1. Cell Biology
  2. Immunology and Inflammation
Download icon

Identification of a super-functional Tfh-like subpopulation in murine lupus by pattern perception

  1. Stefanie Gryzik  Is a corresponding author
  2. Yen Hoang  Is a corresponding author
  3. Timo Lischke
  4. Elodie Mohr  Is a corresponding author
  5. Melanie Venzke
  6. Isabelle Kadner
  7. Josephine Poetzsch
  8. Detlef Groth
  9. Andreas Radbruch
  10. Andreas Hutloff
  11. Ria Baumgrass  Is a corresponding author
  1. German Rheumatism Research Center (DRFZ), A Leibniz Institute, Germany
  2. University of Potsdam, Germany
  3. Charité, Campus Mitte, Germany
Research Article
  • Cited 0
  • Views 731
  • Annotations
Cite this article as: eLife 2020;9:e53226 doi: 10.7554/eLife.53226

Abstract

Dysregulated cytokine expression by T cells plays a pivotal role in the pathogenesis of autoimmune diseases. However, the identification of the corresponding pathogenic subpopulations is a challenge, since a distinction between physiological variation and a new quality in the expression of protein markers requires combinatorial evaluation. Here, we were able to identify a super-functional follicular helper T cell (Tfh)-like subpopulation in lupus-prone NZBxW mice with our binning approach "pattern recognition of immune cells (PRI)". PRI uncovered a subpopulation of IL-21+ IFN-γhigh PD-1low CD40Lhigh CXCR5- Bcl-6- T cells specifically expanded in diseased mice. In addition, these cells express high levels of TNF-α and IL-2, and provide B cell help for IgG production in an IL-21 and CD40L dependent manner. This super-functional T cell subset might be a superior driver of autoimmune processes due to a polyfunctional and high cytokine expression combined with Tfh-like properties.

Introduction

CD4 T cells play a major role in autoimmune diseases such as systemic lupus erythematosus (SLE) by accumulating as autoreactive disease-promoting memory cells and amplifying inflammation as well as providing B cell help (Suárez-Fueyo et al., 2016). It is known that dysregulated cytokine production, metabolic alterations and aberrant cell signalling cause abnormalities in T cell differentiation and hyperactivation of B cells and thus contribute to lupus disease (Moulton et al., 2017). For SLE, there are several studies in mouse models and humans, showing that the progressive disease is associated with lower expression of TNF-α, IFN-γ, and IL-2 (Humrich et al., 2010) as well as higher expression of PD-1 (Jiao et al., 2014), IL-21 (Wang et al., 2014), and IL-10 (Facciotti et al., 2016). Despite the detection of these individual parameters, a comprehensive combinatorial characterization of T-cell subsets in SLE is still lacking.

Recent articles have revealed enormous and unexpected heterogeneity and complexity in T cell subpopulations with partially overlapping properties and functions of the cell subsets, such as follicular T helper (Tfh) and Tfh-like cell subsets (Trüb et al., 2017; Wong et al., 2015; Wong et al., 2016). Of particular interest are the recently observed human T peripheral helper cells (Tph), which expand in autoimmune diseases (Bocharnikov et al., 2019; Christophersen et al., 2019; Ekman et al., 2019; Rao et al., 2017). These and other studies show on the one hand the enormous potential to gain new biological insights from multidimensional cytometry data. On the other hand, they reveal the tremendous demand for new analysis and visualization approaches, especially for the identification of small but different characteristics in apparently homogeneous subpopulations, which is emphasized by numerous review articles (Kvistborg et al., 2015; Mair et al., 2016; Newell and Cheng, 2016; Saeys et al., 2016). Most of the current analysis strategies are based on clustering and the respective frequencies of the clusters (Levine et al., 2015; Qiu et al., 2011; Spitzer et al., 2015; Wong et al., 2016). The main problem with these methods is that they tend to find mainly cell-populationclusters which are significantly high in phenotypical contrast. If the cell subpopulations are very similar or even partially overlapping, as in the present investigation, then a cluster approach is usually not very meaningful (Newell and Cheng, 2016; O'Neill et al., 2013; Saeys et al., 2016).

To circumvent this problem, we developed the bin-based ‘pattern recognition of immune cells (PRI)’ strategy. PRI is a reproducible analysis and visualization approach to examine the combinatorial expression patterns of proteins in a mixture of similar cell subpopulations. Here, we have combined PRI with comprehensive cytometric measurements of Th cells from NZBxW F1 mice, a common lupus model. The results of our approach provide evidence for a super-functional Th cell subpopulation mainly characterized by IL-21+ IFN-γhi PD-1low, TNF-αhi IL-2hi, and functional properties of Tfh-like cells.

Results

Introduction into visualization of multi-parametric flow cytometry data with pattern recognition of immune cells (PRI)

To analyze pathogenic memory Th cell subsets (CD4+ CD44+ cells (Swain, 1994), in the following called Tmem cells), we used the NZBxW mouse model, which is one of the best-established models for human SLE, and visualized the combinatorics of three proteins from splenic T cells of old diseased mice (chronic inflammation, high proteinuria). We plotted IFN-γ and TNF-α expression of Tmem cells with conventional tools such as contour plots (Figure 1A), a pie chart (Figure 1C), a stacked bar chart (Figure 1D), and the recent tool ‘color mapping of dots’ by FlowJo (Figure 1B). As expected, IFN-γ is almost exclusively produced by CD44+ cells (Figure 1D).

PRI allows the visualization of the combinatorial protein expression.

Flow cytometry data from stimulated (PMA/ionomycin) splenic T cells from NZBxW mice were analyzed either conventionally or by bin plots. (A) Overlay of IFN-γ+ cells on all T cells. (B) FlowJo‘s color maps showing MFI of IFN-γ. (C, D) Frequencies of the co-production of TNF-α, CD44+ and IFN-γ are depicted as categories (C) and as individual combinations (D). (E) For bin plots, the x-y-plane of TNF-α and CD44 is divided into small, equally sized bins (asinh = 0.2). If a bin contains the minimum number of cells (10), a statistical feature is calculated such as cell density (left), frequency (middle) or MFI (right) of IFN-γ. The scale of the feature is represented as a pseudo-color code. (F) Relative mean expression level per bin (MFI+) of IFN-γ only in IFN-γ producing cells (left), of IL-2 in IL-2+ (middle) and PD-1 in PD-1+ cells (right). Grey bins contain less than 10 Z+ cells. Data represent one old diseased mouse (A–F). (E, F) Cell frequencies per quadrant are calculated from the number of cells per sample (black) and number of Z+ cells per sample (green). Data represent three independent experiments with (C, E) n = 4 mice per group.

We confirmed this fact with our PRI visualization tool by plotting different features of IFN-γ expression as heat maps onto the x-y plane of TNF-α vs. CD44 (Figure 1E). To this end, the x-y plane was divided into equally sized bins. The resulting plots were called bin plots. Per bin, we calculated and depicted the features (i) cell density, (ii) frequency of IFN-γ-producing cells (%), (iii) mean fluorescence intensity (MFI) of IFN-γ, and (iv) the relative expression level of IFN-γ calculated as mean fluorescence intensity of IFN-γ only of IFN-γ+ cells (MFI+ (IFN-γ)) (Figure 1F). Apparently, our bin plot comparison shows that the bins with the highest expression levels of IFN-γ are TNF-αhi CD44+ (Figure 1F) and are in an area with low cell density and low-to-intermediate frequency of IFN-γ (Figure 1E). Such information cannot be extracted from the conventional plots.

Next, we used PRI for ‘pseudo-multi-parametric viewing’ by color-coding different parameters of the same sample (old diseased NZBxW mouse) on the same x-y plane. This visualized that the IFN-γhighest bin area contains cells that are also high in IL-2, but low in PD-1 (Figure 1F). The applied pseudo-multi-parametric viewing of PRI supports the discovery and characterization of specific subpopulations with certain or unique properties not only in an easy to understand and comprehensive manner but also on a reproducible basis.

Disease kinetics reveal a subsequent reduction of poly-functional memory T cells

Despite the suspected importance of combinatorial expression of inhibitory receptors (such as PD-1) and cytokines for immunity (McKinney et al., 2015), to our knowledge, there are no comprehensive co-expression studies with CD4+ T cells at single-cell protein level for SLE.

To characterize the disease kinetics of lupus development, we scored NZBxW mice according to their age and proteinuria (Figure 2A), determined their auto-antibody concentrations and analyzed the expression of PD-1 and four main cytokines in CD4+ Tmem cells. Increased frequencies of PD-1+ and IL-10+ cells as well as decreased frequencies of IL-2+ and TNF-α+ cells were age- and disease-dependent (Figure 2A–C). The co-production of the cytokines changed in the course of the disease to a lower degree of poly-functionality, visible by the accumulation of non-producing and IFN-γ single-producing cells (Figure 2—figure supplement 1).

Figure 2 with 3 supplements see all
The cytokine co-expression is altered in PD-1+ T cells in young and old mice.

(A) Disease status of NZBxW mice was scored according to their age and proteinuria (PU). The disease score was correlated with the frequencies of PD-1+ T cells and serum levels of anti-ds-DNA-IgG antibodies. (B) Disease-associated changes of the protein expression shown by histogram overlays (top) and statistical analyses (bottom). (C) Representative gating scheme of PD-1-, PD-1low and PD-1hi subpopulations in CD4+ CD44+ T cells from mice of disease score 1 and 5, respectively. (D) Statistical analyses of the frequencies of IL-2, TNF-α, IFN-γ and IL-10 producers in PD-1 subpopulations from mice with disease score of 1 (light grey) and 5 (dark grey). (E) Frequencies of TNF-α + cells of the PD-1+ IFN-γ+ subpopulation in disease score 1 and 5. (F) Representative bin plots of disease score 1 and 5 with PD-1 (x-axis), IFN-γ (y-axis) displaying the frequencies of TNF-α, IL-2 and IL-10, respectively, per bin. Cut-off for PD-1hi cells is marked with dashed lines. Data represent two independent experiments with (A), n = 2 mice for each score, (B– F) n = 6–7 mice per group. Samples were compared using the Mann Whitney test (B), a repeated measure two-way ANOVA with Geisser-Greenhouse correction and Dunnett‘s multiple comparison test (D) and a two-sided unpaired t-test (E). Data are presented as the mean ± SEM.

As loss of poly-functionality is considered as a hallmark of chronic infections and often used as an end-point to evaluate exhaustion of CD8+ and CD4+ T cells (Larsen et al., 2012; Tilstra et al., 2018), we investigated the combinatorial expression of the four cytokines with PD-1.

PRI facilitates the combinatorial characterization of chronically activated T cells

We classified the disease scores 1–2 as 'young', and 3–5 as 'old diseased’ and compared these scores using combinatorial analyses. According to publications on chronic infections (Hwang et al., 2016), we divided the PD-1 expression into three ranges (negative (-), low to intermediate (low), and high (hi) (Figure 2C), using FMO controls and loss of function in IL-2 expression and plotted the diagrams for conventional analysis (Figure 2D and E) and PRI visualization (Figure 2F). During disease progression of lupus nephritis, a higher expression of inhibitory receptor PD-1 was associated with a lower degree of CD4+ Tmem-cell functionality regarding IL-2, TNF-α and IFN-γ expression (Figure 2D—figure supplement 2A and B). This was underlined when comparing PD-1+ (low and hi) and PD-1 cell populations in mice of score 5: the frequencies of IL-2- and TNF-α-producing cells were lower but that of IFN-γ- and IL-10-producing cells were higher in the PD-1+ cell populations (Figure 2F—figure supplement 2B-D). The frequencies of TNF-α+ Tmem cells within the IFN-γ+ PD-1+ CD4+ subpopulations were reduced with lupus nephritis (Figure 2E), similar to CD8+ Tmem cells in chronic infections (Utzschneider et al., 2016).

In contrast to pie charts (Figure 2—figure supplement 2E), pseudo-multi-parametric viewing of bin plots allows to compare whole patterns (frequencies and expression level) and bin areas of different cytokines which are color-plotted on the same x-y-plane. Apparently, bins in the IFN-γhiPD-1–/low area contain the highest frequencies and expression levels of TNF-α and IL-2, but almost no IL-10 production (Figure 2F—figure supplement 2C and D). The low co-expression of IL-10 and IL-2 as well as TNF-α was confirmed by contour plots (Figure 2—figure supplement 3A). Compared to the conventional analysis of the PD-1 subgroups (Figure 2D—figure supplement 2A, B and E), the higher combinatorial complexity of PRI enabled the identification of more informative subpopulations like PD-1low IFN-γhi cells (Figure 2F—figure supplement 2C).

Obviously, the combinatorial pattern of three parameters in bin plots (Bendfeldt et al., 2012) supports a more conclusive statement than just determining the frequencies of quadrants after subsequent gating, which demonstrates a particular strength of our semi-continuous bin plotting. This might explain the contrary results of Kasagi et al. which report ‘higher levels of IFN-γ in PD-1hi than in PD-1low cells’ in diseased NZBxW mice in the CD4+ population (Kasagi et al., 2010).

Reproducible and statistically robust bin plot patterns exemplarily shown with combinatorial IL-10 expression

We discovered a similar bin plot pattern of IL-10 (higher frequencies associated with a PD-1hi IFN-γlow phenotype) in all studied mice despite high differences in the frequencies of IL-10 expression per mouse and disease score (Figure 2—figure supplement 3B and D), and even when mice from a different experiment were concatenated (three mice each) (Figure 2—figure supplement 3C). Using IL-10 bin plots, we also demonstrated some general advantages of the PRI approach: (i) Varying the cut-off value for IL-10+ cells had a high impact on IL-10 frequencies, but almost no influence on their bin pattern (Figure 2—figure supplement 3E). (ii) The plotted bin area with 10 cells per bin provides statistically reliable data, as demonstrated by the relative standard error of the mean (RSEM) (Figure 2—figure supplement 3F). (iii) A comparison of different samples regarding IL-10 expression, here different donor mice, different experiments and different disease scores, is easily feasible due to their bin pattern similarities.

Bin plots revealed an IL-21+ super-functional T cell subpopulation

Recently, IL-10 was found to be highly co-expressed with IL-21 (60%) in Tmem cells in SLE patients (Facciotti et al., 2016). Therefore, we additionally included the measurement of IL-21 in the combinatorial expression study of NZBxW T cells, which led to 32 possible combinations for the co-production of five cytokines (Figure 3—figure supplement 1A). Contrary to Facciotti et al., 2016, however, in NZBxW mice we observed neither substantial IL-10 and IL-21 co-expression with conventional contour plots nor strongly overlapping bin regions in bin plots on the TNF-α and IFN-γ plane (Figure 3A and D). In contrast, IL-21 is particularly co-expressed with IL-2 and TNF-αhi as demonstrated with contour plots and bin plots analyzing two (Figure 3A), three (Figure 3D) or four cytokines (Figure 3E) simultaneously.

Figure 3 with 1 supplement see all
Most IL-21 producers in the NZBxW model strongly co-express Th1 cytokines but no CXCR5.

(A) Co-production of IL-21 with cytokines and the Tfh markers PD-1 and CXCR5 in concatenated CD4+ CD44+ cells from young and old diseased mice. (B) Tfh cells were identified based on their high expression of PD-1 and CXCR5 positivity. (C) Statistical analyses of IL-21+ subpopulations extracted from FlowJo. (D, E) Cytokine co-production is analyzed by bin plots showing the frequencies of IL-21+, IL-2+, IL-10+ (D) and IL-2+ IL-21+ cells (E). Green numbers indicate the frequency of IL-2+ IL-21+ double producers relative to all cells per quadrant (E). (F, G) Bin plots of concatenated samples of young and old diseased mice with PD-1 (x-axis), IFN-γ (y-axis) displaying the cell density and frequencies of IFN-γ+ and IL-10+ (F) and CXCR5 (G), respectively, per bin. Percentages written in italic letters represent the population size of PD-1low IL-21+ IFN-γ+ cells. Data are representative of three independent experiments (A–B). (C) Pooled data from two experiments involving n = 4 (young) and n = 6 (old) mice per group. Samples were compared using Mann-Whitney-U-test. Data are presented as the mean ± s.e.m. (D–G) The samples were concatenated from n = 3 young and old mice of the same experiment. Cut-off for PD-1hi cells is marked with dashed lines. (E) If a bin contains the minimum number of cells (5), the frequency of the third marker is shown in pseudo-colors. Cell frequencies per quadrant are calculated on the number of cells per sample (black) and number of Z+ cells per sample (green).

When searching for IL-21+ Tmem cell subpopulations with differences in frequencies between young (score 1–2) and old diseased (score 3–5) NZBxW mice, we identified a tendency towards higher frequencies of IL-21+ and IL-21+ IFN-γ+ cells in old compared to young mice using conventional gating analysis. However, they were not statistically significant in contrast to PD1hi cells of Tmem and CD44+ cells of CD4+ subset (Figure 3C).

Gating for PD-1hi and PD-1low populations revealed that all populations studied had higher frequencies in older mice, but only the PD-1low IL-21+ IFN-γ+ population was significantly higher (Figure 3C). These data are comparable with the PRI-data (Figure 3—figure supplement 1B), even if the PD-1low IL-21+ IFN-γ+ population was not significantly different between young and old diseased mice. Although IL-21 is a hallmark of Tfh cells, in the NZBxW model, the expression of IL-21 is not correlated with a PD-1hi CXCR5+ Tfh cell phenotype (Figure 3G—figure supplement 1D).

Obviously, the frequencies of IL-21+ cells among Tmem cells do not change significantly with higher disease scores, but rather the properties of these cells to co-produce IFN-γ and PD-1 (Figure 3C and F—figure supplement 1B and C). The identified PD-1low IL-21+ IFN-γhi cell subpopulation even expressed high amounts of IL-2 and TNF-α in both young and old diseased mice. Therefore, in the following they are referred to as super-functional T (Tsh) cells.

Pseudo-multi-parametric viewing of bin plots to analyze co-expression properties

Based on our experience (Fuhrmann et al., 2016) that the co-expression of cytokines and receptors matters, we further characterized the identified IL-21+ IFN-γhi PD-1low Tmem cells by studying the bin pattern of 42 functionally relevant proteins on the PD-1 and IFN-γ plane (17 out of them are shown in Figure 4A; Figure 4—figure supplement 1A and B). We used IFN-γ instead of IL-21 as the second parameter on the y-axis because the intensity range of IFN-γ is much wider than the range of IL-21 and therefore better for visualizing preferred expression areas of the most Z-parameters. Due to the reproducible expression patterns, PRI allowed the analysis of proteins from different mice and experiments. Again, PD-1low IFN-γhi bins showed the highest probability of IL-21+ cells. T cells in these bins predominantly co-expressed CXCR3 and T-bet and lacked FoxP3 expression. The IL-21+ expression pattern overlapped only partially with that of Bcl6, CXCR5 or CXCR4 (Figure 4A—figure supplement 1A). Furthermore, the IL-21+ IFN-γhi PD-1low Tmem cells additionally produced CD40L and ICOS, two receptors involved in the interaction of T and B cells. CD40L was even produced in the highest amounts by IFN-γhi cells (Figure 4B—figure supplement 1A). In contrast, the stimulatory receptors ICOS, OX40, GITR and CD27 showed lower intensities of expression in the PD-1low IFN-γhi bin area but higher intensities in the Tfh (Bcl6hi) bin area. The same is true for the inhibitory receptors TIGIT, CTLA4 and BTLA, which were mainly produced by Tfh cells, chronically activated T cells and/or Treg cells (Figure 4B—figure supplement 1). All these properties and their intermediate-to-high PSGL-1 expression suggest a Tfh-like cell subpopulation with Th1 properties instead of extrafollicular Tfh cells (Choi et al., 2015; Odegard et al., 2008).

Figure 4 with 1 supplement see all
Super-functional T cells exhibit Th1 characteristics and are CD40Lhi ICOS+.

(A) Bin plots visualize the co-expression of PD-1 and IFN-γ with various proteins in stimulated (PMA/ionomycin) splenic T cells of old mice. (B) Distribution of the top 50% Z+ cells of selected markers from (A). If a bin contains the minimum number of cells (10), the frequency of the third marker is shown in pseudo-colors. Cell frequencies per quadrant are calculated on the number of Z+ cells in the quadrant (red) and number of Z+ cells per sample (green). Data are representative for at least two independent experiments with n ≥ 3 mice.

As the IL-21+ IFN-γhi PD-1low Tmem cells lacked CXCR5 expression as other Tfh-like cells, we addressed their contribution to IL-21+ cells (Figure 5A and B) in spleen and their tissue distribution (Figure 5C). We were able to localize them in spleen and non-lymphoid organs, liver and lungs at comparable frequencies (Figure 5C and D). We could also detect them in kidneys and peripheral blood (Figure 5C and D and data not shown). Again, IL-21+ cells co-produced high levels of IFN-γ in all these organs, but almost no CXCR5 (Figure 5D). Interestingly, the Tsh cell subset in diseased NZBxW mice is two times bigger than the Tph and 10 times bigger than the Tfh cell subset.

Super-functional T cells in peripheral organs exceed extrafollicular T cells and Tph cells in terms of frequency.

(A) Absolute numbers of PD-1 subsets in spleens. (B) Frequency of IL-21 producers in spleens. (C, D) Frequency of IL-21 producers in terms of localization and in terms of PD-1 subset. Data represent two independent experiments with n = 4 mice per organ. Data are presented as the mean ± s.e.m.

PRI reveals differences between Tfh and Tfh-like cell subpopulations

To better define and compare important properties of Th cell subpopulations, we determined expression values of bin areas with multiple samples. To this end, we covered the PD-1-IFN-γ plane with a 3 × 3 grid, which led to nine categories according to their PD-1 and IFN-γ expression levels (Figure 6A). The statistical analysis of many markers within these categories (Figure 6B—figure supplement 1A) confirms a particular pattern of Th cell subpopulations (shown as a schematic distribution in Figure 6D). These include the highest frequencies of IL-21+ T cells in PD-1low T cells, which was demonstrated by their normalization to CD44+ T cells (Figure 6C) and their positive correlation with the IFN-γ expression level (Figure 6B).

Figure 6 with 1 supplement see all
Comparison of Tfh and Tfh-like cell subpopulations by bin patterns.

(A) Subdivision of Tmem cells into nine categories based on their PD-1 and IFN-γ expression (compare to D). (B) Barplots depicting the frequencies of CXCR5, Bcl6 and IL-21 producers in respective subpopulations. (C) Frequency of the IL-21 producers of each subpopulation relative to CD44+ cells. (D) Areas with highest probability for Treg, Tfh, Th1 and IL-21+ cells. (E) Representative bin plots of the frequencies of double producers for CXCR5/Bcl6, IL21/Bcl6, IL21/CXCR5, IL21/ICOShi, IL21/CD40L in splenic T cells of old diseased mice. (F) Representative bin plot of the frequency of IL21 producing, but CXCR5 non-producing cells in splenic T cells of old diseased mice. (G) T cells were sub-divided according to their CXCR5 and Bcl6 expression levels into negative (-), low and high (hi) expressors (left). Their IL-21 production was investigated by histogram overlays (right). (H) 3D heatmap showing frequencies of Bcl6 (left) and IL21 producers (right) with PD-1 (x-axis) and IFN-γ (y-axis). All data represent three independent experiments with (B, C), n = 3–9 mice and (E–G), 3–11 mice. E-F, Frequencies of ‘double’ producers are calculated per quadrant (red) and per all CD44+ cells (green). Grey bins contain less than 10 Z+ cells. Data are presented as the mean ± s.e.m.

Using viSNE plots, the association of IL-21+ cells with high IFN-γ expression and their separation from PD-1hi T cells could also be demonstrated (Figure 6—figure supplement 1B). However, viSNE was only partially appropriate for dim markers such as CXCR5 or Bcl6, because it illustrates the MFI over all cells. Color Maps (FlowJo 7.2) have similar issues with suboptimal color distributions for dim markers and only represent MFIs and not expression levels (Figure 6—figure supplement 1D). PRI, instead, allows the calculation of relative expression levels of marker Z of Z+ cells (MFI+ (Z)) (Figure 6—figure supplement 1C).

To further delineate Tfh-like and Tfh cell subpopulations, we investigated the co-expression of IL-21 with the Tfh marker proteins CXCR5 and Bcl6 as well as the T/B cell interaction receptors ICOS and CD40L. Again, we used PRI to calculate the frequencies of positive cells, this time for two defined marker proteins per bin to visualize double producers (Figure 6E). CXCR5+ Bcl6+ (Tfh cells) predominantly appeared in bins at the PD-1hi side, while there were only few bins with IL-21+ Bcl6+, IL-21+ CXCR5+ and even IL-21+ ICOShi T cells. Instead, the co-production of IL-21 and CD40L is localized in bins with IFN-γhi T cells (Figure 6E). IL-21 expression levels did not correlate with Bcl6 and negatively with CXCR5 (Figure 6G).

Plotting only IL-21+ CXCR5- cells in PRI-bins with a PD1hi threshold visualized Tsh and Tph cell subsets together. Most of these cells are not PD-1hi but PD-1lo (Figure 6F).

The discrepancy between IL-21 and Tfh marker expression became evident again when the frequencies (Figure 6H) or expression levels (Figure 6—figure supplement 1E) of IL-21 and Bcl6 producers were visualized by 3D surface plots, since positive cells for each protein appeared in different areas.

Super-functional T cells provide B cell help

Next, we asked for similarities and differences between already described Tfh-like cell subpopulations (Choi et al., 2015; Hutloff, 2018; Vu Van et al., 2016) and the specified IL-21+ IFN-γhi PD-1low Tmem cells identified here. The main difference is the high PD-1 expression which has been described for all previously identified Tfh-like subsets, including the human Tph cells (Bocharnikov et al., 2019; Christophersen et al., 2019; Ekman et al., 2019; Rao et al., 2017).

Expression level, number and frequency of cytokines were usually not characterized in parallel with Tfh-like marker proteins. The most important similarities between the here specified and already known Tfh-like cells, besides IL-21 expression, are the absence of CXCR5 and expression of ICOS and CD40Lhi (Figure 4A and B—figure supplement 1). These characteristics have prompted us to investigate whether the IL-21+ Tmem cells are able to induce antibody production by co-culture of CD44- naive B cells and CD44+ CXCR3+ non-Tfh non-Treg cells (Figure 7A). CXCR3+ served as a surrogate marker for IFN-γ production in FACS live sorting and helped to enrich IL-21+ IFN-γhi Tmem cells by a factor of five compared to CD44+ T cells (Figure 4A, Figure 7—figure supplement 1). Polyclonal stimulation of these CXCR3+ T cells induced a high IgG production by B cells (Figure 7B) which was reduced either by blocking IL-21 or CD40L signaling and was abrogated by blocking both pathways (Figure 7B). We could rule out B cell helping activity of high IFN-γ levels of these cells, as reported for IgG2a induction (Snapper and Paul, 1987), since the blockade of IFN-γ did not diminish total IgG production (Figure 7B).

Figure 7 with 2 supplements see all
Super-functional T cells provide help for IgG production in co-cultures with B cells and expand B cells with activated GL7+ Fas+Germinal Center-like phenotype.

(A) Splenocytes were FACS-sorted for naïve B cells (CD19+ B220+ CD44low) and Th1 cells (CD4+ CXCR5- CD44+ CD25- CXCR3+). Both fractions were co-cultured for 5 days in presence or absence of antibodies against B cell activating cytokines. T cells were stimulated with agonistic antibodies against CD3 and CD28. As positive control, B cells were stimulated with IL-21 and an agonistic anti-CD40 antibody. IgG in the supernatant was detected by an ELISA. (B) ELISA results. One representative experiment of three is shown. Statistical analysis was performed by one-way ANOVA with Dunnett’s multiple comparison test. (C) Total CD19+ B220+ B cells were co-cultured for 5 days either alone (negative control, -), or with IL-21 + anti-CD40 antibody (positive control, +), CXCR3+ PD-1lo Tsh (PD-1lo CXCR3+), CXCR3- PD-1lo CD4 T cells (PD-1lo CXCR3-), or PD-1hi CXCR5+/- cells (PD-1hi) FACS-sorted from pooled splenocytes of 2-year-old C57Bl/6 mice. At day 5, viable B cells were identified and percentage of activated Germinal Center (GC)-like B cells was determined by co-expression of GL7 and Fas, antibody-producing plasma cells (PC) by expression of CD138. Data on B cell count and viability are representative of one out of two experiments involving six replicates per condition. Data on GL7+ Fas+ GC-like B cells and CD138+ antibody-producing cells are pooled data from two independent experiments involving 1–7 replicates. Statistical significance of differences between CXCR3+ PD-1lo Tsh co-cultures and other conditions were assessed by Mann Whitney test.

To measure the relative helper functions of Tsh cells, we compared their ability to provide B cell help with that of other T cell subsets in vitro (Figure 7C—figure supplement 2). We sorted (CD4+ CD44+ PD-1hi) cells containing CXCR5+ Tfh helper cells with well-established helper functions (Vinuesa et al., 2016), Tsh (CD4+ CD44+ CXCR5- PD-1low CXCR3+), and as control population also the CD4+ CD44+ CXCR5- PD-1low CXCR3- subset, and co-cultured them with B cells in vitro for 5 days (Figure 7—figure supplement 2A–B). To obtain enough T cells, in particular PD-1hi cells that represent a minute population in NZBxW mice (Figure 3C), we resorted to old (15–24 months) C57BL/6 mice readily available and known to develop autoimmune features (Hayashi et al., 1989; Nusser et al., 2014). We had found that these mice also present with increased numbers of Tsh cells co-producing high levels of IFN-γ, IL-21 and CD40L (Figure 7—figure supplement 2C). At the endpoint of the co-culture, we analyzed the B cells for their numbers, viability and phenotype by flow cytometry (Figure 7—figure supplement 2D). Strikingly, PD-1low CXCR3+ Tsh cells promoted larger expansion of B cells than any other conditions, even outnumbering the positive control (IL-21 and anti-CD40) and PD-1hi cells by a factor of two and three, respectively (Figure 7C—figure supplement 2D). However, PD-1hi cells were clearly superior in preserving more viable B cells than the other conditions (Figure 7C—figure supplement 2D). The phenotype of B cells cultured with Tsh cells was in stark contrast compared to that of B cells cultured with PD-1hi cells. While the proportion of CD138+ antibody-producing B cells was higher in the co-cultures with PD-1hi cells (20% against 2–6% in other cultures), Tsh cells induced 33% of the B cells to express the activation markers GL7 and Fas, typically used to define germinal center (GC) B cells, against 1.5% of B cells with a GC-like phenotype in PD-1hi co-cultures (Figure 7C—figure supplement 2D). The proportion of GC-like B cells recovered with Tsh cells was similar to that found with PD-1low CXCR3- cells and the positive control (Figure 7C—figure supplement 2D). Nevertheless, given the superiority of Tsh cells to promote B cell expansion, this experiment indicates that Tsh cells have highly potent helper capacity and suggests complementary functions to that of PD-1hi and Tfh cells during B cell differentiation.

All this suggests that the IL-21 producing super-functional Th cells (IL-21+ IFN-γhi PD-1low Tmem cells) represent a novel Tfh-like subpopulation that can provide B-cell help in lupus nephritis mice in an IL-21- and CD40L-dependent manner.

Discussion

For many autoimmune diseases, the underlying cellular mechanisms are still incompletely understood. To develop novel targeted therapies, it is important to identify potentially pathogenic immune cell subsets. To this end, appropriate methods for the analysis and visualization of high-dimensional cytometric data are crucial to document global changes in the immune cell pattern during the course of the disease. Our study has generated three important findings: it showed the added value of our PRI-visualization, it identified a super-functional Tfh-like cell subpopulation and it indicated its potential functional effect on SLE.

First, PRI enabled the identification and characterization of the super-functional IL-21+ IFN-γhi cell subset by its main functionalities: (i) simple visualization of the combinatorial properties of three markers and heat map representation of different statistical attributes of the third marker by binning conventional two-parametric dot plots (e.g. Figure 1E, Figure 6E); (ii) semi-continuous, combinatorial display that allows straightforward perception of correlating patterns (e.g. Figure 3D–E); (iii) reproducible pseudo-multi-parametric display of different markers so that subpopulations can be intuitively observed and compared, even between samples of different staining panels, which is not applicable with cluster algorithms (e.g. Figure 4A and BQiu et al., 2011Shen et al., 2018) and (iv) exploitation of auxiliary information in percentages is enabled by the combinatorial approach, which would otherwise only be accessible with further gating steps. It should also be noted, however, that with PRI many cells and a broad bin range together with the right marker combination are necessary to obtain meaningful information from the plots.

PRI paves the way for reproducible and automatable cytometric analyses, which are useful not only for basic research but also for clinical studies to identify disease mechanisms and progressions, to discover biomarkers, to analyze therapy responder/non-responder, and to monitor stable quality and effectiveness of manipulated cells. In addition, PRI provides a statistically robust pattern map for visual inspections and serves as input for prospective pattern recognition in machine learning approaches with feature engineering steps.

Second, the super-functional IL-21+IFN-γhi cell subset identified here is a novel Tfh-like subpopulation, which differs from the subpopulations described so far by a low expression of PD-1. In concordance with Tfh, Tph (Rao et al., 2017) and other Tfh-like cell subsets (Bubier et al., 2009; Hutloff, 2018; Vu Van et al., 2016; Weinstein et al., 2016) splenic super-functional T cells are IL-21+, CD40Lhi, and ICOS+ (Figure 4—figure supplement 1). Their affiliation to Tfh or extrafollicular Tfh cells could be ruled out, as the majority of super-functional T cells do neither express Bcl6 or CXCR5, nor CXCR4 or PSGL-1low (Figure 4Odegard et al., 2008). They rather seem to exhibit polyfunctional characteristics such as a strong co-expression of Th1 and Tfh-like effector proteins with high activation potential.

Poly-functional T cells were first described as being protective in viral infections (Kannanganat et al., 2007; Seder et al., 2008). However, polyfunctional T cell subpopulations were identified as pathogenic in autoimmune diseases (Basdeo et al., 2015). Most of the previous studies on poly-functional T cells focused on IFN-γhi, TNF-αhi, IL-2hi cells in the context of infections and immunizations (Mateus et al., 2019; Sauzullo et al., 2014; Westerhof et al., 2019). Compared to the triple positive poly-functional T cells, super-functional T cells here were even superior regarding their additional expression of IL-21, ICOS, and CD40Lhi and their demonstrated B cell helper activity (Figure 7). In vitro, PD-1low CXCR3+ CXCR5- Tsh cells and PD-1hi cells, containing the CXCR5+ Tfh cells, displayed dichotomic functions regarding the induction of GC-like and plasma cell phenotypes, respectively. In keeping with the chemokine profile of these two CD4 T cell subsets, this result suggests that in vivo PD-1low CXCR3+ CXCR5- Tsh cells could provide help in the inter-follicular space juxtaposed to the GC and enriched in the CXCR3 ligand CXCL9 (Matloubian and Cyster, 2012). There, B cells become blasts, proliferate and differentiate into GC B cells that migrate to CXCR5 ligand-rich follicles, where further support by the Tfh promotes affinity maturation and plasma cells differentiation (Haynes et al., 2007; Vinuesa et al., 2016). Alternatively, Tsh could participate to expand activated B cells in an extrafollicular GC-like reaction, as recently described during Salmonella infections (Di Niro et al., 2015).

Considering the observed extrafollicular co-localization of T cells and B cells in the spleen of NZBxW mice (data not presented) and the peripheral localization of super-functional T cells (Figure 5D), we speculate that they can even provide B cell help directly in inflamed tissues, thereby exacerbating tissue destruction. Due to their low expression of inhibitory receptors (especially PD-1, CTLA4), their activation is potentially facilitated compared to Tfh cells, as these express significantly higher amounts of PD-1, CTLA4, TIGIT and BTLA (Figure 4Butte et al., 2007; Walunas et al., 1994).

Third, the super-functional IL-21+IFN-γhi cell subset could promote SLE progression through B cell help dependent and independent effects. Similar to Tfh-like cells, which were found in SLE and other autoimmune diseases (Horiuchi and Ueno, 2018; Hutloff, 2018; Rao, 2018; Vu Van et al., 2016), the super-functional subset delivers effective B-cell help for IgG production in an IL-21- and CD40L-dependent manner. Tissue-resident Tfh-like cells are proposed to be 'the most pathogenic T cell subset, as they select autoreactive B cells in the uncontrolled environment of lymphocytic tissue infiltrates and drive the local differentiation of plasmablasts producing pathogenic antibodies directly in the affected tissues' (Hutloff, 2018).

Due to their presence also in non-lymphoid organs (Figure 5) and their IL-21 and IFN-γ co-expression, the super-functional cell subset might be superior in activation-induced processes that drive autoimmunity by expanding pathogenic T cell subsets in non-lymphoid tissue and promoting inflammation. This hypothesis is supported by several studies focusing either on IL-21 or IFN- γ effects. 

IL-21 uniquely contributes to SLE and other autoimmune diseases, because pharmacological and genetic abrogation of IL-21 signaling in mice stopped inflammation in non-lymphoid organs and protected them from autoimmune diseases, respectively (Choi et al., 2017; Liu and King, 2013). In line with the reported IL-21–mediated proliferation (Zeng et al., 2007) and its autocrine and paracrine action (Nurieva et al., 2007) is our observation that the IL-21+ IFN-γ+ double-producing population had the highest proliferative capacity in vivo compared to single-producing (IL-21+ IFN-γ, IL-21 IFN-γ+) and double-negative (IL-21 IFN-γ) populations measured by Ki67 expression (data not shown).

IFN- γ is also considered to be a key molecule in the pathogenesis of SLE (Peng et al., 2002; Pollard et al., 2013; Tsokos et al., 2016). Consequently, excessive IFN-γ production was required to sustain lupus-associated Tfh cell accumulation in mice (Lee et al., 2012). Contrary to this and other reports on lupus-prone mice (Enghard et al., 2006; Humrich et al., 2010), we observed no increase but a slight decrease in IFN-γ frequencies with disease progression (Figure 2), possibly due to different subpopulations of Tmem and Th cells being studied.

In addition to IL-21 and IFN-γhi co-expression, the super-functional cell subset is characterized by high IL-2 and TNF-α but no IL-10 expression. IL-2 is discussed to be rather protective than pathogenic in SLE because the acquired IL-2 deficiency is a crucial general event in the pathogenesis of lupus and mainly leads to a disorder of the homeostasis of regulatory T cells. Correspondingly, a low dose IL-2 therapy selectively corrected Treg cell defects and greatly expanded the Treg cell subpopulation (Humrich et al., 2015; von Spee-Mayer et al., 2016). In contrast, anti-TNF-α therapy with infliximab has produced inconsistent results so far and is not routinely used to treat SLE (Aringer and Smolen, 2012). IL-10 has pleiotropic effects, which can be anti-inflammatory and pro-inflammatory. In SLE patients, however, the higher IL-10 level associated with the disease is considered pathogenic and its blockage improves the disease (Llorente et al., 2000). Recently, an IL-10/IFN-γ co-producing pathogenic Th cell sub-population (CXCR5- CXCR3+ PD-1hi) was identified in SLE patients (Caielli et al., 2019). These cells differ from our super-functional cells primarily because they do not co-express IL-21, rather they provide B cell help independently of IL-21, and are PD-1 high.

Transient co-expression of IL-21 and IFN-γ has already been described for subsets of early Th1-Tfh and germinal centre Tfh cells (Fang et al., 2018; Nakayamada et al., 2011; Weinstein et al., 2016). However, it is still unclear to what extent double producing cells functionally differ from IFN-γ and IL-21 single producers (Song and Craft, 2019). It is worth mentioning in this context that the majority of IL-21+ Tmem cells in this study were IFN-γ+ (Figure 3F; young: 56 ± 3% versus old diseased: 71 ± 4%) and non-Tfh cells (Figure 6). They were found in both lymphoid and peripheral organs of NZBxW mice (Figure 5).

Interestingly, IL-21 was reported to inhibit TCR-induced Th1 differentiation (Kastirr et al., 2014; Suto et al., 2006; Wurster et al., 2002) and, vice versa, forced expression of the Th1 master transcription factor T-bet inhibited IL-21 expression (Kastirr et al., 2014). Therefore, the special co-expression of IL-21 and IFN-γ could be triggered by chronic inflammatory conditions, leading to a dysregulated cytokine landscape (Humrich et al., 2010), specific expansion of double positive cells (Kastirr et al., 2014), and plasticity of Th cell phenotype (Carpio et al., 2015).

Several reports underline a high diversity and plasticity of Tfh and Tfh-like cells (Lüthje et al., 2012; Song and Craft, 2019; Wong et al., 2015). Song and Craft even proposed in between the two archetypes are differentiated cells that have varied degrees of Th and Tfh cell characteristics which may not exist in distinct subsets" (Song and Craft, 2019). They proposed to search for alternatives to further dividing Th, Tfh and Tfh-like cell subsets. Our PRI approach is an attempt in this direction to analyze and visualize a continuum between cell subsets without further dividing subpopulations by gating (Figure 4, Figure 3—figure supplement 1D).

In this study, we defined a novel subset of Tfh-like cells – super-functional IL-21+ IFN-γhi PD-1low Tmem cells – that exhibit superior production of cytokines in both quantity and number and provide B cell help. This suggests that efforts to decrease their frequency or manipulate their main properties such as IL-21/IFN-γ co-expression may represent an important therapeutic strategy.

Materials and methods

Mice

NZBxNZW (NZBxW) mice were bred under specific-pathogen-free conditions in the animal facility of the Federal Institute for Risk Assessment (Berlin, Germany). The female filial one generation was used for experiments. Animal experiments were approved by the local authority LAGeSo (Landesamt für Gesundheit und Soziales) Berlin under animal experiment licenses T0187-01 and G0070/13.

Severity of disease was monitored weekly by scoring the mice according to levels of proteinuria (Uristix, Siemens) and weight.

C57Bl6/J mice were purchased from Janvier (France) and kept for 15 to 24 months in the SPF facility of the DRFZ.

Cell preparations for flow cytometry

Request a detailed protocol

We isolated splenocytes by crushing the spleen through a 200 µm metal mesh in order to create single-cell suspensions. Erythrocytes were lysed by hypotonic shock and splenocytes were filtered through a 30 µm cell strainer.

Kidney, liver and lung were perfused with PBS. From the liver, single-cell suspensions were created by crushing through a 200 µm metal mesh. To isolate hepatic lymphoid cells, a density gradient centrifugation in the presence of 40% Percoll was performed followed by erythrocyte lysis.

Lungs and kidneys were perfused with PBS and placed in ice-cold PBS. Organs were cut into small pieces and digested with 0.25 mg/ml collagenase (Sigma) and 0.25 mg/ml collagenase D (Roche) and 10 U/ml DNAse I (Sigma) in RPMI 1640 medium + 0.5% BSA. The tissue was gently crushed through a 200 µm metal mesh followed by a 70 µm cell strainer. Cells were washed and resuspended in RPMI1640 supplemented with 10% FCS, b-mercaptoethanol, penicillin (100 U/ml) and streptomycin (100 U/ml).

Flow cytometry

Request a detailed protocol

The analysis was conducted according to the guidelines for the use of flow cytometry and cell sorting in immunological studies (Cossarizza et al., 2017).

For intracellular cytokine staining, cells were stimulated with 10 ng/ml phorbol 12-myristate 13-acetate (PMA) and 1 µg/ml ionomycin (Sigma) for 1 hr followed by additional 3 hr in the presence of 5 µg/ml Brefeldin A (Sigma). Cell surfaces were stained with antibodies in the presence of 100 µg/ml 2.4G2 (anti-FcγRII/III; purified from hybridoma supernatants) to reduce unspecific antibody binding. Cells were fixed with 2% paraformaldehyde followed by an intracellular staining in 0.5% saponin. For intracellular staining of transcription factors, cells were fixed with FoxP3 fixation buffer (eBioscience) and stained in FoxP3 permeabilization buffer (eBioscience). Live cells were discriminated from dead cells with a fixable LIVE/DEAD stain (Life Technologies). The expression of phenotypic markers was determined with a BD LSR Fortessa (BD Biosciences) and analyzed with FlowJo (Treestar).

Detection of serum antibodies by enzyme-linked immunosorbent assay (ELISA)

Request a detailed protocol

Serum was collected from mice of different age. Anti-dsDNA-IgG antibodies were measured by ELISA as described previously using biotin-labeled goat anti-mouse IgG (γ chain specific) antibodies for detection (Southern Biotech) (Cheng et al., 2013).

Co-culture experiments

Request a detailed protocol

Splenocytes were stained with anti-CD19-FITC, anti-CD25-PerCP-Cy5.5, anti-CXCR5-PE-Cy7, anti-CXCR3-APC, anti-B220-APC-Cy7, anti-CD4-PacB, anti-CD44-BV785, DAPI and sorted for DAPI naive B cells (CD19+ B220+ CD44low) and Th1 cells (CD4+ CXCR5 CD44+ CD25 CXCR3+) using FACS Aria II (BD Biosciences) with a purity ≥95%. Th1 cells and naive B cells (2:1) were sterily co-cultured for five days in 96-well round-bottom plates in the presence or absence of blocking antibodies: IL-21R-Fc chimera (R and D systems), anti-CD40L (clone MR1, DRFZ) and anti-IFN-γ (clone AN18.17.24, DRFZ). T cells were stimulated polyclonally with anti-CD3 (clone KT3) and anti-CD28 antibodies (clone 37.51, both DRFZ). As positive control, B cells were stimulated with IL-21 (Peprotech) and anti-CD40 (clone FGK-45, DRFZ). After 5 days, the IgG concentration in the supernatant was determined by the Mouse IgG ELISA Kit (LSBio).

T cell populations for T helper functions’ comparison were FACS-sorted from splenocytes stained with anti-PD-1 conjugated to PercP-Cy5.5 or APC-Cy7, anti-CXCR5-BV650, anti-CD44 conjugated with PacO or BV786, anti-CXCR3 conjugated with APC or BV421, anti-CD4-PE, anti-CD8-Al488, DAPI or PI. B cells were sorted from splenocytes after enrichment with B cell purification kit using magnetic beads (Miltenyi, Germany), and stained with anti-B220-FITC, anti-CD19-Al647 and PI. PI- CD19+ B220+ B cells, live CD4+ PD-1lo CXCR5- CXCR3+ or CXCR3- and PD-1hi CXCR5+/- were sorted using FACS Aria II (BD Biosciences) with a purity ≥90%. T and B cells were cultured as above. On day five, cells were harvested, washed in PBS, and stained with L/D aqua fixable dye (ThermoFisherScientific) and with anti-B220-FITC, anti-CD4-PE, anti-GL7-PerCP-Cy5.5, anti-Fas-PE-Cy7, anti-CD138-BV421. For each well, the totality of the cells were acquired on BD LSR Fortessa (BD Biosciences) and analyzed with FlowJo (Treestar).

Flow cytometry analysis

View detailed protocol

Files were pre-processed in FlowJo (compensation, elimination of doublets and dead cells, gating). For tSNE, CD44+ T cells were down-sampled to 10.000 cells and analyzed with the plugin ‘Tsne’ (1000 iterations, perplexity 20, theta 0.5, eta 200) in FlowJo. For PRI, T cell populations were exported as fcs-files. Fluorescent intensities were transformed with inverse hyperbolic sine (arcsinh), which is a common transformation method for flow cytometry data (Finak et al., 2010). The x and y axis of parameters X vs. Y was categorized into bins of size 0.2 × 0.2. Over all cells in each bin, different statistical methods are calculated and plotted in a color-coded manner as a heat map (low values are represented by shades of blue, median values by yellow and high values by red). The frequency (%) of parameter Z+ cells per bin, mean fluorescence intensity of parameter Z+ cells (MFI+), MFI over all cells per bin or the relative standard error of the mean of parameter Z (RSEM) were used. Auxiliary information are displayed as black, red, green and blue percentage numbers in each quadrant. Black percentage numbers indicate the frequency of cells relative to total cells. The frequency of Z+ cells is given as relative to cells inside the respective quadrant (red), relative to total cells (green) and relative to total Z+ cells (blue). Frequencies of double producing cells (two parameters) per bin are color-coded in green scales. All parameters were included in the analysis and bins with less than 10 cells are not displayed. This minimum count of cells was used to balance the impact of identifying comparatively rare subpopulations while retaining the statistical robustness. We operated on R with package ‘flowCore’ to read the fcs-files. In order to generate 3D-surface plots with package plotly v4.7.1, frequency bin tables were exported from PRI’s R source code.

Quantification and statistical analysis

Request a detailed protocol

Statistical analyses were conducted using GraphPad Prism7 Software. For bar and scatter plots, data are shown as mean ± s.e.m. The statistical tests included unpaired, two-tailed Student’s t-test and Mann-Whitney test. For comparisons of three or more groups, data were subjected to one-way or two-way ANOVA, followed by Sidak’s or Dunnett’s multiple comparison test (as indicated in each figure legend). p<0.05 was considered significant.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    Guidelines for the use of flow cytometry and cell sorting in immunological studies
    1. A Cossarizza
    2. HD Chang
    3. A Radbruch
    4. M Akdis
    5. I Andrä
    6. F Annunziato
    7. P Bacher
    8. V Barnaba
    9. L Battistini
    10. WM Bauer
    11. S Baumgart
    12. B Becher
    13. W Beisker
    14. C Berek
    15. A Blanco
    16. G Borsellino
    17. PE Boulais
    18. RR Brinkman
    19. M Büscher
    20. DH Busch
    21. TP Bushnell
    22. X Cao
    23. A Cavani
    24. PK Chattopadhyay
    25. Q Cheng
    26. S Chow
    27. M Clerici
    28. A Cooke
    29. A Cosma
    30. L Cosmi
    31. A Cumano
    32. VD Dang
    33. D Davies
    34. S De Biasi
    35. G Del Zotto
    36. S Della Bella
    37. P Dellabona
    38. G Deniz
    39. M Dessing
    40. A Diefenbach
    41. J Di Santo
    42. F Dieli
    43. A Dolf
    44. VS Donnenberg
    45. T Dörner
    46. GRA Ehrhardt
    47. E Endl
    48. P Engel
    49. B Engelhardt
    50. C Esser
    51. B Everts
    52. A Dreher
    53. CS Falk
    54. TA Fehniger
    55. A Filby
    56. S Fillatreau
    57. M Follo
    58. I Förster
    59. J Foster
    60. GA Foulds
    61. PS Frenette
    62. D Galbraith
    63. N Garbi
    64. MD García-Godoy
    65. J Geginat
    66. K Ghoreschi
    67. L Gibellini
    68. C Goettlinger
    69. CS Goodyear
    70. A Gori
    71. J Grogan
    72. M Gross
    73. A Grützkau
    74. D Grummitt
    75. J Hahn
    76. Q Hammer
    77. AE Hauser
    78. DL Haviland
    79. D Hedley
    80. G Herrera
    81. M Herrmann
    82. F Hiepe
    83. T Holland
    84. P Hombrink
    85. JP Houston
    86. BF Hoyer
    87. B Huang
    88. CA Hunter
    89. A Iannone
    90. HM Jäck
    91. B Jávega
    92. S Jonjic
    93. K Juelke
    94. S Jung
    95. T Kaiser
    96. T Kalina
    97. B Keller
    98. S Khan
    99. D Kienhöfer
    100. T Kroneis
    101. D Kunkel
    102. C Kurts
    103. P Kvistborg
    104. J Lannigan
    105. O Lantz
    106. A Larbi
    107. S LeibundGut-Landmann
    108. MD Leipold
    109. MK Levings
    110. V Litwin
    111. Y Liu
    112. M Lohoff
    113. G Lombardi
    114. L Lopez
    115. A Lovett-Racke
    116. E Lubberts
    117. B Ludewig
    118. E Lugli
    119. HT Maecker
    120. G Martrus
    121. G Matarese
    122. C Maueröder
    123. M McGrath
    124. I McInnes
    125. HE Mei
    126. F Melchers
    127. S Melzer
    128. D Mielenz
    129. K Mills
    130. D Mirrer
    131. J Mjösberg
    132. J Moore
    133. B Moran
    134. A Moretta
    135. L Moretta
    136. TR Mosmann
    137. S Müller
    138. W Müller
    139. C Münz
    140. G Multhoff
    141. LE Munoz
    142. KM Murphy
    143. T Nakayama
    144. M Nasi
    145. C Neudörfl
    146. J Nolan
    147. S Nourshargh
    148. JE O'Connor
    149. W Ouyang
    150. A Oxenius
    151. R Palankar
    152. I Panse
    153. P Peterson
    154. C Peth
    155. J Petriz
    156. D Philips
    157. W Pickl
    158. S Piconese
    159. M Pinti
    160. AG Pockley
    161. MJ Podolska
    162. C Pucillo
    163. SA Quataert
    164. T Radstake
    165. B Rajwa
    166. JA Rebhahn
    167. D Recktenwald
    168. EBM Remmerswaal
    169. K Rezvani
    170. LG Rico
    171. JP Robinson
    172. C Romagnani
    173. A Rubartelli
    174. B Ruckert
    175. J Ruland
    176. S Sakaguchi
    177. F Sala-de-Oyanguren
    178. Y Samstag
    179. S Sanderson
    180. B Sawitzki
    181. A Scheffold
    182. M Schiemann
    183. F Schildberg
    184. E Schimisky
    185. SA Schmid
    186. S Schmitt
    187. K Schober
    188. T Schüler
    189. AR Schulz
    190. T Schumacher
    191. C Scotta
    192. TV Shankey
    193. A Shemer
    194. AK Simon
    195. J Spidlen
    196. AM Stall
    197. R Stark
    198. C Stehle
    199. M Stein
    200. T Steinmetz
    201. H Stockinger
    202. Y Takahama
    203. A Tarnok
    204. Z Tian
    205. G Toldi
    206. J Tornack
    207. E Traggiai
    208. J Trotter
    209. H Ulrich
    210. M van der Braber
    211. RAW van Lier
    212. M Veldhoen
    213. S Vento-Asturias
    214. P Vieira
    215. D Voehringer
    216. HD Volk
    217. K von Volkmann
    218. A Waisman
    219. R Walker
    220. MD Ward
    221. K Warnatz
    222. S Warth
    223. JV Watson
    224. C Watzl
    225. L Wegener
    226. A Wiedemann
    227. J Wienands
    228. G Willimsky
    229. J Wing
    230. P Wurst
    231. L Yu
    232. A Yue
    233. Q Zhang
    234. Y Zhao
    235. S Ziegler
    236. J Zimmermann
    (2017)
    European Journal of Immunology 47:1584–1797.
    https://doi.org/10.1002/eji.201646632
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
    Spontaneous development of organ-specific autoimmune lesions in aged C57BL/6 mice
    1. Y Hayashi
    2. M Utsuyama
    3. C Kurashima
    4. K Hirokawa
    (1989)
    Clinical and Experimental Immunology 78:120–126.
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
    Interferon-gamma and systemic autoimmunity
    1. KM Pollard
    2. DM Cauvi
    3. CB Toomey
    4. KV Morris
    5. DH Kono
    (2013)
    Discovery Medicine 16:123–131.
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79

Decision letter

  1. Bernard Malissen
    Reviewing Editor; Centre d'Immunologie de Marseille-Luminy, Aix Marseille Université, France
  2. Satyajit Rath
    Senior Editor; Indian Institute of Science Education and Research (IISER), India
  3. Joe Craft
    Reviewer; Yale University School of Medicine, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Some T helper cell populations have been shown to expand in autoimmune diseases such as systemic lupus erythematosus and might thus contribute to the unfolding of the disease condition. To zoom on those T cells that have phenotype very similar to other T cell subpopulations, the authors combined a bin-based "pattern recognition of immune cells" strategy with comprehensive cytometric measurements of T helper cells from found in lupus-prone mice. On that basis, they defined a novel subset of super functional T helper cells that exhibit superior B cell help and may constitute therapeutic targets.

Decision letter after peer review:

Thank you for submitting your article "Identification of a super-functional Tfh-like subpopulation in murine lupus by pattern perception" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Satyajit Rath as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Joe Craft (Reviewer #1).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The authors used CytoBinning, a recently described flow cytometry analysis approach, to explore the phenotype and cytokine expression profile of splenic memory CD4+ T cells from NZBxW mice at different stages of disease. The analysis approach combines automation of a traditional workflow and machine learning, linking high dimensional data back to two biomarkers which can be represented as 2D scatter plots and enables the identification of subtle shifts in immune phenotypes. Using this approach, the authors identify a subpopulation of IL21+ IFNγ+ PD1low CD40Lhigh CXCR5- Bcl6- CD4+ T cells expanded in older, diseased mice that expresses high levels of TNFα and IL2. Considering that in co-culture experiments they observe that it provides B cell help though IL21 and CD40L, the authors define this population as a "super-functional Tfh-like cells". Overall, this interesting manuscript surmises that these newly identified cells are critical for the disease phenotype in lupus, including promotion of autoantibody production.

As a major aim of the manuscript is the presentation of a new method to analyse flow cytometry data, further information should be made available to allow a direct comparison of this new method with previous methods. Moreover, the "super-functional Tfh-like cells" should be more precisely compared to the T peripheral helper (Tph) cells, first described in rheumatoid arthritis by Rao and colleagues (Rao et al., 2017), and this year in SLE (Bocharnikov et al., 2019), celiac disease (Christophersen et al., 2019) and T1DM (Ekman et al., 2019).

Essential revisions:

1) Methodological issues.

1. 1) While the insights revealed using the combinatorial approach with data binning are of interest, sifting through the data is more challenging than it need be. Certain panels of figures are mislabeled (2H does not exist, 3C or D is incorrectly noted, and so forth), and the authors are asked to proof the text and figures. As well, Figure 2E and Figure 2—figure supplement 1B are nigh on impossible to parse, given overlap in colors (not even accounting for red-green discernment). In a like vein, certain figures are challenging to interpret, given the light shades of gray that appear to overlap. The concerns stand out all the more, given the nice visuals of the binning data.

1.2) Each PRI plot superimposes data from several samples (biological replicates). While the authors show that data are reproducible for individual mice, it is likely that for some markers or some biological experiments, great inter-individual variation can occur. The provided statistics fail to represent this issue. A better clarification on how PRI plots can be used to show the range of variability among a population should be provided.

1.3) The audience of this manuscript will be far more familiar with traditional methods of FACS analysis than PRI. As a consequence, in order to establish the robustness of PRI-based quantification and familiarity with the approach, Figure 3A should include also samples from young mice. It seems from the analysis of the flow cytometry plots of old mice (Figure 3A) that all IL-21+ cells co-express IL-2, TNF, and IFN while being CXCR5-, PD-1int, and IL-10-. However, PRI assessment of the same samples (Figure 3F) provides a visual representation suggesting that the vast majority of IFNγ+ cells are also producing IL-21, while in the dot plots (Figure 3A) only ~1/7 of IFNγ+ cells produce IL-21. Furthermore, looking at the PRI of young mice it is unclear whether the IL-21 producing cells maintain the same phenotype as the cells from old mice, like the authors claim (subsection “Bin plots revealed an IL-21+ super-functional T cell subpopulation”). For all of the above Figure 3A should allow independent assessment of young/old phenotype in addition to PRI plots of Figure 3F.

1.4) Additionally, the most biologically relevant statistical assessment is in respect to cell populations (defined to the best of our ability) rather than individual markers. If the authors claim the importance of a novel cell subset defined as IL-21+, IFNγ+… the relevant statistical analysis of Figure 3E should be based on the frequency of that cell subset in young vs. old mice rather than (or in addition to) the presented analysis based on individual markers.

1.5) The phenotype of IL-21-producing cells displayed in Figure 4 is not easy to interpret given the poor identification of the IL21+ cells, as they are contained in bins together with other cells. Lack of statistics is also a concern. As a key objective of the manuscript is to show how PRI compares with FACS analysis, the figure would improve with traditional dot plots of gated IL-21+ cells (or even how IL21+IFNγ++ compared with IL-21-IFN++ for the different markers – to a large extent these two cell populations are in the same bin region). Statistics could then be provided and compared with PRI data.

1.6) It is also puzzling why the authors do not compare the phenotype of IL-21+ cells and IL-21-IFN+ and IL21-IFN- instead of a complex matrix of comparisons based on IFNγ and PD-1 expression (Figures 5, 6 and Figure 6—figure supplement 1). A major goal of the manuscript is to define a IL-21+IFNγ+ cell population that can be easily gated. Data should be provided for the population of interest in addition to arbitrary quadrants.

1.7) Figure 2C. How was the determination of PD-1-, PD-1low and PD-1 positivity determined? The gating does not follow the cell density plots.

1.8) The authors note the data are reproducible. but how was reproducibility shown statistically?

2) Biological issues

2.1) The authors focused only on the "super-functional Tfh-like cells", even though in the sicker mice there is not only and expansion of IL21+ IFNγ+ PD1low cells but also of IL21+ IFNγ+ PD1high cells (Figure 3E) as well IL10+ IFNγ+ PD1high cells. To prove that this IL21+ INFγ+ PD1low CD40Lhigh CXCR5- Bcl6- T cell population helps B cells though IL21 and CD40L, they sort non Tfh-Th1 like cells (CD4+, CXCR5-, CD44hi, CXCR3+ cells) and check their ability to support B cell activation and Ig production. Two main concerns arise from these set of experiments. First, the sorting strategy should include PD1 in order to enrich the non Tfh-Th1 like mem T cells (CD4+, CXCR5-, CD44hi, CXCR3+) within the PD1low population (CD4+, PD1low, CXCR5-, CD44hi, CXCR3+ cells). Second, the B cell helper assay should be repeated and compared with other cell populations, including conventional Tfh1 cells (CD4+, PD1hi, CXCR5+, CD44hi, CXCR3+ cells) and include naïve versus total B cells. In addition, the authors should perform a more thorough characterization of the main three expanded cell subpopulations by RNAseq to better define the uniqueness of the novel "super- functional" cells.

2.2) The authors claim IFNγ+IL-21-producing cells represent a novel population distinct from previous defined populations with "super-functional" properties.

2.2.1) One cell population that has been shown important in human SLE and bears some similarities was not discussed: T peripheral helper (Tph) cells, first described in rheumatoid arthritis by Rao and colleagues (Rao et al., 2017), and this year in SLE (Bocharnikov et al., 2019), in celiac disease (Christophersen et al., 2019) and T1DM (Ekman et al., 2019). These cells are CXCR5- Bcl6- and characterized by IL-21 production, a critical difference to the population discussed in this manuscript is that Tph are PD-1hi. However, if the two populations are different, they should both be present. Do the authors claim the only T cell population producing IL-21 is the newly discovered? Is there an explanation for poor identification of other IL-21-producing T cells among CXCR5- cells, namely Tph?

2.2.2) Considering that the super-functional T cell population identified by the authors is similar to Tph cells, it would be of interest to know if the super-functional T cells are Blimp1+, which is likely the case, recognizing the potential limitation of flow staining for this transcriptional repressor in mouse.

2.2.3) While the T-B cell helper assay adds a functional component to the paper, how physiological this experiment really is? Do the authors think that their T cells drive "naïve" B cells to produce Ig in vivo? Do super-functional T cells come into contact with naïve B cells in secondary lymphoid organs or elsewhere? Wouldn't naïve B cells need an IgM as well as TLR and/or cytokine signals first, to move into proximity with activated T cells?

2.2.4) The functional assay cannot be easily interpreted with the available information. Since sorting was done with an indirect strategy (based on CXCR3+CXCR5- T cells) it will be important to show what is the frequency of the IL-21+IFNγ+PD-1int cells among the sorted population. A positive control with bona fide CXCR5+PD1hi Tfh cells should be provided to compare the potency of Tfh and the claimed "super-functional" population. Titrations may be provided to assess relative potency, if needed. It is hard to claim super-functional properties without a functional comparison with the current standard of B cell help (i.e. Tfh cells). FACS plots with B cells stained with anti-GL7 and anti-IgG1 could also make the results more compelling, given the low levels of IgG antibodies produced.

2.2.5) Frequently, BCR agonists (anti-IgM for instance) are provided in addition to T cell stimulation for in vitro assays (see for instance Sage, Nature Immunol 2016 17:1436), this may improve the assay read-out.

2.2.6) More clarity about disease definition would be helpful. "Chronic inflammation, high proteinuria", is not so specific, not is age of mice. Likewise, were autoantibodies and renal histology assessed?

https://doi.org/10.7554/eLife.53226.sa1

Author response

Summary:

The authors used CytoBinning, a recently described flow cytometry analysis approach, to explore the phenotype and cytokine expression profile of splenic memory CD4+ T cells from NZBxW mice at different stages of disease. […] Overall, this interesting manuscript surmises that these newly identified cells are critical for the disease phenotype in lupus, including promotion of autoantibody production.

Binning is a way of discretizing data and can be done by adaptive binning (e.g. quantiles) or fixed-width binning (e.g. rounding). The mentioned CytoBinning approach, first described by Roederer et al., 2001 (Cytometry 2001, 45:37), use the percentile-based binning. In contrast, our approach uses fixed-width bins with a varying number of cells. This visualisation was chosen, because the cells tend to accumulate near the center of the data range but flatten out fast near the boundary of the range. A percentile-based binning would emphasize on the cell distribution. In contrast, we focus on the expression distribution and pattern mainly at the range boundaries. We do not only use the regular cell count, but we particularly make use of the mean fluorescence intensities and the frequency of positive cells of an additional third marker or several markers to provide valuable additional information. This is the combinatorics of 3 or more markers. Furthermore, a fixed-width binning allows for a good intuitive comparison of different samples.

As a major aim of the manuscript is the presentation of a new method to analyse flow cytometry data, further information should be made available to allow a direct comparison of this new method with previous methods.

We accepted the advice and integrated additional conventional analyses into several figures. In Figure 3A and B we included all respective contour plots for young mice. We confirmed the bin plot data of new Figure 3F and G by conventional gating analysis shown in Figure 3A and B, and the results of statistical analysis in Figure 3C. For comparison we show the statistical analysis from the bin plots in Figure 3—figure supplement 1B.

Moreover, the "super-functional Tfh-like cells" should be more precisely compared to the T peripheral helper (Tph) cells, first described in rheumatoid arthritis by Rao and colleagues (Rao et al., 2017), and this year in SLE (Bocharnikov et al., 2019), celiac disease (Christophersen et al., 2019) and T1DM (Ekman et al., 2019).

Thank you for the recommendation. Tfh and Tph cells are both characterized by very high PD-1 expression (which distinguishes them from the PD-1low super-functional Tfh-like (Tsh) cells), high production of IL-21, and either high or absent expression of CXCR5. We included new data and figures for these three markers (Figure 3C lower row, Figure 3G, Figure 3—figure supplement 1B and Figure 5D). In addition, we inserted additional references into the main text of the manuscript.

Essential revisions:

1) Methodological issues.

1. 1) While the insights revealed using the combinatorial approach with data binning are of interest, sifting through the data is more challenging than it need be. Certain panels of figures are mislabeled (2H does not exist, 3C or D is incorrectly noted, and so forth), and the authors are asked to proof the text and figures.

Sincerely sorry for the mistake. We corrected it.

As well, Figure 2E and Figure 2—figure supplement 1B are nigh on impossible to parse, given overlap in colors (not even accounting for red-green discernment). In a like vein, certain figures are challenging to interpret, given the light shades of gray that appear to overlap. The concerns stand out all the more, given the nice visuals of the binning data.

We totally see the point. We improved the colours and the graphics for both pie charts. For simplicity, we moved the pie chart (former Figure 2E) to Figure 2—figure supplement 2E).

1.2) Each PRI plot superimposes data from several samples (biological replicates). While the authors show that data are reproducible for individual mice, it is likely that for some markers or some biological experiments, great inter-individual variation can occur. The provided statistics fail to represent this issue. A better clarification on how PRI plots can be used to show the range of variability among a population should be provided.

In the original version of the manuscript, each bin plot visualised the cytometric data of a single mouse. To minimize inter-individual deviations, data from different mice were combined after measurement and were visualised in one bin plot. Therefore, for the new Figure 3 pictures we used concatenated cells from 3 mice each (young and old diseased). To exemplary demonstrate the robustness of the method, we compared the pattern of concatenated cells from the new experiments with patterns of 4 single mice from older experiments (Figure 2—figure supplement 3B). Thus we inserted the new Figure 2—figure supplement 3C.

1.3) The audience of this manuscript will be far more familiar with traditional methods of FACS analysis than PRI. As a consequence, in order to establish the robustness of PRI-based quantification and familiarity with the approach, Figure 3A should include also samples from young mice.

We included conventional plots of young mice into Figure 3A and B.

It seems from the analysis of the flow cytometry plots of old mice (Figure 3A) that all IL-21+ cells co-express IL-2, TNF, and IFN while being CXCR5-, PD-1int, and IL-10-. However, PRI assessment of the same samples (Figure 3F) provides a visual representation suggesting that the vast majority of IFNγ+ cells are also producing IL-21, while in the dot plots (Figure 3A) only ~1/7 of IFNγ+ cells produce IL-21.

We admit that too many frequencies given in the old Figure 3F were confusing. Therefore, we depicted here (Author response image 1) and in the novel Figure 3F and G only the most important green numbers (Z+ producers of all CD44+ cells in the respective quadrants).

The old figures showed that 5.7% (Author response image 1 left picture) and 6.8% (Author response image 1 right picture) of CD44+ cells are IL21+IFN-γ+ (DP) cells and 5.4% are triple-producers (with PD-1+). Most of the IFN-γ producing cells (32%) were IL-21- and PD-1+.

Author response image 1

Furthermore, looking at the PRI of young mice it is unclear whether the IL-21 producing cells maintain the same phenotype as the cells from old mice, like the authors claim (subsection “Bin plots revealed an IL-21+ super-functional T cell subpopulation”). For all of the above Figure 3A should allow independent assessment of young/old phenotype in addition to PRI plots of Figure 3F.

We included plots of young mice into Figure 3A and G.

The frequencies of IL-21+ and IL21+IFN-γ+ among the Tmem cells (CD44+) are not significantly different between young and old mice but show a slight tendency to be a bit higher in old diseased mice as shown both with PRI-data (Figure 3—figure supplement 1B) and FlowJo-data (Figure 3C). However, the properties of these cells to co-produce PD-1 increased significantly (Figure 3—figure supplement 1C).

1.4) Additionally, the most biologically relevant statistical assessment is in respect to cell populations (defined to the best of our ability) rather than individual markers. If the authors claim the importance of a novel cell subset defined as IL-21+, IFNγ+… the relevant statistical analysis of Figure 3E should be based on the frequency of that cell subset in young vs. old mice rather than (or in addition to) the presented analysis based on individual markers.

In fact, this consideration is really logical. Therefore, we provided two new experiments (with 10 mice in total) and included all relevant markers within one staining panel. This allowed us to analyse in parallel the subsets Tfh, Tph, and Tsh. The new Figure 3F, G, and Figure 3—figure supplement 1B show the PRI data of concatenated measured cells (3 mice each for young and old mice). The conventional flow cytometric analysis is shown in Figure 3C.

1.5) The phenotype of IL-21-producing cells displayed in Figure 4 is not easy to interpret given the poor identification of the IL21+ cells, as they are contained in bins together with other cells.

We agree to this. However, we experienced that taking IFN-γ instead of IL-21 is better for visualisation of preferred expression areas of the most Z-parameters. The range of IFN-γ is much greater than the range from IL-21. Therefore, tendencies and differences in the populations are better visualised as can be seen e.g. for CD40L intensities (Figure 4—figure supplement 1A) which correlates with the IFN-γ intensities.

Lack of statistics is also a concern. As a key objective of the manuscript is to show how PRI compares with FACS analysis, the figure would improve with traditional dot plots of gated IL-21+ cells (or even how IL21+IFNγ++ compared with IL-21-IFN++ for the different markers – to a large extent these two cell populations are in the same bin region). Statistics could then be provided and compared with PRI data.

The new Figure 3C and the new Figure 3—figure supplement 1B, offer the possibility of a direct comparison of the statistical data obtained by conventional gating and PRI analysis, respectively.

1.6) It is also puzzling why the authors do not compare the phenotype of IL-21+ cells and IL-21-IFN+ and IL21-IFN- instead of a complex matrix of comparisons based on IFNγ and PD-1 expression (Figures 5, 6 and Figure 6—figure supplement 1). A major goal of the manuscript is to define a IL-21+IFNγ+ cell population that can be easily gated. Data should be provided for the population of interest in addition to arbitrary quadrants.

The results from subsequent gating are shown in the new Figure 3C.

1.7) Figure 2C. How was the determination of PD-1-, PD-1low and PD-1 positivity determined? The gating does not follow the cell density plots.

For PD-1+/- gating we used the FMO controls. For determination of PD-1high, we took the decline in co-expression of PD-1high and IL-2 intensities in old diseased mice.

1.8) The authors note the data are reproducible. but how was reproducibility shown statistically?

Plotting the same data file produces the same pattern and numbers (100% identical) if all parameters, such as the defined threshold value of the parameter Z, the minimum number of cells per bin and the bin width are maintained. By applying different threshold values for parameter Z (IL-10) as shown in Figure 2—figure supplement 3D, the reviewer can get an impression about the changes of the pattern.

2) Biological issues

2.1) The authors focused only on the "super-functional Tfh-like cells", even though in the sicker mice there is not only and expansion of IL21+ IFNγ+ PD1low cells but also of IL21+ IFNγ+ PD1high cells (Figure 3E) as well IL10+ IFNγ+ PD1high cells.

Yes, the reviewer is right (seen also in the new Figure 3C), there are also more IL21+ IFNγ+ PD1high and IL10+ IFNγ+ PD1high cells in old mice compared to young mice. However, the majority of IL-21+ cells are PD-1low and IFNγ+ , which defines a previously unknown subset. Therefore, we concentrated on characterisation of this novel cell subset.

To prove that this IL21+ INFγ+ PD1low CD40Lhigh CXCR5- Bcl6- T cell population helps B cells though IL21 and CD40L, they sort non Tfh-Th1 like cells (CD4+, CXCR5-, CD44hi, CXCR3+ cells) and check their ability to support B cell activation and Ig production. Two main concerns arise from these set of experiments. First, the sorting strategy should include PD1 in order to enrich the non Tfh-Th1 like mem T cells (CD4+, CXCR5-, CD44hi, CXCR3+) within the PD1low population (CD4+, PD1low, CXCR5-, CD44hi, CXCR3+ cells). Second, the B cell helper assay should be repeated and compared with other cell populations, including conventional Tfh1 cells (CD4+, PD1hi, CXCR5+, CD44hi, CXCR3+ cells) and include naïve versus total B cells.

To address this concern, we conducted a new set of T/B co-culture assays with cells sorted according to their PD-1 expression.

We performed new T/B co-culture assays to compare directly the B helper abilities of Tsh cells defined as CD4+, PD1low CXCR5- CD44+ CXCR3+ cells, to that of CD4+PD1low CXCR5- CD44+ CXCR3- cells and PD-1high cells, defined as CD4+PD1high CXCR5+/- CD44+ containing both Tfh and Tph cells. These experiments are reported in Figure 7C and Figure 7—figure supplement 2. Due to the rarity of Tfh and Tph populations and the limited availability of old enough NZBxW F1 mice, we had to resort to aged (15-24 months) C57Bl/6 mice known to have increased autoimmune features (see for example Nusser et al., 2014). We had found that 24-month old C57Bl/6 mice presented with increased numbers of CD44+ cells and Tsh with the same features as sick NZBxW F1 mice. The strategy to sort the compared populations, as well as their respective productions of CD40L, IL-21 and IFN-g are presented in Figure 7—figure supplement 2A and C.

The limiting numbers of PD-1hi cells restricted our study to the use of total B cells defined as CD19+ B220+ cells (Figure 7—figure supplement 2B).

These experiments clearly show dichotomic roles for Tsh and PD-1high cells. While Tsh cells promote higher B cells expansion and GC-like differentiation of the B cells (GL7+ Fas+), PD-1high cells promote cell survival and sustains better plasma cell differentiation (Figure 7C and Figure 7—figure supplement 2D). These results showing the superiority of Tsh cells in providing help for B cell proliferation is now described and discussed in the manuscript.

In addition, the authors should perform a more thorough characterization of the main three expanded cell subpopulations by RNAseq to better define the uniqueness of the novel "super-functional" cells.

We agree that RNAseq data might reveal additional unique markers of this novel subset. Here, we rather concentrated on comprehensive protein pattern (in particular cytokine pattern) characterisation. We feel that this analysis already provided unique markers not only for definition of this subset but also unique functional properties.

2.2) The authors claim IFNγ+IL-21-producing cells represent a novel population distinct from previous defined populations with "super-functional" properties.

2.2.1) One cell population that has been shown important in human SLE and bears some similarities was not discussed: T peripheral helper (Tph) cells, first described in rheumatoid arthritis by Rao and colleagues (Rao et al., 2017), and this year in SLE (Bocharnikov et al., 2019), in celiac disease (Christophersen et al., 2019) and T1DM (Ekman et al., 2019). These cells are CXCR5- Bcl6- and characterized by IL-21 production, a critical difference to the population discussed in this manuscript is that Tph are PD-1hi. However, if the two populations are different, they should both be present. Do the authors claim the only T cell population producing IL-21 is the newly discovered? Is there an explanation for poor identification of other IL-21-producing T cells among CXCR5- cells, namely Tph?

The reviewer is right, both cell subsets Tph and Tsh are increased in aged NZBW mice and produce IL-21. To analyse Tph cells in parallel to Tfh and Tsh cells, we performed 2 new experiments which included staining for CXCR5 (seen in Figure 3). In addition we re-analysed data of Figure 5 to directly compare the frequencies of both IL-21-producing subsets. Tsh cells are more than twice the number as Tph cells in the NZBW model.

2.2.2) Considering that the super-functional T cell population identified by the authors is similar to Tph cells, it would be of interest to know if the super-functional T cells are Blimp1+, which is likely the case, recognizing the potential limitation of flow staining for this transcriptional repressor in mouse.

We tried Blimp1+ staining, but were not successful. An example is shown where we compared staining obtained for PD-1low CXCR5- IFN-g+ IL-21+ (Tsh) cells, PD-1high CXCR5+ (Tfh) cells and PD-1high CXCR5- cells. Compared to FMO staining, the antibody gave a similar signal for the entire population, including supposedly negative Tfh-cells, whereas T-bet staining as control gave a differential signal.

Author response image 2

2.2.3) While the T-B cell helper assay adds a functional component to the paper, how physiological this experiment really is? Do the authors think that their T cells drive "naïve" B cells to produce Ig in vivo? Do super-functional T cells come into contact with naïve B cells in secondary lymphoid organs or elsewhere? Wouldn't naïve B cells need an IgM as well as TLR and/or cytokine signals first, to move into proximity with activated T cells?

We agree with the reviewer. However, we would like to point out that the coculture experiments have now been repeated with total B cells and that it is quite impressive that Tsh cells can help naive B cells.

2.2.4) The functional assay cannot be easily interpreted with the available information. Since sorting was done with an indirect strategy (based on CXCR3+CXCR5- T cells) it will be important to show what is the frequency of the IL-21+IFNγ+PD-1int cells among the sorted population. A positive control with bona fide CXCR5+PD1hi Tfh cells should be provided to compare the potency of Tfh and the claimed "super-functional" population. Titrations may be provided to assess relative potency, if needed. It is hard to claim super-functional properties without a functional comparison with the current standard of B cell help (i.e. Tfh cells). FACS plots with B cells stained with anti-GL7 and anti-IgG1 could also make the results more compelling, given the low levels of IgG antibodies produced.

A new set of co-culture assays with a sorting strategy based on the expression of PD-1 and CXCR3 (Figure 7C and Figure 7—figure supplement 2) addresses the reviewer’s concern. This new test confirmed the high helper functions of PD-1low CXCR3+. Moreover, we characterized the respective cytokines and CD40L productions of the sorted CD4 T cell subsets to allow a direct correlation between their profiles and helper capacity (Figure 7—figure supplement 2C). These experiments provide a solid proof that Tsh cells have superior helper ability to support B cell proliferation over Tfh, suggesting a role in the production of self-reactive B cells in autoimmune contexts.

We took the advice of the reviewer and analysed the co-cultured B cells at the endpoint of the We took the advice of the reviewer and analyzed the co-cultured B cells at the endpoint of the experiment (day 5) by flow cytometry, using GL7 and Fas as marker of activated GC-like cells, and CD138 as an alternative to IgG1 for definition of antibody-producing cells. We preferred CD138 over IgG1 because IgG1 is mainly produced in Th2 contexts and Tph display a more Th1 phenotype with IFN-g, which induces diverse Ig class switching. Measuring IgG1 may have led to underestimate the production of Ig-producing cells. This new readout highlighted differences in the quality of T cell help provided by the different subsets, but also provided an explanation for the low levels of IgG detected by ELISA, given the low percentage of CD138+ cells.

2.2.5) Frequently, BCR agonists (anti-IgM for instance) are provided in addition to T cell stimulation for in vitro assays (see for instance Sage, Nature Immunol 2016 17:1436), this may improve the assay read-out.

Thank you for the advice. However, because we worked with limiting numbers of T cells, we could not endeavour to set an experimental system in which T help and BCR agonists are mutually titrated to optimize the co-culture assay. We plan to explore Tsh helper functions in antigen-dependent contexts involving BCR engagement in the future.

2.2.6) More clarity about disease definition would be helpful. "Chronic inflammation, high proteinuria", is not so specific, not is age of mice. Likewise, were autoantibodies and renal histology assessed?

Disease definition concerning age, proteinuria, and auto-antibodies is depicted in Figure 2A and in the main text. Auto-antibodies increase with age and disease score. Renal histology was not assessed.

https://doi.org/10.7554/eLife.53226.sa2

Article and author information

Author details

  1. Stefanie Gryzik

    German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft
    Contributed equally with
    Yen Hoang
    For correspondence
    s_gryzik@web.de
    Competing interests
    No competing interests declared
  2. Yen Hoang

    1. German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    2. University of Potsdam, Potsdam, Germany
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft
    Contributed equally with
    Stefanie Gryzik
    For correspondence
    yen.hoang@drfz.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8956-1709
  3. Timo Lischke

    German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    Contribution
    Conceptualization, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0413-4252
  4. Elodie Mohr

    German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    Contribution
    Formal analysis, Investigation, Methodology, Writing - review and editing
    For correspondence
    elodie.mohr@drfz.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3406-7302
  5. Melanie Venzke

    German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    Contribution
    Resources
    Competing interests
    No competing interests declared
  6. Isabelle Kadner

    1. German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    2. University of Potsdam, Potsdam, Germany
    Contribution
    Software
    Competing interests
    No competing interests declared
  7. Josephine Poetzsch

    1. German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    2. University of Potsdam, Potsdam, Germany
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4238-3866
  8. Detlef Groth

    University of Potsdam, Potsdam, Germany
    Contribution
    Methodology
    Competing interests
    No competing interests declared
  9. Andreas Radbruch

    1. German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    2. Charité, Campus Mitte, Berlin, Germany
    Contribution
    Funding acquisition
    Competing interests
    No competing interests declared
  10. Andreas Hutloff

    German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    Contribution
    Supervision, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0572-8151
  11. Ria Baumgrass

    1. German Rheumatism Research Center (DRFZ), A Leibniz Institute, Berlin, Germany
    2. University of Potsdam, Potsdam, Germany
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Writing - original draft, Project administration
    For correspondence
    baumgrass@drfz.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3289-1608

Funding

Bundesministerium für Bildung und Forschung (0316164A)

  • Ria Baumgrass

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank all members of the lab, and Tobias Alexander, Gabriele Riemekasten, and Jens Humrich for helpful comments and discussions. We thank Falk Hiepe and Manuela Frese-Schaper for providing NZBxW mice. We are grateful to Jenny Kirsch and Toralf Kaiser for operating the flow cytometry core facility (FCCF) and also to the staff of the animal facility and the lab managers.

Ethics

Animal experimentation: Animal experiments were approved by the local ethics committee LaGeSo (Landesamt für Gesundheit und Soziales) Berlin under animal experiment licenses T0187-01 and G0070/13.

Senior Editor

  1. Satyajit Rath, Indian Institute of Science Education and Research (IISER), India

Reviewing Editor

  1. Bernard Malissen, Centre d'Immunologie de Marseille-Luminy, Aix Marseille Université, France

Reviewer

  1. Joe Craft, Yale University School of Medicine, United States

Publication history

  1. Received: October 31, 2019
  2. Accepted: May 20, 2020
  3. Accepted Manuscript published: May 22, 2020 (version 1)
  4. Version of Record published: June 5, 2020 (version 2)

Copyright

© 2020, Gryzik et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 731
    Page views
  • 137
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    Kanji Okumoto et al.
    Research Article Updated

    Most of peroxisomal matrix proteins including a hydrogen peroxide (H2O2)-decomposing enzyme, catalase, are imported in a peroxisome-targeting signal type-1 (PTS1)-dependent manner. However, little is known about regulation of the membrane-bound protein import machinery. Here, we report that Pex14, a central component of the protein translocation complex in peroxisomal membrane, is phosphorylated in response to oxidative stresses such as H2O2 in mammalian cells. The H2O2-induced phosphorylation of Pex14 at Ser232 suppresses peroxisomal import of catalase in vivo and selectively impairs in vitro the interaction of catalase with the Pex14-Pex5 complex. A phosphomimetic mutant Pex14-S232D elevates the level of cytosolic catalase, but not canonical PTS1-proteins, conferring higher cell resistance to H2O2. We thus suggest that the H2O2-induced phosphorylation of Pex14 spatiotemporally regulates peroxisomal import of catalase, functioning in counteracting action against oxidative stress by the increase of cytosolic catalase.

    1. Cell Biology
    Natalia Felipe-Medina et al.
    Research Article Updated

    Primary Ovarian Insufficiency (POI) is a major cause of infertility, but its etiology remains poorly understood. Using whole-exome sequencing in a family with three cases of POI, we identified the candidate missense variant S167L in HSF2BP, an essential meiotic gene. Functional analysis of the HSF2BP-S167L variant in mouse showed that it behaves as a hypomorphic allele compared to a new loss-of-function (knock-out) mouse model. Hsf2bpS167L/S167L females show reduced fertility with smaller litter sizes. To obtain mechanistic insights, we identified C19ORF57/BRME1 as a strong interactor and stabilizer of HSF2BP and showed that the BRME1/HSF2BP protein complex co-immunoprecipitates with BRCA2, RAD51, RPA and PALB2. Meiocytes bearing the HSF2BP-S167L variant showed a strongly decreased staining of both HSF2BP and BRME1 at the recombination nodules and a reduced number of the foci formed by the recombinases RAD51/DMC1, thus leading to a lower frequency of crossovers. Our results provide insights into the molecular mechanism of HSF2BP-S167L in human ovarian insufficiency and sub(in)fertility.