A novel myeloid cell in murine spleen defined through gene profiling

Abstract A novel myeloid antigen presenting cell can be generated through in vitro haematopoiesis in long‐term splenic stromal cocultures. The in vivo equivalent subset was recently identified as phenotypically and functionally distinct from the spleen subsets of macrophages, conventional (c) dendritic cells (DC), resident monocytes, inflammatory monocytes and eosinophils. This novel subset which is myeloid on the basis of cell surface phenotype, but dendritic‐like on the basis of cell surface marker expression and antigen presenting function, has been tentatively labelled “L‐DC.” Transcriptome analysis has now been employed to determine the lineage relationship of this cell type with known splenic cDC and monocyte subsets. Principal components analysis showed separation of “L‐DC” and monocytes from cDC subsets in the second principal component. Hierarchical clustering then indicated a close lineage relationship between this novel subset, resident monocytes and inflammatory monocytes. Resident monocytes were the most closely aligned, with no genes specifically expressed by the novel subset. This subset, however, showed upregulation of genes reflecting both dendritic and myeloid lineages, with strong upregulation of several genes, particularly CD300e. While resident monocytes were found to be dependent on Toll‐like receptor signalling for development and were reduced in number in Myd88‐/‐ and Trif‐/‐ mutant mice, both the novel subset and inflammatory monocytes were unaffected. Here, we describe a novel myeloid cell type closely aligned with resident monocytes in terms of lineage but distinct in terms of development and functional capacity.

CD8 − subsets. 4 CD8 + cDC are distinct as CD11c hi CD11b − CD8α + MHCII + cells, while CD8 − cDC have a CD11c hi CD11b lo CD8α − MHCII + phenotype. 5 These subsets differ in immune function, including cytokine production and ability to cross-present antigen. 6 The plasmacytoid (p) DC is another common splenic DC subset, existing as a plasmacytoid preDC in the steady-state. 7 Under inflammatory conditions, monocyte-derived (mo)-DC can develop when inflammatory stimuli recruit circulating classical or inflammatory monocytes from blood into spleen where they differentiate. [8][9][10] Dendritic cells are located mainly within the white pulp region of spleen where immune responses against blood-borne antigens and pathogens are initiated, while myeloid cells are primarily located within the red pulp region.
Spleen contains several subsets of tissue-resident macrophages, namely marginal zone and marginal metaphyllic macrophages, red pulp macrophages and tingible body macrophages in the white pulp region. 11 While the yolk sac origin of red pulp macrophages is understood, [12][13][14] their relationship with splenic monocytes resident mainly in the red pulp region is still unclear. To date, no distinct phenotypic markers are available which can be used to distinguish red pulp macrophages from other myeloid subsets present in the red pulp. The distinction and lineage relationship between splenic monocytes and red pulp macrophages is not well understood.
Monocytes in blood and spleen are thought to derive from myeloid precursors in bone marrow. 7 Two main subsets of circulating monocytes in blood are also present in spleen: the CX 3 CR1 lo Ly6C hi inflammatory or classical monocytes, and the CX 3 CR1 hi Ly6C − resident or non-classical monocytes. 9 Phenotypic identity of the two main monocyte subsets in spleen was recently clarified in this lab through marker phenotype and functional analysis. [15][16][17] That study also classified splenic macrophages as CD11b lo cells, with distinct macrophage subsets identifiable through staining with specific subset markers of SIGNR1, MOMA-1, CD69 and F4/ 80. 17 Under inflammatory conditions, both classical or inflammatory monocytes and non-classical or resident monocytes are selectively mobilized from spleen to the site of inflammation. Here, classical monocytes clear damaged tissues, while non-classical monocytes promote wound healing. 18 Inflammatory monocytes can also home to sites of infection where they differentiate to give mo-DC, 19 while resident monocytes home to non-inflammatory sites where they are thought to differentiate to give macrophages in some tissues, eg liver and spleen. 19 Deployment of a reservoir of splenic monocytes was hypothesized as a mechanism for faster initiation of an immune response. Information on the function of non-classical (resident) monocytes, and whether or not they differentiate to give macrophages within tissues, is still debatable. However, all tissue-resident macrophages are not derived from resident monocytes and evidence for a yolk sac or foetal origin for tissue resident macrophages is forthcoming for some tissues. 14 All evidence points to a major role for spleen in myelopoiesis.
Our own previous work has identified a novel CD11b hi CD11c lo MHC-II − cell type in spleen. This was investigated as an in vivo equivalent to the named "L-DC" subset of dendritic-like antigen presenting cells produced in long-term co-cultures of haematopoietic progenitors over splenic stromal lines. The original studies on "L-DC" produced in vitro described a dendritic-like cell type which was distinct in terms of its phenotype as a CD11b hi CDllc lo MHC-II − cell antigen presenting cell, having very strong capacity to cross-present antigen and to activate CD8 + cytotoxic T cells. [20][21][22] It was predicted that such a cell type located in spleen may play a unique role in the induction of CD8 T cell immunity to blood-borne antigens like pathogens and dead cancer cells (REF). 21,22 Their inability to activate CD4 + T cells would be consistent with their location and function at the level of the spleen and the blood stream because CD4+ T cell activation and cytokine production might be toxic.
In light of their unique functional capacity, studies were initiated to identify this specific cell type in spleen. The absence of specific markers made this process difficult. However, through a series of studies dissecting the myeloid cell populations in both murine and human spleen, 15,23 this novel splenic subset has been identified and analysed in terms of function equivalent to the in vitro-derived cell type. The "L-DC" subset in mice has been shown to be phenotypically distinct from the four splenic macrophage subsets, 17 and both phenotypically and functionally distinct from the two splenic monocyte subsets. 15,16 It was also clearly distinguished from the main splenic DC subsets. [15][16][17]24 In terms of antigen presenting capacity, this novel subset is superior in capacity to cross-present antigen to CD8 T cells and to activate cytotoxic function, although incapable of activating of CD4 T cells. 16 These cells were shown by gene profiling to reflect a distinct cell type expressing genes common to both myeloid and dendritic lineages. 16 This novel subset closely resembles dendritic-like cells produced in vitro in long-term cultures of spleen (LTC-DC) and similar cells produced in co-cultures of bone marrow progenitors over splenic stromal lines. 20,[25][26][27] For this reason alone, it has been referred to as "L-DC" in these studies.
Transcriptome analysis has been used here to analyse the relationship between this new subset and resident and inflammatory monocyte subsets in spleen, eosinophils and the CD8 + cDC and CD8 − cDC subsets. These subsets were isolated in line with a recently published subset identification method which redefined the resident monocyte subset in spleen, and also identified splenic eosinophils more fully. 15 Gene profiling now clearly distinguishes this novel subset. While it is found to be closely related to resident and inflammatory monocytes, evidence presented here also distinguishes it as developmentally and functionally distinct.
was conducted according to protocols approved by the Animal Experimentation Ethics Committee at the ANU. Mice were sacrificed by cervical dislocation.

| Cell preparation
Following dissociation of spleen tissue through crushing, T and B lymphocytes were depleted through column separation performed with MACS ® technology (Miltenyi Biotec: Auburn, California, USA).
Cells were resuspended at 10 8 cells/mL in MACS labelling buffer

| Antibody staining and flow cytometry
Procedures for staining cells with antibodies for flow cytometry have been described in detail in earlier studies. 15 Anti-CD16/32 (FcBlock: 5 µg/mL) (Biolegend, San Diego, CA, USA) was used to block non-specific antibody binding through Fc receptors. Fluorochrome-or biotin- FACSDiva software (Becton Dickinson) was used to acquire data. FloJo software (Tree Star, Ashland, OR, USA) was used for data analysis.

| Transcriptome analysis
Cell sorting of dendritic and myeloid subsets in spleen was performed as described previously 15

| Statistical analysis
Where indicated, data has been obtained from multiple animals, data are presented as mean ± SE for sample size (n). The Student's t test has been used to determine significance (P ≤ 0.05).

| Transcriptome analysis of splenic dendritic and myeloid subsets
The clearly defined populations of resident monocytes, inflammatory monocytes, eosinophils, CD8 + cDC and CD8 − cDC were sorted from spleens of C57BL/6J mice according to the staining procedure developed previously (Table 1). 15 Figure 1A). "L-DC" and resident monocytes showed the least variance with very few differentially expressed genes evident as outliers ( Figure 1A). A comparison of "L-DC" and inflammatory monocytes showed more differentially F I G U R E 1 Variability in gene expression amongst dendritic and myeloid subsets. Transcriptome analysis was performed on subsets of cells sorted from murine spleen. RNA was extracted and labelled for hybridization to Murine Gene ST1.0 genechips (Affymetrix). Following scanning to collect signal values from samples prepared in duplicate experiments, data were analysed using Partek and ANOVA by pairwise comparison. A, Mean signal values were calculated and plotted for a total of 35 556 genes in pairwise subset comparisons. The darker blue inner polygon contains 50% of data points, while the pale blue outer polygon contains all other data points which are not outliers (shown in red outside the polygon). The bivariate median is shown by the red asterisk at the centre of the polygon. B, Principle component analysis was used to determine variability in gene expression for each subset. Three principle components are shown for each subset prepared for analysis in duplicate experiments. C, Hierarchical clustering was used to analyse the relationship between subsets on the basis of average gene expression. The dendrogram displays distance between subsets based on clustering of 8508 genes selected for analysis on the basis of mean signal value ≥100 for any one subset expressed genes. The "L-DC" and cDC subsets showed more variation with many differentially expressed genes evident as outliers.
The resident and inflammatory monocyte subsets showed very few differentially expressed genes, indicating a close relationship. The CD8 + cDC and CD8 − cDC subsets gave a tight bivariate plot with a high number of differentially expressed genes. Eosinophils showed the greatest difference in gene expression in comparison with all other subsets ( Figure 1A).
Differences in overall gene expression between the subsets were also evident through principal component analysis (PCA). This showed close grouping of resident monocytes, inflammatory monocytes, "L-DC" and cDC subsets in the first principal component, but separation of "L-DC" and monocyte subsets from cDC subsets in the second principal component ( Figure 1B). In addition, CD8 + cDC were clearly differentiated from CD8 − cDC in the second principal component. Lastly, eosinophils were distinct from all other subsets on the basis of the first and second principal components. This analysis indicated similarity between "L-DC" and monocytes and clearly differentiated "L-DC" from eosinophils and cDC subsets.
Hierarchical clustering was then used to map the relationship between subsets based on gene expression ( Figure 1C). Average signal values from duplicate samples were used for clustering. Genes were included which showed expression in at least one subset (signal value ≥100), giving a sample set of 8508 genes. The analysis indicated a close relationship between "L-DC" and resident monocytes and then between these two subsets and inflammatory monocytes.
CD8 + cDC and CD8 − cDC were closely related and distinct from the cluster of "L-DC," resident monocytes and inflammatory monocytes.
As predicted from PCA analysis, eosinophils were quite distinct as a subset and lineage.

| Investigation of gene expression specific to subsets
The lineage origin of the different subsets was investigated by data mining and comparing the expression level of known genes to functional categories of "DC and APC," "Chemokines," "Cell surface mark-  (Figure 2A and 2). 31,32 In terms of genes encoding chemokines, inflammatory factors and related genes, the two cDC subsets showed common expression of Ccr7, Ccl5, IL-1b, Itgb2, IL2rg and IL-10ra ( Figure 2C and 2). 35 Figure   S1). This suggests that none of the subsets analysed here reflect red pulp macrophages. Further investigation of "L-DC" has shown it to be readily distinguishable from red pulp macrophages through phenotype 15,17 and lack of expression markers like ITGA9 and VCAM-1 and genes like SpiC and Mertk, previously associated with red pulp macrophages (data not shown). 51

| Genes upregulated in the novel subset
Data sets were extracted to identify genes upregulated at least threefold in either "L-DC" or the two monocyte subsets. "L-DC" and inflammatory monocytes were found to be the most distinct subsets, while "L-DC" and resident monocytes were the most closely related ( Figure 3A). The data also predict a close relationship between resident monocytes and inflammatory monocytes, consistent with PCA and clustering evidence (Figure 1). Only four genes were found to be uniquely up-regulated in one of the three subsets. Upregulation of transmembrane residue which interacts with DAP12. 55 It was previously shown to be expressed by macrophages/monocytes, mo-DC, and at lower levels in in vitro-derived macrophages and DC. 55 Upon binding, CD300E induces activation signals calcium mobilization and release of reactive oxygen species by monocytes. 56,57 In addition, CD300E binding also induces cytokine release by monocytes and promotes survival of monocytes and mo-DC. 56,57 DC activated via CD300E have stronger capacity to stimulate T cells. 57 CD300E upregulation is consistent with the superior antigen presenting capacity of "L-DC" over the other monocyte subsets. 16,17 Other genes which were upregulated at 2.5-fold included Cd300ld, Serpinb6a and Dnahc12 ( Figure 3C). CD300LD belongs to the same family as CD300E and participates in signal transduction and production of pro-inflammatory cytokines. 58,59 Serpinb6a encodes a protein essential for protection against cytotoxic granules, 60 while Dnahc12 encodes a protein that forms part of dynein. 61 Genes upregulated ≥2-fold in "L-DC" over other subsets were mainly proteases, adhesion proteins and transmembrane proteins. Upregulation of Fcgr4 is of interest because CD300E has been shown to physically interact with FcRγ. 62

| Identification of genes which distinguish the novel subset from resident monocytes
Both PCA and gene expression analyses revealed a close developmental relationship between resident monocytes and "L-DC" which differ from resident monocytes through low expression of CD43 and absence of Ly6C expression (Table 1). To further investigate this relationship, genes specifically upregulated in either "L-DC" or resident monocytes were identified (Figure 4). Cd300e and Cd9 were shown to be upregulated in "L-DC" over resident monocytes, along with Dnahc12, Tgm2, Pecam1, Fabp4, Rab11, Serpinb6a, Abhd2 and Sash3 ( Figure 4A). Both CD300E and CD9 regulate the ability of DC and monocyte/macrophages to activate T cells. 56,57,63 In addition, CD9 also modulates cell adhesion and migration 64  SASH3, also known as SLY1, participates in the regulation of marginal zone B cell development via the Notch signalling pathway. 70 All of these genes reflect the function of APC and are consistent with the defined functional role of "L-DC" in CD8 + T cell activation and cytotoxic function. 16,17 Consistent with the reports in the literature, resident monocytes expressed Ly6c1, Ly6c2 and Ccr2 ( Figure 4A). 9,41 Expression of Ly6C1/2 reflects the sorting strategy used here, whereby resident monocytes were separated from "L-DC" (Table 1). In addition, resident monocytes also showed upregulation of several genes known to be expressed by myeloid cells including Chi3l3, Ifi205, Gm11428 encodes AMWAP which is expressed by tissue macrophages, microglia and retinal cells and regulates proinflammatory microglia and macrophage activation. 83 This phenotype distinguishes resident monocytes from the novel APC subset in that it reflects activated monocytes with wound-healing capacity perhaps related to M2 macrophages.
As very few genes were identified as upregulated by "L-DC" over resident monocytes, genes specifically expressed by "L-DC" or resident monocytes were sought. Genes were selected according to the criteria of signal value ≥125 in one subset and ≤50 in the other. This gave a subset of seven genes specific to resident monocytes, but none for "L-DC" ( Figure 4B). Amongst genes specifically expressed by resident monocytes, Ngp, S100a8 and S100a9 encode monocyte and macrophage markers. 84,85 NGP regulates monocyte functions of activation and recruitment into sites of infection. 86 Both S100A8 and S100A9 have been described as activators of endogenous TLR4, so promoting proinflammatory responses. 84 F I G U R E 4 Genes upregulated or specifically expressed between "L-DC" and resident monocytes. ANOVA was used to make pairwise comparisons of average gene expression (n = 2) between subsets and to calculate relative fold changes. A, Genes upregulated in either "L-DC" or resident monocytes were selected as those for which the signal value in one subset was ≥50, and the signal value in the second subset was ≥125. Data shown reflect genes with ≥2.5-fold difference in signal value. B, Genes specifically expressed in either "L-DC" or resident monocytes were selected on the basis of the mean signal value in one subset ≤50, and mean signal value in the second subset ≥125. No genes were found to be specifically expressed by "L-DC"

| Identification of markers which distinguish resident and inflammatory monocyte subsets
Both resident (non-classical) monocytes and inflammatory (classical) monocytes are shown here to be closely related. Previously, bloodderived inflammatory monocytes were described as precursors of resident monocytes, 88,89 although that relationship is still unclear.
Genes upregulated in one or other subset were therefore identified to further distinguish these two spleen monocyte subsets. In line with earlier gene expression profiles of murine blood monocytes, 54 resident monocytes from spleen upregulated genes encoding known markers such as Ccl5, Itgax, Cd300e, Dusp16, Cd36, H2-Ab1 and Fabp4 ( Figure 5). Upregulation of Itgax (CD11c) by resident monocytes is consistent with the staining and gating strategy used here ( Table 1).
Upregulation of H2-Ab1 and H2-Aa by resident monocytes could indicate potential to express MHCII and act as APC. Dusp16 is also upregulated and encodes a dual-specificity phosphatase that can regulate mitogen-activated protein kinase for signal transduction and gene transcription which selectively regulates cytokine production by myeloid cells. 90,91 Resident monocytes also show upregulation of Ccl5 which encodes a chemokine involved in recruitment of leukocytes to sites of inflammation and promotes recruitment and survival of human macrophages. 92 Inflammatory monocytes were found to upregulate several genes involved in inflammatory monocyte function including Mmp8, F13a1 and Fn1 described previously, as well as Vcan, Cd14, Capg, Ms4a8a and Cxcl10 ( Figure 5). Cd14 encodes a marker on human inflammatory monocytes which acts as a co-receptor for TLR4 signalling. 19,93 Capg encodes a protein which participates in control of actin-based motility in macrophages, 94,95 Ms4a8a encodes a tetraspanin as a marker of activated M2 macrophages, 96 and Cxcl10 encodes a chemokine produced mainly by neutrophils and inflammatory monocytes. 97

| Toll-like signalling during development distinguishes "L-DC" and resident monocytes
Previously we showed that "L-DC" can develop in vitro from hematopoietic stem cells (HSC) overlaid above splenic stroma. 27  F I G U R E 5 Genes upregulated in either resident monocytes or inflammatory monocytes. ANOVA was used to make pairwise comparisons between average gene expression (n = 2) in inflammatory and resident monocytes. Genes were selected which showed ≥4-fold change in mean signal value in either resident monocytes (Resi mono) or inflammatory monocytes (Infl mono), where mean signal value in both subsets was ≥50 the total dendritic and myeloid subset in spleen was, therefore, measured in MyD88 −/− , Trif −/− and MyD88 −/− /Trif −/− mice compared with wild-type control mice. In MyD88 −/− mice, a significant 2.5-fold increase in the percentage of CD8 − cDC was seen compared with wild-type mice ( Figure 6A). Eosinophils showed a significant but small decrease. The populations of inflammatory monocytes, resident monocytes, "L-DC" and neutrophils in MyD88 −/− mice were not significantly different from the wild-type mice. In Trif −/− mice, a significant reduction in percentage of both CD8 + cDC and CD8 − cDC was observed in Trif −/− mice compared with wild-type mice and is associated with TLR3 signalling ( Figure 6B) From combined studies on the three mutants, it was concluded that TLR signalling is important in the development of CD8 − cDC, CD8 + cDC, eosinophils and resident monocytes. However, the development of "L-DC" and inflammatory monocytes occurred independently of TLR signalling, such that the latter two subsets develop in steady-state spleen. It is important to note that our protocol for delineation of the CD8 − cDC subset could also capture inflammatory or mo-DC whose development would be lost in mutant mice. As "L-DC" development occurs independently of inflammatory signals, these results serve to distinguish "L-DC" as a distinct subset from resident monocytes, and to definitively distinguish resident monocytes from inflammatory monocytes.

| D ISCUSS I ON
This study has made a number of contributions towards better understanding dendritic and myeloid subsets present in murine spleen.
In particular, a novel subset equivalent to the in vitro generated "L-DC" cell type has been characterized in terms of gene expression and shown to be distinct from other known DC subsets and monoctyes. Full and complete analysis of splenic subsets initially required that splenic macrophages were first gated out, and that the monocyte subsets were redefined. [15][16][17] Inflammatory monocytes are now clearly distinguishable as a separate lineage from resident monocytes, and are also distinct from the novel subset of interest. The resident monocyte subset defined in spleen was previously shown to be phenotypically distinct from resident monocytes previously defined in murine blood. 15 Spleen resident monocytes are now shown to be closely related to a novel APC subset described here as "L-DC," such that the two subsets may be derived from a common progenitor or lineage. 104 The possibility that "L-DC" reflect a macrophage subset was considered but refuted previously. 17 (Table 1).
Ly6C hi inflammatory monocytes can give rise to mo-DC-like TNFα and iNOS-producing DC (Tip-DC) in murine tissues during inflammation. Tip-DC have also been described as classically activated M1 macrophages. [109][110][111] It is notable that "L-DC" development occurs independently of inflammatory signals essential for generation of Tip-DC (Figure 6), and the "L-DC" phenotype is distinct from that of Tip-DC through lack of Ly6C and MHCII expression (Table 1).
These findings clearly distinguish "L-DC" from mo-DC which develop in response to inflammation.
Based on gene expression data obtained here and phenotypic and functional data obtained previously, [15][16][17]112 "L-DC" can be distinguished as a unique myeloid subset in spleen. They are more closely F I G U R E 6 "L-DC" development occurs independently of Toll-like receptor signalling. Splenocytes were harvested from C57BL/6J mutant and C57BL/6J (wild type) mice. Cells were stained with antibodies to delineate subsets as described in Table 1. Gates were set based on fluorescence minus one controls, to estimate % cells amongst the total myeloid and dendritic subset (CD11b + and/or CD11c + ) cells. Individual mice were analysed (n = 4 or 5 related to monocytes than to cDC, although the reason for this could relate to progenitor origin rather than function as an APC. Indeed, their function as APC is distinct from cDC subsets in that they activate only CD8 + T cells and not CD4 + T cells, and appear to have capacity to cross-present antigen. 16,17 The resident (non-classical) monocyte population in spleen quite distinct from the inflammatory (classical) monocyte subset, despite evidence for a common myeloid phenotype. Both CD8 + cDC and CD8 − cDC were closely linked in terms of gene profile, and quite distinct from monocytes and the "L-DC" subset. Lastly, the gene profile of eosinophils was quite distinct from other subsets isolated, suggesting a distinct lineage origin, consistent with evidence that the eosinophil develops from a granulocyte/macrophage progenitor instead of the macrophage/dendritic progenitor. [113][114][115] Gene profiling studies were conducted with a view to identification of distinguishing markers for "L-DC" for better classification of this subset. However, markers were not found, and "L-DC" were shown to be closely related to resident monocytes differing only through upregulation of markers related to T cell activation capacity, namely CD300E, CD300LD, SERPINb6a and CD9. SERPINb6a is widely expressed, and CD300E and CD300LD have expression aligned with DC subsets, 59 non-classical monocytes and macrophages. 62 While no specific genes were found to distinguish "L-DC" from resident monocytes, a number of specifically expressed genes did distinguish resident monocytes from "L-DC." These genes reflect myeloid cells rather than DC including Ly6C, S100A8 and CD209. Although this expression pattern could be consistent with mo-DC, 116 no evidence was found for upregulated CD206, or for production of TNF and iNOS, which are delineating markers of mo-DC. 116,117 The possibility that "L-DC" reflect mo-DC was considered, and refuted on several accounts. Firstly, "L-DC" do not express markers identified for mo-DC including SIRPA, S100A8, CD206 and CD209a. 116 Secondly, "L-DC" development both in vivo 15  consistent with the common progenitor origin. 112 In vitro studies to define the hematopoietic progenitors which generated "L-DC" when cocultured above a splenic stromal line which supported hematopoiesis, revealed that "L-DC" arose only from HSC or multipotential progenitors and not from other myeloid progenitors, 27 suggesting that "L-DC" may differentiate directly from HSC in the absence of formation of a myeloid progenitor. Our hypothesis therefore is that "L-DC" and resident monocytes may have a common progenitor origin in spleen but arise by divergent differentiation.

| CON CLUS ION
The close relationship between gene profiles for "L-DC" and resident or non-classical monocytes raises questions about a possible common lineage origin. They are shown here to be quite distinct subsets in that resident monocytes require TLR signalling for their development, while "L-DC" do not. The latter could reflect a steady-

ACK N OWLED G EM ENTS
This work was supported by project grant #585443 from the National Health and Medical Research Council of Australia to HO.

CO N FLI C T S O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

AUTH O R CO NTR I B UTI O N S
YH designed, performed and analysed experiments and wrote the paper. TO analysed data and reviewed the paper. HO designed, supervised and analysed experimental work and wrote the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors declare that data will be made available upon request from the authors.