Integrated single cell multiomic profiling and functional validation reveal distinct cellular routes to human plasma cell differentiation

Reviewer #1 (Public review):

Summary:

Fields et al. investigated the heterogeneity and kinetics of human antibody secreting cell (ASC) differentiation by analyzing ex vivo tonsil samples and using in vitro differentiation modeling. They discovered that a CD30+ intermediate subset emerges in transition from B cell to ASC in both contexts, but not from germinal centers, and they identified cytokines that promote this state. They also identified an isoform of CD44, CD44v9, that is expressed on some ASCs.

Strengths:

The strengths are the novelty of the findings and the identification of two new markers that may be useful for tracking ASC heterogeneity.

Weaknesses:

However, some of this work seems preliminary and would need to be further validated. Some of the data presented was only representative, with limited controls and biological repeats, limiting the interpretation. For example, the role of Mef2c for CD30 expression was not robustly demonstrated. It was not clear if Figure 1 scRNAseq/ATACseq was from multiple donors or just one. Future studies may extend these novel findings and determine the functional relevance of these factors, CD30, and CD44v9 for ASC differentiation and physiology.

Reviewer #2 (Public review):

Summary:

Bhattacharya and colleagues here use cell culture, single-cell RNA and ATACseq sequencing of such in vitro cultures and of ex vivo isolated B-lineage cells to infer an ontogeny for extra-germinal centre B cell differentiation. The manuscript presents a useful potential ontogeny for plasma cells, wherein in vitro cultured naïve human B cells enter a CD30+ intermediate state before moving in subsequent days through a CD44v9+ state before ultimately obtaining a 'mature' antibody-secreting plasma cell phenotype. Ex vivo isolated germinal centre B cells obtain the plasma cell state without expressing CD30 in their development. Phenotype analysis of tonsillar B-lineage cells supports the same phenotype conversion in vivo, although the intermediate cell population was smaller in vivo. The link to CD44v9 expression on developing plasma cells is inferred to be for extra-GC (T-independent) responses, but the data presented leave this equivocal, and the functional importance of developing via a CD30+CD44v9+ intermediate is not investigated.

Strengths:

The article presents a solid potential ontogeny for PC development, wherein some differentiating B cells acquire a CD30+ state, transition through a CD44v9+CD30+ state, then downmodulate CD30 before obtaining canonical CD38+ 'PC' status. A strength is the integration of in vitro cultured B cell results with tonsillar B-lineage cell data sets, and careful flow cytometry of the in vitro cultures over several days to infer lineage. The data provide reasonable support for the concept. CD30+ cells are shown to develop readily from naïve B cells in culture, but uncommonly from GC B cell cultures. A nice piece of data is Figure 6B, which shows reasonably strong correlative changes in phenotype through the assumed ontogeny, and this fits with the expected trajectory of maturation.

Weaknesses:

The most important weakness throughout is the non-absolute nature of the relationship. An example is seen in that the sorted ex vivo GC B cells also give rise to the 'extra-GC' phenotype of plasma cell, suggesting that while the profile is enriched, it is not absolute. There is a further weakness, as while cultures are run for several days, division-associated shifts in PC phenotype are not mapped; such would greatly strengthen the weight of the argument, and show conditional shifts in phenotype associated with division, an uncontrolled parameter in the mix. For example, for the MEF2C A388 inhibition experiments, it would be strong evidence of the pathway/process contributing if a by-division peak increase in CD30+ population was demonstrated in the early days of culture.

There are some basic sort experiments performed (e.g. 3C-3F), which show that the CD30+ cells do give rise to PC preferentially, but what is missing is the step-wise phenotype shifts in these sorted populations, which should support the trajectory shown in Figure 3B and (the in vitro equivalent of) 6B. It would emphatically support the trajectory to show the cellular phenotypes on the PC with sorting based on CD30, CD44v9, CD27, and CD20 expression, and following outcome phenotypes 24-48 hours later, if the inferred maturation trajectory is true.

There are also specific weaknesses with the bioinformatics, in that, while the analyses are likely appropriate, unpresented data is necessarily used to shape the argument. For example, Figure 1C shows bubble plots for two plasma cell sets, yet, of archetypal PC-expressed genes, only IRF4 is demonstrated to confirm they are true PC, and the gene is not universally expressed in cells in the clusters. For this figure, it would help to expand the bubble plot to show J-CHAIN, XBP-1, CIITA and PRDM1 or other appropriate PC demarcating molecules. Similarly, in Fig 2B, more evidence of a bifurcation in state is needed than that CD44v9 distinguishes PC1 from PC2 clusters-this is the stated conclusion, but 2A depicts that 50% of PC1 relatively weakly express CD44, while <25% of PC2 express it. Demonstrating additional molecules or genes distinguishing the clusters would improve veracity. Figure 2F shows clonal lineages, but it would be helpful to see somatic hypermutation burdens and learn if they differ between the demarcated subsets. I also find the pseudotime analyses of limited value, as some of the branches follow trajectories that are unrealistic biologically, so less weight should be placed on the pathways to which they do or do not point (i.e., the notion that GC B cells do or do not give rise to particular PC subsets).

Statistically, some of the experiments are single wells from single donors, so there is a low level of confidence and no reproducibility demonstrated for some aspects of the study, which is a weakness.

Paradoxical to the argument that it is the TI response process being modelled, it is presented that CpG stimulation, plus proxy T cell help (CD40L), drives the CD30+ phenotype best with the addition of the GC-associated cytokine IL-21. This should be carefully considered and discussed.

Overall, in addition to presenting more contextual information from the bioinformatics, the best way to solidify the data set, in my vie,w would be to revisit the hypothesis with two additional experimental approaches: (1) to incorporate division tracing into the ontogeny studies and (2) to perform lineage tracing on sort-purified populations at different stages of the maturation process.