Dynamic single cell transcriptomics defines kidney FGF23/KL bioactivity and novel segment-specific inflammatory targets

Abstract

Fibroblast growth factor 23 (FGF23) via its coreceptor αKlotho (KL) provides critical control of phosphate metabolism, which is altered in both rare and very common syndromes. However, the spatial-temporal mechanisms dictating kidney FGF23 functions remain poorly understood. Thus, developing approaches to modify specific FGF23-dictated pathways has proven problematic. Herein, wild type mice were injected with rFGF23 for one, four and 12h and kidney FGF23 bioactivity was determined at single cell resolution. Computational analysis identified distinct epithelial, endothelial, stromal, and immune cell clusters, with differential expressional analysis uniquely tracking FGF23 bioactivity at each time point. FGF23 actions were sex independent but critically relied upon constitutive KL expression mapped within proximal tubule (segments S1-S3) and distal convoluted tub/connecting tubule cell sub-populations. Temporal KL-dependent FGF23 responses drove unique and transient cellular identities, including genes in key MAPK-signaling and vitamin D-metabolic pathways via early- (transcription factor AP-1-related) and late-phase (initiation factor EIF2 signaling) transcriptional regulons. Combining ATACseq/RNAseq data from a cell line stably expressing KL with the in vivo scRNAseq pinpointed genomic accessibility changes in MAPK-dependent genes, including the identification of FGF23-dependent early growth factor-1 distal enhancers. Finally, we identified unexpected crosstalk between FGF23-mediated MAPK signaling and pro inflammatory TNF receptor activation via transcription factor NF-κB, which blocked FGF23 bioactivity in vitro and in vivo. Collectively, our findings have uncovered novel pathways at the single cell level that likely influence FGF23-dependent disease mechanisms.

Graphical Abstract

Introduction

Human and mouse kidneys are formed with one million and 14,000 nephrons, respectively,^1,2 which are individually composed of distinct functional segments designed to control mineral homeostasis.³ Systemic phosphorus handling is regulated by FGF23, a hormone produced by osteocytes in response to elevated circulating phosphate or 1,25(OH)₂ vitamin D (1,25D)(reviewed in⁴). FGF23 acts through high affinity binding to its co-receptor αKlotho (KL) and FGF receptors (FGFRs) to mediate its renal effects.⁵ Under elevated blood phosphate, FGF23 promotes phosphaturia through the downregulation of the phosphate co-transporters Slc34a1 and Slc34a3 (NPT2A/-C) via their internalization in proximal tubular cells.^6,7 Additionally, FGF23 controls 1,25D in coordinated regulation of decreasing the expression of the 1,25D anabolic Cyp27b1 (vitamin D 1alpha-hydroxylase) and increasing the catabolic Cyp24a1 (vitamin D 24-hydroxylase), with the net effect of these actions to reduce both serum Pi and 1,25D. However, the full understanding of the transcriptional activity upstream of these pathways have been explored only through individual candidate regulatory studies, and are thus poorly understood.

In the kidney, FGF23 signaling through FGFR-KL complexes⁸ stimulates MAPK cascades coupled with the activation of the phospho-early response kinase (p-ERK1/2) to induce target genes including the transcription factor early growth response gene-1 (Egr1).^5,9 Genetic mouse models have been generated through conditional deletion of KL from nephron proximal tubule (PT) and distal tubule (DT)^10,11 to uncover segment-specific roles of FGF23. However phenotypic differences across conditional mouse models have been observed, potentially due to the variability in Cre-mediated recombination efficiency and cell-type specificity. Overall, KL deletion in PT or DT resulted in normophosphatemia or moderate hyperphosphatemia, compared to the hyperphosphatemic phenotype observed in global KL knock out mice.^12,13 Further, the molecular mechanisms responsible the altered phosphate and 1,25D regulation in diseases associated with FGF23 overexpression, such as ADHR¹⁴ and XLH,¹⁵ as well as in multi-factorial disorders such as chronic kidney disease (CKD),¹⁶ where patients have markedly elevated FGF23 and reduced KL expression, remain incompletely understood. Finally, FGF23 signaling may be influenced by disease states, as interactions between inflammation and FGF23 have been proposed. Indeed, in a large cohort study, higher quartiles of FGF23 were associated with elevated mean concentrations of IL-6 and IL-10.¹⁷ However, how inflammatory signals may influence FGF23 bioactivity specifically in kidney cells are not well understood.

Here, we tested the molecular function of FGF23 in kidney, the localization of these actions with regard to KL expression, and project translational manifestations to diseases associated with FGF23. We identified spatial and temporal transcriptional reprogramming within the nephron during FGF23-mediated signaling. We also identified in vivo and confirmed in vitro novel crossover activity of FGF23 with TNF family members and NF-κB signaling. Collectively, this work isolated kidney FGF23 function at single cell resolution and identified novel pathways, genes, and transcription factors regulated by this hormone, with potential therapeutic implications for acute and chronic mineral diseases.

Material and Methods (see Supplementary Methods for additional details)

Research animals

Animal studies were approved by the IUSM Institutional Animal Care and Use Committee.

FGF23 administration:

C57BL/6J male and female mice (9-weeks-old) were purchased from the Jackson Laboratory and housed for 1 week. At 10-weeks of age, mice were treated with a single intraperitoneal injection of 500 ng/g body weight of rhFGF23 and euthanized at baseline (0h), or after 1, 4, 12, and 24 h. For scRNAseq experiments, male-female kidney pairs were digested, and cells were isolated as previously described.¹⁸ Confirmatory analyses were tested in total kidney RNA using n = 5-6 mice per time point and per gender.

BMS, TWEAK and FGF23 administration:

C57BL/6 10-week-old female mice were treated with BMS-345541 (0.5 mg/mouse i.p. injection, one time), and after 1 h, mice received a second single i.p. injection of FGF23; mice were euthanized after 1 h. The combined treatment duration of BMS-345541 to the mice was 2 h. Hyp mice and littermates (8-12 weeks of age) were treated with a single dose of BMS-345541 (0.5 mg/mouse, by one i.p. injection) for 3 h. C57BL/6 10-week-old female mice were treated daily with rTWEAK (2 ug per mouse) for 4 days followed by a single i.p. injection of FGF23 and euthanized 4 h later.

Single cell library preparation and data processing

For each sequencing set, a 10,000 cell recovery was targeted, and a single cell master mix with lysis buffer and reverse transcription reagents was used according to 10X Genomics Chromium Single Cell 3’ Reagent Kit V3. This was followed by cDNA synthesis, library preparation, and sequencing on NovaSeq6000 platform in paired-end mode (28bp + 91bp). Data were processed as previously described.¹⁸

Single nuclei Assay for Transposase-Accessible Chromatin sequencing (ATAC) data analysis

A publicly available male kidney scATAC-seq dataset accessible in the GEO database under the number GSE210938 (GSM443123)¹⁹ was used for mapping mouse KL genomic regions.

RNA isolation and qPCR

Kidneys were harvested, homogenized, followed by RNA extraction and qPCR as previously performed.²⁰

In vitro studies

HEK293 cells and HEK293 stably transfected with membrane bound KL (‘HEK293-mKL’ cells) were cultured similarly to our previously published protocol.²⁰

Bulk mRNA and ATAC sequencing

After treatment with FGF23 (50 ng/ml for 4 h), total RNA from HEK293-mKL cells was extracted and evaluated for its quantity and quality using a Bioanalyzer 2100; 100 ng of total RNA was used for cDNA libraries. For ATACseq, nuclei isolation and ATAC sequencing was performed according to published protocols.²¹

Immunoblotting

HEK293-mKL cells were lysed and 30 μg of cellular lysate protein was loaded for immunoblots, as previously performed.²⁰

Quantification and statistical analysis

Updated R software packages with robust affiliated statistics were used to analyze scRNAseq and snATACseq datasets. Statistical analyses of the in vitro data were performed by one-way ANOVA to assess the differences between the same cell line responses to FGF23, and a two-way ANOVA to assess the response differences between two different cell lines. Statistical analyses of the in vivo data were performed by one-way ANOVA to assess the differences between same genders in response to treatment and two-way ANOVA to assess differences between genders and treatments. Means and standard deviations were used in bar graphs and kinetic curves. Significant changes were considered at P < 0.05.

Results

Renal scRNAseq enables sex-specific sample demultiplexing

To test FGF23 bioactivity in the kidney, male and female C57BL/6J mice were injected with FGF23 (400 ng/g) for 0, 1, 4, and 12 h followed by euthanasia, and the generation of a pooled male/female scRNAseq dataset (Figure 1a) was demultiplexed by time and sex (Supplementary Figure S1a and S1b). We mapped sexually dimorphic kidney phenotypes by counting the number of sequencing reads of male- versus female-specific genes.^22,23 Male and female cell types were identified within clusters using sex-specific markers, and analyzed together as there were no sex-dependent transcriptional changes in response to FGF23 (see Supplementary Methods). We identified 21 distinct clusters in kidney including epithelial and immune cells, which were classified into unique populations based upon known expressional markers of specific renal cell types (Figure 1b-1d).^18,22,24-27 Of note, proximal tubule (PT) S1-S3 cells represented the most abundant types (Figure 1e).

Figure 1. Renal sex-multiplexing scRNA sequencing.

a. Experimental Overview. Pooled male and female kidneys from FGF23-injected mice were dissociated into single-cell suspensions followed by scRNAseq and downstream computational and molecular analysis. b. Unsupervised UMAP clustering identified various renal cell populations divided into 21 distinct cell types. c. Known sex-specific markers (described in Results) were used to identify male and female cell populations. d. Dot plot of representative genes show sex-specific cell types. e. Stacked bar plot displays the relative proportions of each cell type after FGF23 treatment. The different cell types identified are color-coded and annotated for clustering as in Fig. 1b.

Klotho expression coordinated with FGF23 activity in the nephron

The FGF23 bioactivity within the nephron remains unresolved as its co-receptor Klotho (KL)⁵ is expressed in multiple segments. KL had the highest expression in DCT/CNT cells, in agreement with previous studies^9,11,28,29 (Figure 2a). The expression of Fgfr1 was detected in renal epithelial cells (Supplementary Figure S2), confirming the ability of these cells to form FGF23-FGFR1-KL complexes, known to initiate FGF23 signaling. To determine cell-specific genomic predisposition associated with KL expression, we analyzed mouse kidney snATACseq (Supplementary Figure S3a) and found that chromatin accessibility at KL promoter regions were limited to PT, loop of Henle (LOH), DCT and CNT cells, supporting our findings of KL mRNA distribution (Figure 2a-2b). Interestingly, accessible chromatin regions between exons 1 and 2 which corresponded to a VDR binding site were present in PT but not in LOH, DCT, and CNT cells (Figure 2b and Supplementary Figure S3b-S3d). Taken together, the differential genomic predisposition within the KL gene supported FGF23 site-specific regulation of mineral metabolism.

Figure 2. KL defines a specific FGF23 signature in renal tubules.

a. Ridgeplots show KL mRNA expression in select renal cells. b. snATACseq detected a differential chromatin accessibility between PT, LOH, DCT, and CNT versus stromal and endothelial cells at the KL promoter region. The red bracket shows differential chromatin accessibility in a region located between exon 1 and exon 2 when comparing PT cells types versus LOH, DCT, CNT, and CD-IC. c. Cell states in PT and DCT cells in response to FGF23 with a time-dependent clustering in PT versus DCT. d. Ridgeplot analysis shows KL expression in PT(S1-S3) and DCT. The vertical red line demarcates cells with low KL expression (KL^low) or high KL expression (KL^high). e. The Monocle program was used to cluster PT-S1 and PT-S2-S3, to separate cells into KL^high and KL^low groups, and display cells by treatment condition (0, 1, 4, and 12 h). f. Overview of nephron segment specific pathways activated in proximal tubule and distal tubule cells in response to FGF23. Ingenuity Pathway Analysis (IPA) was used to predict the statistically significant canonical pathways upregulated in PT and DT in response to FGF23 at 1, 4, and 12 h. The z-score represents the ratio of the number of upregulated genes found in each pathway versus the total number of genes known to be involved in that pathway; the size of the dot represents the p value.

To test the effects of KL expressional heterogeneity, we applied pseudotime analysis³⁰ to the dataset, and identified a heterogenous cell-state in PT and a more homogeneous cell-population (DCT) that clustered depending upon FGF23 treatment (Figure 2c). In DCT, a profound phenotype was observed at 4 h when compared with untreated cells (Figure 2c). It was then hypothesized that FGF23-dependent cell-state clustering was driven by differential KL expression. Interestingly, using the visualization tool Ridgeplot, KL mRNA in the PT and DCT cell subpopulations showed a clear cutoff if KL expression was >0.4 or <0.4 (Figure 2a), allowing the assignment of renal cells as KL^high or KL^low, respectively (Figure 2d). To determine the role of KL-dependent FGF23 bioactivity in PT cells, Monocle 3 was used to discover cell subtypes that clustered based upon KL expression specifically within PT-S1 and PT-S2/S3 (Figure 2e, top and bottom left). Interestingly, KL^high subclusters were driven by spatio-temporal separation of cells in response to FGF23. When stimulated by FGF23, in contrast to KL^low clusters, KL^high cells possessed distinct transient cellular phenotype changes, suggesting differential gene signatures (Figure 2e, top and bottom right). Monocle analysis on DCT cells, however, revealed only a single partition, consistent with the KL homogeneity observed in DCT (Supplementary Figure S4a-S4b). To confirm whether KL expression was heterogeneously distributed across PT cells, a duplex RNA in-situ hybridization for KL and Slc34a1 (Npt2a) expression was performed. As expected, an oversaturation of KL signal was detected in DCT/CNT (SLC34a1 negative cells) consistent with these segments having high KL expression. In contrast, Slc34a1-positive cells harbored differing KL mRNA expression (Supplementary Figure S4c-S4g). These findings support the concept that KL production within a segment-specific cell population can dictate downstream FGF23 bioactivity.

To identify potential common and unique pathways driven by FGF23 bioactivity in Slc34a1 positive cells (highest expression observed in PT-S1) and Slc12a3 positive cells (primarily DCT), differentially upregulated gene sets in PT-S1 cells vs DCT cells were first isolated (Supplementary Figure S5a), then the Ingenuity Pathway Analysis (IPA) tool was used to identify enriched and upregulated pathways. FGF23 signaling was associated with MAPK pathways such as EIF2 (Figure 2f). In addition, other stress-related pathways such as NRF2-mediated oxidative stress response and the unfolded protein response were increased in both PT-S1/S2 and DCT. Further, known FGF23-mediated pathways such as VDR signaling was induced by FGF23 in PT cells (Vdr expression in PT-S1 shown in Supplementary Figure S5b). Collectively, our temporal data support that FGF23-mediated vitamin D metabolism is preceded by overlapping biological process initiated by MAPK and EIF2 signaling (Supplementary Figure S5c).

Dynamic, cell specific FGF23 bioactivity

We next sought to localize FGF23 bioactivity across the nephron. At the mRNA level, KL expression remained stable in response to acute FGF23 administration (Figure 3a). Egr1, a transcription factor and well-known FGF23 target²⁰ acutely rose primarily in PT, DCT and CNT 1 h after FGF23 delivery. These increases in Egr1 observed at 1 h decreased at 4 h before resetting to baseline at 12 h, suggesting a resolution of the 1 h “early phase” of FGF23 actions (Figure 3b). Only epithelial nephron cells responded to FGF23 delivery as we observed no significant increase of Egr1 in immune or endothelial cells which were KL^low cell populations. In whole kidney, Egr1 expression was elevated in response to FGF23 at 1 h. In bulk renal RNA analysis, KL expression was not statistically different after FGF23 treatment (Figure 3c-3d).

Figure 3. FGF23 bioactivity across the nephron.

a-j. The mRNA expression of FGF23 target genes Kl, Egr1, Cyp24a1, Cyp27b1, and Slc34a1, respectively, was tested in the kidney scRNAseq dataset and by qPCR analysis. Feature plots show the expression of FGF23 targets in the different renal cell populations. The kinetic curves represent the qPCR analysis in total kidney with male data shown in blue and female data represented in red. The qPCR data are expressed as fold change (2-ΔΔCt) relative to the housekeeping gene β-actin. Data are shown as mean +/− standard deviation. k. Dot plot of representative genes that define FGF23 bioactivity in PT(S1-S3), DCT, and CNT.

To pinpoint the cells involved in FGF23 regulation of the 1,25D metabolic enzymes, Cyp24a1 and Cyp27b1 mRNAs were mapped and localized to PT. In response to FGF23, Cyp24a1 increased, whereas Cyp27b1 expression decreased (Figure 3e and 3f). These changes of Cyp24a1 and Cyp27b1 were confirmed in total kidney (Figure 3g and 3h). The expressional mapping of Slc34a1 (Figure 3i and 3j) showed little change at the transcriptional level. However, FGF23 treatment resulted in a time-dependent decrease of Slc34a1 protein expression over a 24 h time course (Supplementary Figure S6). Additionally, genes such as Id1-3 were highly upregulated in PT whereas Nab2, and Mcl1 were increased in DT (Figure 3k). Other genes such as Egr1, and Hbegf were induced in both PT and DT (Figure 3k). Finally, a cell survival phenotype was associated with the upregulation of Hes1, Tmem208, and Myof, in parallel with up-regulation of TNFR family members and NF-κB dependent genes such as Tnfrsf12a (Figure 3k). These results support that KL-dependent FGF23 signaling induced an acute phase of signaling that overlaps in PT and DT cells, as well as a more latent phase that includes nephron site-specific, dynamic control of 1,25D metabolism.

FGF23-induced NF-κB is a negative regulator of MAPK signaling

To determine the transcription factors (TFs) activated in response to FGF23 signaling, early (1 h) and late (4 h and 12 h after FGF23) phase TF regulon activity was examined using Single-Cell rEgulatory Network Inference and Clustering (SCENIC).^31,32 In PT-S1, the top 15 regulons during early phase included Egr1, Fos, and Jund (Figure 4a; and Supplementary Data). In the late phase, Nfkb1 and Nfkb2 regulons were increased (Figure 4a; and Supplementary Data). Egr1, Fos and Bhlhe40 regulon activity and mRNA expression were activated in specific cell subsets (Figure 4a-4c; cells surrounded in red). Similar to the in vivo regulation of PT-S1 cell Egr1, Fos, and Bhlhe40, in vitro kinetic studies confirmed that these same mRNAs were upregulated in human HEK293-mKL cells with FGF23 treatment (Figure 4d-4f). In sum, these data demonstrate that FGF23 bioactivity coordinated an overlapping panel of early-and later-phase transcriptional activators and repressors that were specific to distinct nephron cell populations.

Figure 4. NF-kB is a negative regulator of FGF23 bioactivity.

a-c. t-SNE plots illustrate regulon activities of Egr1, Fos, and Bhlhe40 (blue) in PT-S1 cells in parallel with the transcription factor mRNA expression detected in the SCENIC-generated dataset (orange color in t-SNE plot). The insets illustrate the predicted binding motifs. d-f. The mRNA expression of EGR1, FOS, and BHLHE40 obtained from in vitro studies were measured by qPCR. The red kinetic curve represents the HEK293-mKL cell studies whereas the black kinetic curve shows the data from native HEK293 cell experiments. g. Integration of ATACseq and RNAseq data displays selected genes that showed both increased chromatin accessibility (x-axis) with gene expression alterations (y-axis). h. Functional enrichment analysis using the Database for Annotation, Visualization and Integrated Discovery (DAVID) on upregulated DEGs with FGF23 treatment (numbers in parenthesis are gene counts). Heatmap displays the genes associated with NF-κB pathway activation. i. HEK293-mKL cells were pretreated with BMS-34554 (10 μM) for 1h followed by FGF23 (50 ng/ml) for 4h. Immunoblots were then performed to assess the phosphorylation of ERK, cJUN, and cFOS. j. HEK293-mKL cells were pretreated with BMS-34554 (10 μM) for 1h followed by FGF23 (50 ng/ml) for 4 h and subsequent EGR1 and FOS mRNA expression analysis by qPCR. k. Experimental design shows the workflow of the treatment performed by treating C57BL/6 or Hyp mice with BMS and/or FGF23. l-m. qPCR analysis of mouse Egr1, Fos, Cyp27b1, and Cyp24a1 following treatments.

Next, we focused on FGF23-mediated MAPK and NF-κB crossover signaling as supported by observations including: 1) increased NF-κB targets such as Tnfrsf12a (Figure 3k); 2) sustained enrichment of NF-κB TFs in response to FGF23 (Supplementary Data); and 3) the important translational impact of NF-κB signaling in renal diseases. To test the transcriptomic and genome accessibility responses to FGF23, we performed RNAseq and ATACseq experiments using HEK293-mKL cells treated with FGF23 (quality control metrics in Supplementary Figure S7a-S7c). To determine the correlation of the epigenetic landscape and mRNA expression, the ATACseq and RNAseq data were integrated and revealed higher chromatin accessibility within a 10kb promoter region were associated with elevated expression of FOSL1, NAB2, VDR, HBEGF, and NFKB1 (Figure 4g). Within the EGR1 genomic region, a transient increase of chromatin accessibility at a distal enhancer region was present (Supplementary Figure S7d-S7e). Thus, FGF23 may activate long range promoter-enhancer interactions to regulate EGR1. Functional analysis also identified activated FGFR1, NF-κB and MAPK signaling pathways (Figure 4h). In response to FGF23, among the genes associated with NF-κB signaling were CXCL8, RELB, and TNFRSF1A (Figure 4h). Specifically related to NF-κB signaling, a positive regulation of I-kappa B, involved in propagating the cellular response to inflammation, was predicted to be activated (Figure 4h).

To test the interrelationships between MAPK and NF-κB in the context of FGF23 signaling, a selective inhibitor of I-kappa B kinase phosphorylation,³³ BMS-345541, was used. Pretreatment of HEK293-mKL cells with BMS-345541 followed by FGF23 resulted in sustained FGF23-dependent MAPK bioactivity as evidenced by the increase of ERK1/2, c-JUN, and c-FOS phosphorylation, as well as enhancing c-FOS protein stabilization (Figure 4i). Further, NF-κB inhibition increased FGF23-dependent EGR1 and FOS mRNAs (Figure 4j). These data support that FGF23-induced NF-κB activity acts as a negative regulator of FGF23/KL-driven MAPK signaling. To test this hypothesis, C57BL/6J mice were treated with BMS-345541 for 1 h followed by FGF23 for 1 h which revealed that NF-κB inhibition promoted FGF23-induced renal Fos and Egr1 mRNAs as well as potentially suppressing 1,25D by decreasing Cyp27b1 mRNA when compared to mice treated with FGF23 alone (Figure 4l). Using an X-linked hypophosphatemia mouse model (Hyp mouse),³⁴ known to mimic human hypophosphatemia with elevated FGF23, BMS-345541 increased renal FOS and EGR1 mRNAs (Figure 4m) and normalized Cyp27b1 and Cyp24a1 transcription. Further, investigation of Slc34a1 protein by immunofluorescence in kidney cortex supported that a chronic treatment of normal mice with BMS-345541 may be required to reach the Slc34a1 decrease and activity observed in Hyp (Supplementary Figure S8a-S8b). These data support that NF-κB is a novel regulator of FGF23-dependent MAPK activity and kidney 1,25D metabolic enzymes.

TNF-related family members as novel KL-dependent FGF23 gene targets

Elevated FGF23 has been associated with pro-inflammatory states in CKD, thus it was next sought to isolate inflammatory biocomponents regulated by FGF23 bioactivity. A pronounced increase of TNF-associated genes following FGF23 delivery included elevated TNF receptors in the PT, DCT, and CNT, which corresponded with the expression of KL. In response to FGF23, maximal expression of Tnfrsf12a and Tnfrsf1a was detected at 4 h (Figure 5a; bulk mRNA, Figure 5b, left). In contrast, TNF ligands Tnfα and Tnfsf12 were not induced by FGF23 (Supplementary Figure S9a). The upregulation of Tnfrsf1a and Tnfrsf12a was restricted to PT-S1 and DCT (Supplementary Figure S9b-S9c), suggesting KL-and FGF23-dependent activity may regulate coordinated roles between epithelial and immune cells under inflammatory conditions.

Figure 5. KL-dependent FGF23 bioactivity regulates Tnfrsf12a.

a. Violin plots display the expression of Tnfrsf12a and Tnfrsf1a in PT-S1 and DCT. The kinetic curves show Tnfrsf12a mRNA expression in total kidney samples in male (blue) and female (red) mice treated with FGF23 for 0, 1, 4, 12, and 24h. b. In vitro analysis highlights the expression of TNFRSF12A, EGR1, and TNFRSF1A in HEK293 (black line) and HEK293-mKL (red line) cells treated with FGF23 for different times. c. Correlation analysis shows the positive interrelationship between EGR1 and TNFRSF12A. d. HEK293-mKL cells were treated with FGF23 (50 ng/ml) for four hours and RNAseq was performed. The heatmap graphic displays differential gene expression of TNFRs genes that change with FGF23 treatment. e. HEK293-mKL cells were treated with FGF23 (50 ng/ml) for 4 h and ATACseq was performed. Representative ATACseq peaks of cells treated with vehicle (top track) compared to cells treated with FGF23 (lower track) highlights the increase of chromatin accessibility across the TNFRSF12A gene body. f. HEK293-mKL cells were pretreated with the MEK inhibitor U0126 (5, or 10 μM) for 1 h prior to FGF23 administration (50 ng/ml) for 10 min; p-ERK immunoblot. g. HEK293-mKL cells were pretreated with U0126 (5, or 10 μM) for 1h prior to FGF23 (50 ng/ml) for 4 hours followed by RNA extraction and assessment of EGR1 and TNFRSF12A mRNAs by qPCR. h. HEK293-mKL cells were pretreated with BMS-34554 (10 μM) for 1h followed by FGF23 (50 ng/ml) for 4h then TNFRSF12A and CXCL8 mRNA expression analysis. i. HEK293-mKL cells were treated with 50 ng/ml of FGF23 for 0, 1, 2, 4, 8, and 24 h. The mRNA expression of CCL5, CXCL8, and IL6 obtained from in vitro studies were measured by qPCR. The red kinetic curve represents the HEK293-mKL cell studies whereas the black kinetic curve shows the data from HEK293 cell line experiments. j. HEK293-mKL cells were treated with the ligand of TNFRSF12A (TNFSF12, TWEAK 100 ng/ml) for 1h followed by FGF23 treatment for 4h and pERK immunoblot. k. HEK293-mKL cells were pretreated with TWEAK (100 ng/ml) for 1h followed by FGF23 (50 ng/ml) for 16 h. EGR1 expression was then tested by qPCR to assess FGF23 bioactivity. l. In vivo studies were performed by treating C57BL/6 mice with recombinant mouse TWEAK for 4 days at the rate of one injection per day. On the fourth day mice received FGF23 treatment for 4 h and Tnfrsf12a was assessed by qPCR. m-p. HEK293-mKL cells were pretreated with TNF (100 ng/ml) for 16 h followed with FGF23 (50 ng/ml) for 4 h. pNF-κB, total NF-κB, and β-actin were evaluated by immunoblot (m). The mRNA expression of NFKB (n), CXCL8 (o), and EGR1 (p) were assessed by qPCR.

To test whether TNFRs are direct target genes of KL-dependent FGF23 bioactivity, parent HEK293 cells and HEK293-mKL cells were treated with FGF23. In response to FGF23, TNFRSF12A and TNFRSF1A (Figure 5b, right) were increased in HEK293-mKL cells versus HEK293 cells (Figure 5b, right). At 4 h, TNFRSF12A mRNA was positively correlated with EGR1 expression, highlighting TNFRSF12A as a potential novel FGF23 target gene (Figure 5c). By screening the TNFRs that were differentially expressed with FGF23 treatment in the RNAseq dataset, it was found that TNFRSF1A, NGFR, TNFRSF12A, TNFRSF21, RELT, TNFRSF9, and TNFRSF10B were significantly increased (Figure 5d). As highlighted in Figure 5e, chromatin accessibility for TNFRSF12A was increased with FGF23 treatment.

To assess whether the regulation of TNFRSF12A by FGF23 was MAPK-dependent as well as downstream of pERK signaling, HEK293-mKL cells were pretreated with the MEK inhibitor U0126 followed by FGF23. As assessed by immunoblot, pERK was inhibited in U0126-treated cells (Figure 5f) and confirmed as associated with EGR1 mRNA suppression (Figure 5g). Importantly, MEK inhibition suppressed FGF23-dependent TNFRSF12A mRNA (Figure 5g). These findings implicate the KL-MAPK axes in the regulation of TNF-associated pathways. Further, inhibition of NF-κB using BMS-345541 suppressed FGF23-induced TNFRSF12A and CXCL8 (Figure 5h). Interestingly, TNFRSF12A and NF-κB target genes CXCL8, CCL5, and IL6 were upregulated in response to chronic FGF23 treatment (24 h; Figure 5i), during which the MAPK target EGR1 returned to baseline (Figure 4d).

Because FGF23 induced TNFRs, the presence of a coordinated crosstalk between FGF23 and TNF to regulate FGF23 bioactivity was hypothesized. To isolate TNF signaling on FGF23 bioactivity, HEK293-mKL cells were treated with TWEAK followed by FGF23. Activation of TNFRSF12A suppressed FGF23-induced ERK1/2 phosphorylation (Figure 5j), and enhanced EGR1 (Figure 5k). To test the effects of epithelial TWEAK signaling in vivo, C57BL/6J mice were treated daily with TWEAK for 4 days followed by 4 h FGF23 delivery. In confirmation of the scRNAseq, FGF23 increased renal Tnfrsf12a, however pretreatment with TWEAK suppressed FGF23-mediated Tnfrsf12a (Figure 5l). There were no effects on Slc34a1 protein following 4 h FGF23 treatment (Supplementary Figure S10a-S10b). To isolate whether TNFα may play similar roles as TWEAK on FGF23 bioactivity, HEK293-mKL cells were treated with TNFα followed by FGF23. Cells receiving FGF23 had increased phosphorylation of NF-κB (p65; Figure 5m) consistent with predicted NF-κB activation. FGF23 and TNF co-treatment led to an additive effect with further increased NF-κB phosphorylation and associated elevation of NF-κB targets NFKB and CXCL8, and decreased EGR1 (Figure 5n-5p). Thus, TWEAK and TNFα via NF-κB suppressed FGF23-induced MAPK activity, supporting that impairment of FGF23 bioactivity can occur during prevailing inflammatory-related signaling.

In sum, our findings identified key molecular events involved in renal FGF23 bioactivity at the single cell level. KL expression drove cis functional responses in PT and DT cells, including overlapping and unique signaling events within these nephron segments. We also identified novel functional segment-specific pathways and identified FGF23-mediated transcription factor regulon induction. Finally, crossover NF-κB and MAPK-dependent interactions may potentiate TNFR signaling and restrict FGF23 cellular responses, potentially critical for mineral metabolism during disease states.

Discussion

The cell-specific actions of FGF23 via its co-receptor KL are largely unknown, and to date have been solely based upon assumptions of segment-specific candidate gene expression. A primary goal herein was to leverage scRNAseq capabilities to facilitate the identification of FGF23-dependent target genes that may impact kidney disease. Our novel “sex-hashing” design demonstrated that FGF23 responses elicited in males and females were similar. Further, transcriptomic data at the single cell level showed that the control of phosphate and 1,25D metabolism by FGF23 requires a temporally coordinated set of nephron cell-specific programs. To this end, signaling regulons were identified in PT and DT segments associated with KL-dependent FGF23 bioactivity. Using a temporal approach, we found that FGF23 rapidly induced known and unique gene sets, some associated with MAPK signaling,³⁵ prior to the regulation of transcripts associated with the control of 1,25D metabolizing enzymes in PT-S1, but not in DCT/CNT. These results support that in response to acute FGF23 treatment, FGF23 induced distinctive ‘‘early’- and ‘late-stage’ signaling events. Further exploration of these associations through pairing scRNAseq with scATACseq could provide additional insight into FGF23 function by identifying cell type-specific chromatin accessibility changes.

The FGF23 co-receptor KL is required for high affinity FGF23 bioactivity.¹³ KL was transcribed at highest levels in the homogeneous DCT/CNT compared to the levels observed in more heterogenous PTS1-S3 segments. PT cells had varying KL mRNA expression, consistent with prior molecular mapping of the nephron by bulk analysis of isolated tubule sections.^9,29 Our findings suggest that PT segments have a broad but lower KL expression, however whether this provides the ability to be more rapidly adaptive to minute-to-minute changes in FGF23 and phosphate/1,25D handling remains to be determined. To test the importance of KL for FGF23 bioactivity, sub-setting analysis was employed based upon single cell KL mRNA expression levels. We found that in response to FGF23, KL^high cells exhibited a marked and parallel phenotypic change in gene expression, whereas KL^low cells showed an unchanged transcriptional profile, and clustered more closely with untreated cells. Moreover, cells that responded to FGF23 segregated more closely to each other due to similarly induced gene regulatory control than at the original initiation point of being defined by nephron segment cell type. In the broader sense, as opposed to neutralization of circulating FGF23, it has been therapeutically difficult to directly target FGF23-responsive pathways in kidney. Thus, our findings support the idea that different targeting strategies may have to be employed depending upon the temporal gene regulation across nephron segments.

Through pathway analysis, inflammatory responders including TNFRSF12A, were identified as novel receptors controlled by FGF23, and in parallel with KL localization, were induced primarily in PT, DCT, and CNT. Transient activation of TNFRSF12A promotes tissue repair following acute injury,³⁶ however prolonged TNFSF12/TNFRSF12A signaling initiates tissue damage and degeneration.³⁷ Although FGF23 induced TNFRs, the corresponding ligands were not up-regulated in any kidney cell, in agreement with previous work showing that TNFSF12 is primarily produced in immune cells.³⁸ Thus, our findings suggest that FGF23 controls responses in KL-expressing cells, and through these actions may influence phenotypes from cells that are not direct targets of FGF23. Indeed, NF-κB inhibition completely abolished TNFRSF12A up-regulation by FGF23, and in parallel enhanced MAPK-specific signaling and transcriptional events, consistent with the idea that NF-κB is a novel negative regulator of FGF23-dependent MAPK activity. It is thus plausible that chronic inflammation associated with increased NF-κB could inhibit FGF23-dependent mineral metabolism. KL has been shown to be downregulated in CKD,³⁹ leading to the hypothesis that CKD is a state of FGF23 tissue resistance. Whether these alterations in MAPK/NF-κB signaling also contribute to dysfunctional FGF23 bioactivity prior to, and during, reductions in KL remains to be determined.

Our studies had several limitations, including focus upon relatively acute aspects of FGF23 signaling. This approach was primarily undertaken to reduce secondary effects due to sustained changes in systemic mineral metabolism that may occur with prolonged FGF23 delivery. Therefore, expanded studies could test the chronic effects of FGF23 towards understanding how the acute signaling mechanisms found herein may be compensated in the longer term. Further, our work confirmed known reductions in proteins such as Npt2a after FGF23 delivery, but proteomic analysis combined with transcriptional datasets could reveal how the identified mRNA changes are reflected at translational levels for FGF23-dependent genes that may rely on stabilizing or destabilizing mechanisms of protein turnover.

In summary, this work identified new sets of FGF23 target genes with physiological and potentially pathological functions at the single cell level. Our studies thus provide critical insight into the mechanisms of FGF23 actions, which drove transcriptional reprogramming in KL-positive cells. Further, FGF23 initiated common upstream signaling pathways in PT and DT cells, followed by segment-specific actions leading to differential control of mineral ion handling. FGF23 also induced ‘early’ and ‘late’ phase transcriptional events, which revealed that NF-κB signaling suppressed MAPK-dependent activity (see model, Figure 6). Collectively, these novel FGF23 functions may provide key junction points for interventions in diseases of mineral metabolism.

Figure 6. Proposed model.

The graphic shows a proposed model of crosstalk between KL-dependent FGF23 bioactivity and inflammation-induced NF-κB signaling.

Translational statement.

Inflammation and elevated FGF23 in diseases such as chronic kidney disease (CKD) are associated with mortality, however inflammatory effects on FGF23-mediated mineral metabolism within specific nephron segments remain unclear. Herein, we identified transcriptional reprogramming within the nephron during FGF23-mediated signaling and isolated TNF/NF-κB pathways associated with regulating FGF23 bioactivity. These findings may be important in designing future therapeutic approaches for mineral diseases, including potential combination or early intervention treatments. Further studies could test translation of our findings to interactions of chronic inflammation and elevated FGF23 in human CKD.

Data availability

scRNAseq data is deposited in the NCBI’s Gene Expression Omnibus database (GEO GSE246385). The scATACseq analyzed in this paper were extracted from the GEO repository, accession number GSM6443123. The bulk RNAseq and ATACseq data presented are deposited in the GEO repository, accession number GSE254541.