key: cord-0775455-d9kokt56
authors: Wilk, Aaron J.; Rustagi, Arjun; Zhao, Nancy Q.; Martin, Beth A.; Rogers, Angela J.; Blish, Catherine A.
title: Additional analyses exploring the hypothesized transdifferentiation of plasmablasts to developing neutrophils in severe COVID-19
date: 2020-10-16
journal: bioRxiv
DOI: 10.1101/2020.10.15.339473
sha: 45d7ce6663852db589e0571549859af05cf2ff12
doc_id: 775455
cord_uid: d9kokt56

We thank Alquicira-Hernandez et al. for their reanalysis of our single-cell transcriptomic dataset profiling peripheral immune responses to severe COVID-19. We agree that careful analysis of single-cell sequencing data is important for generating cogent hypotheses but find several aspects of their criticism of our analysis to be problematic. Here we respond briefly to misunderstandings and inaccuracies in their commentary that may have led to misinformed interpretation of our results.

Abstract 24 We thank Alquicira-Hernandez et al. for their reanalysis of our single-cell transcriptomic dataset 25 profiling peripheral immune responses to severe COVID-19. We agree that careful analysis of 26 single-cell sequencing data is important for generating cogent hypotheses but find several 27 aspects of their criticism of our analysis to be problematic. Here we respond briefly to 28 misunderstandings and inaccuracies in their commentary that may have led to misinformed 29 interpretation of our results. 30 31

Main 32

Alquicira-Hernandez et al. 1 question the plausibility of the potential lineage relationships 33 between plasmablasts and developing neutrophils that we postulated as a part of our recent 34 work 2 . We appreciate their commentary and concur that careful computational analysis of 35 single-cell RNA sequencing (scRNA-seq) data is necessary. Our study, the first to publicly share 36 scRNA-seq data to profile immunity in COVID-19, was by its design and execution descriptive, 37

correlative, and hypothesis-generating, given the limitations of the dataset acknowledged in our 38 original manuscript. Our goal was to develop a resource for the scientific community to better 39 understand COVID-19, and to identify distinctive immune features for further study. We regret 40 that we may have not adequately conveyed the hypothesis-generating nature of our study; if 41 any reader came away with the impression that we had claimed to have "proven" the existence 42 of a plasmablast to neutrophil transition, this was not our intent, and we apologize. 43 44 We chose our words very carefully when describing our findings, explicitly choosing not to say 45 that we had "proved" or "concluded" new hypotheses with scRNA-seq data alone, particularly as 46 it related to a potential transdifferentiation pathway. In this regard, we point to our original 47 manuscript rather than Alquicira-Hernandez et al.'s paraphrasing, which left out important 48 context for our stated conclusions. To wit, our final statement on this putative pathway reads, 49 "Collectively, we observe a developing neutrophil population that may be characteristic of ARDS 50 in severe COVID-19 infection; our data suggest that these cells may derive from plasmablasts, 51 but they may also represent developing neutrophils derived from emergency granulopoiesis" 2 . 52 53

Alquicira-Hernandez et al. argue that transdifferentiation between plasmablasts and developing 54 neutrophils is biologically implausible and, therefore, that the association between these two cell 55 types in uniform manifold approximation and projection (UMAP) space must represent an 56 artifact of the computational pipeline we selected. there is still a phenotypic association between plasmablasts and developing neutrophils ( Figure  100 1a). This is similar to the UMAP projections generated by Alquicira-Hernandez et al., which still 101

show developing neutrophils closely associated with plasmablasts, with several plasmablasts 102 embedding closer to developing neutrophils than with other plasmablasts (Figure 1a , purple 103 box). To examine this phenotypic relationship outside of the non-linear dimensionality reduction 104 manifold space of UMAP, we additionally hierarchically clustered pseudobulk average gene 105 expression profiles of each cell type in our dataset, which again indicates that plasmablasts and 106 developing neutrophils are phenotypically related (Figure 1b) . It is important to note that this 107

does not indicate relationship in terms of cell lineage, but merely relationship of transcriptional 108 phenotype. Taken together, these analyses re-affirm our previous decision to explore this 109 relationship further with RNA velocity analysis. 110 111 We performed RNA velocity analysis using preprocessing parameters employed in our original 112 manuscript (Figure 1c , green box) and using preprocessing parameters used by Alquicira-113

Hernandez et al. (Figure 1d , purple box). To analyze the dynamic relationship between 114 additional cell types that may be biologically related to plasmablasts and developing neutrophils, 115

we embedded only plasmablasts, developing neutrophils, B cells, and a population of low-116 density mature neutrophils we identified. We found that, with both sets of preprocessing 117 parameters, developing neutrophils appear to transition from plasmablasts and do not occupy 118 similar UMAP manifold space as B cells and mature neutrophils (Figure 1c, d) . 119 120

Alquicira-Hernandez et al. hypothesized that plasmablasts should be more closely related to B 121 cells than developing neutrophils, and that developing neutrophils should be phenotypically 122 associated with mature neutrophils. Upon finding that an embedding of these four cell types 123 alone showed a relative lack in relatedness between plasmablasts and B cells, the authors 124 concluded that plasmablasts and developing neutrophils must be misclassified as related cell 125 types. However, it is incorrect to assume B cells and plasmablasts should be phenotypically 126 related in UMAP space, as these cell types are dramatically different in terms of gene module 127 expression (eg. proliferation) that is easily detected at the transcriptional level (Figure 1b) , and 128 because the kinetics of B cell-to-plasmablast differentiation in these patients may not enable 129 identification of intermediate cell states in the periphery. While it does remain possible that 130 developing neutrophils and plasmablasts are related in UMAP space because they are both 131 proliferative cell types, there are other proliferative T and NK cells in the dataset that are not 132 phenotypically related and this argument does not have bearing on trajectories predicted by 133 RNA velocity. We thus conclude that our selection of preprocessing parameters was reasonable 134

and would have led to the same hypotheses had we chosen different parameters. 135 136

Alquicira-Hernandez et al. also imply that we did not fully consider the possibility that 137 hemophagocytic lymphohistiocytosis (HLH) could have explained our findings because of the 138 difficulty in making this diagnosis. First, we would like to reiterate that such an explanation 139

would be expected to result in an increase in complexity (# genes detected per # UMI 140 sequenced per cell) of developing neutrophils, which we did not observe (Extended Data Figure  141 9 

No evidence that plasmablasts 183 transdifferentiate into developing neutrophils in severe COVID-19 disease

A single-cell atlas of the peripheral immune response in patients with 186 severe COVID-19

RNA velocity of single cells

Generalizing RNA velocity to 189 transient cell states through dynamical modeling

Regulation of B lymphocyte and macrophage development by 192 graded expression of PU

Historical origins of transdifferentiation and reprogramming

C/EBPα induces highly efficient macrophage transdifferentiation of B 196 lymphoma and leukemia cell lines and impairs their tumorigenicity

CCAAT/enhancer binding protein α (C/EBPα)-induced transdifferentiation 199 of pre-B cells into macrophages involves no overt retrodifferentiation

Single cell RNA-seq identifies the origins of heterogeneity in efficient 202 cell transdifferentiation and reprogramming

Examination Findings in SARS-CoV-2 Infection

Peripheral Blood Smear Demonstration of Lymphocyte Changes 206 in Severe COVID-19

Coronavirus disease 2019 induces multi-lineage, morphologic changes in 208 peripheral blood cells