Subsets Reveals Subset-Specific Surface Markers and Differential Inflammasome Function. Cell Rep. (2016) PMID: 27626665

Single-cell RNA-seq reveals new types of human blood dendritic cells, monocytes, and progenitors

Abstract
Dendritic cells (DCs) and monocytes play a central role in pathogen sensing, phagocytosis, and antigen presentation and consist of multiple specialized subtypes. However, their identities and interrelationships are not fully understood. Using unbiased single-cell RNA sequencing (RNA-seq) of ~2400 cells, we identified six human DCs and four monocyte subtypes in human blood. Our study reveals a new DC subset that shares properties with plasmacytoid DCs (pDCs) but potently activates T cells, thus redefining pDCs; a new subdivision within the CD1C+ subset of DCs; the relationship between blastic plasmacytoid DC neoplasia cells and healthy DCs; and circulating progenitor of conventional DCs (cDCs). Our revised taxonomy will enable more accurate functional and developmental analyses as well as immune monitoring in health and disease.

Villani AC, Satija R, Reynolds G, Sarkizova S, Shekhar K, Fletcher J, Griesbeck M, Butler A, Zheng S, Lazo S, Jardine L, Dixon D, Stephenson E, Nilsson E, Grundberg I, McDonald D, Filby A, Li W, De Jager PL, Rozenblatt-Rosen O, Lane AA, Haniffa M, Regev A, Hacohen N. Single-cell RNA-seq reveals new types of human blood dendritic cells, monocytes, and progenitors. Science (2017) PMID: 28428369