Distinct developmental reprogramming footprint of macrophages during acute kidney injury across species

Abstract

Acute kidney injury (AKI) is a common finding in hospitalized patients, particularly those who are critically ill. The development of AKI is associated with several adverse outcomes including mortality, morbidity, progression to chronic kidney disease, and an increase in healthcare expenditure. Despite the well-established negative impact of AKI and rigorous efforts to better define, identify, and implement targeted therapies, the overall approach to the treatment of AKI continues to principally encompass supportive measures. This enduring challenge is primarily due to the heterogeneous nature of insults that activate many independent and overlapping molecular pathways. Consequently, it is evident that the identification of common mechanisms that mediate the pathogenesis of AKI, independent of etiology and engaged pathophysiological pathways, is of paramount importance and could lead to the identification of novel therapeutic targets. To better distinguish the commonly modulated mechanisms of AKI, we explored the transcriptional characteristics of human kidney biopsies from patients with acute tubular necrosis (ATN), and acute interstitial nephritis (AIN) using a NanoString inflammation panel. Subsequently, we used publicly available single-cell transcriptional resources to better interpret the generated transcriptional findings. Our findings identify robust acute kidney injury (AKI-induced) developmental reprogramming of macrophages (MΦ) with the expansion of C1Q⁺, CD163⁺ MΦ that is independent of the etiology of AKI and conserved across mouse and human species. These results would expand the current understanding of the pathophysiology of AKI and potentially offer novel targets for additional studies to enhance the translational transition of AKI research.

NEW & NOTEWORTHY Our findings identify robust acute kidney injury (AKI)-induced developmental reprogramming of macrophages (MΦ) with the expansion of C1Q⁺, CD163⁺ MΦ that is independent of the etiology of AKI and conserved across mouse and human species.

INTRODUCTION

Acute kidney injury (AKI) is a frequently encountered clinical condition with well-documented adverse clinical and expenditure outcomes (1, 2). Despite wide-ranging causes, almost all forms of kidney diseases directly or indirectly involve an immune component, with macrophages (MΦ) playing a central role (3). Acute tubular necrosis (ATN) and acute interstitial nephritis (AIN) are common causes of AKI that can often be accurately diagnosed only with kidney biopsy. To investigate the inflammatory pathways that are involved in the pathogenesis of these two entities of AKI, we explored the transcriptional characteristics of human kidney biopsies from patients with ATN, and AIN using a 255-gene NanoString inflammation panel. As controls, we used kidney biopsies from patients with thin basement membrane (TBM) nephropathy, a condition that typically does not alter kidney function or its future trajectory. We also elucidated these data using publicly available single-cell transcriptional resources to better interpret the generated transcriptional findings.

MATERIALS AND METHODS

Sample Selection

Remnant human kidney biopsies were obtained through the tissue resource available from the UAB-UCSD O’Brien Center for AKI Research. These samples were collected, de-identified, archived together, and analyzed according to protocols approved by the Institutional Review Board of the University of Alabama at Birmingham (UAB). Biopsy samples for this study were selected based on the sole diagnosis of ATN, AIN, or TBM, according to pathologists’ diagnoses and comments. Any other concomitant reported pathological manifestations, such as the presence of fibrosis, arteriosclerosis, diabetic changes, etc., were used to exclude potential samples. As controls, we used kidney biopsies from patients with thin basement membrane (TBM) disease. Diagnosis of TBM was confirmed by electron microscopy, and thickness measurement values used to establish the diagnosis of TBM were as follows: 337 ± 75 nm for females and 355 ± 75 nm for males (Supplemental Fig. S1). Notably, it is now established that TBM should not be considered an entirely benign condition, given some evidence suggesting an increased risk of kidney disease (4). Notwithstanding, the remnant tissue samples used in this study did not reveal any noticeable pathological features apart from the sole diagnosis of TBM.

RNA Isolation and NanoString Analysis

We used three to eight pieces of 10-µM-thick sections of formalin-fixed, paraffin-embedded (FFPE) samples that were put into a sterile, RNase-free microcentrifuge tube. RNase Away was used on the microtome blade between each block to avoid contamination. The PureLink FFPE total RNA isolation kit (Invitrogen, Carlsbad, CA) was used as per the manufacturer’s instructions. Briefly, melting buffer was added to each tube and incubated for 10 min at 72°C to melt the paraffin. Proteinase K was added and samples were incubated for 1 h at 60°C until no visible tissue was seen. Subsequently, the solution was centrifuged and removed from the paraffin layer, combined with binding buffer and ethanol, and bound to a spin cartridge, where it was washed and eluted in 30-µL RNase-free water. The samples were further concentrated using the RNA Clean & Concentrator kit (Zymo Research Corp, Irvine, CA) according to the manufacturer’s protocol without DNase I digestion. Samples were quantified using a DeNovix DS-11 Spectrophotometer (Wilmington, DE), and 100 ng of RNA was used per reaction whenever possible. RNA was hybridized overnight to Reporter and Capture probes of the human Immunology panel (NanoString Technologies, Seattle, WA). This premade panel contains 249 genes involved in inflammation and 6 housekeeping genes, which were used to normalize the data. After the samples are attached to the cartridge and aligned, they are read by the digital analyzer, and subsequently, the data were analyzed using the nSolver 4.0 program (NanoString Technologies, Seattle, WA).

Statistical Analysis of NanoString Data

We controlled technical variability by using standard normalization techniques and applying a sample-specific correction factor derived from the geometric mean expression of reference genes (CLTC, GAPDH, GUSB, HPRT1, PGK1, and TUBB). Gene expression data were analyzed using the R package limma voom, suitable for NanoString data (5). To address the uncertainty associated with low-level gene expression, a cutoff threshold of 12 was set to exclude low expressers, with genes having more than two samples below this cutoff in all three groups filtered out. Following this, limma-voom was applied for both overall and pairwise tests, specifically comparing AIN versus TBM, ATN versus TBM, and ATN versus AIN. The analyses yielded fold changes, P values, and false discovery rates (FDR) for each comparison, with additional reporting of F statistics and P values for both overall and pairwise tests, enhancing the comprehensive gene expression data analysis.

Interpretation of Gene Expression Data Using Publicly Available Transcriptomic Platforms

We used three publicly available resources for transcriptomic analysis. These include the following. 1) Kidney Precision Medicine Project (https://www.kpmp.org), a project funded by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). Data were downloaded between July 2023 and September 2023. See the grants for more information. 2) Kidney interactive transcriptomics platform (http://humphreyslab.com/SingleCell/), developed by Dr. Humphreys. The generated results in this study used a data source that was previously published (6–8). 3) Susztak Laboratory Kidney Biobank (http://www.susztaklab.com) developed by Dr. Susztak. The results generated in this study used a data source that was previously published (9, 10). Data presented in this study received written permission from all three resources mentioned earlier. Specifically, Fig. 2A and Supplemental Figs. S2, D and E, and S4, A–C were generated using the Kidney Precision Medicine Project. Figures 2C, 3, and 4, and Supplemental Fig. S2B, were generated using kidney interactive transcriptomics. Figure 2, B and C, and Supplemental Figs. S2, A and C, and S3, were created utilizing the Susztak Laboratory Kidney Biobank.

Figure 2.

Transcriptional analysis of C1QA, C1QB, and CD163 expression using a publicly available online analyzer for kidney transcriptome datasets. A: enrichment of the selected genes in healthy individuals vs. AKI in monocyte/MΦ component of the UMAP. The UMAP algorithm reduces high-dimensional scRNA-Seq data to a low-dimensional plot that retains much of the original information. B: differential expression of specified genes in healthy human kidneys vs. those diagnosed with kidney disease. C: the expression level of stated genes in developing mouse kidneys. Only cells annotated into MΦ are depicted. D: demonstration of high C1QA, C1QB, and CD163 gene expression levels in human fetal kidney (week 17). AKI, acute kidney injury.

RESULTS

AIN and ATN Trigger Similar Transcriptional Changes in Human Kidneys

The three studied groups, AIN, ATN, and TBM, were each represented by kidney biopsies from five patients. The average maximum serum creatinine in the AIN group was 4.2 ± 2 mg/dL, 8.0 ± 6 mg/dL in the ATN group, and 1.0 ± 0 mg/dL in the TBM group (see details and additional demographic data in Supplemental Table S1). The differences in serum creatinine may be related to various factors including gene enrichment patterns seen in these conditions. We analyzed kidney biopsies from patients with AKI due to histologically confirmed AIN, ATN, and control biopsies with TBM (Fig. 1A) using the 255-gene nCounter Human Inflammation V2 Panel (NanoString Technologies). TBM diagnosis was confirmed using electron microscopy (Supplemental Fig. S1). These analyses revealed 53 differentially expressed genes in AIN (vs. TBM) kidney biopsies and 18 differentially expressed genes in ATN (vs. TBM) biopsies (Fig. 1, B–D). Remarkably, almost all genes (17 of 18) that were differentially expressed in the ATN group were among the significantly modulated genes in AIN (Fig. 1B). Such a substantial overlap points to a common type of immune response in AIN and ATN. The entire list of top modulated genes in AIN versus TBM and ATN versus TBM is presented in Supplemental Tables S2 and S3, respectively.

Figure 1.

NanoString inflammatory analysis of kidney biopsies with a confirmed diagnosis of ATN, AIN, and TBM. A: images demonstrate human kidney biopsy specimens stained by the periodic acid-Schiff stain method. Scale bar = 100 µM. B: illustration of the overlap of 53 differentially expressed genes in AIN (vs. TBM) kidney biopsies and 18 differentially expressed genes in ATN (vs. TBM), P < 0.05, FDR < 0.05. C: hierarchical clustering of genes that are significantly altered in AIN (vs. TBM) and (D) in ATN (vs. TBM) group. Heatmaps display z-transformed expression values. E: relative C1QA, C1QB, and CD163 expression based on NanoString data analysis, fold vs. TBM, n = 5/group. Data presented as means ± SE. *P < 0.05 and FDR < 0.05. ATN, acute tubular necrosis; AIN, acute interstitial nephritis; FDR, false discovery rates; TBM, thin basement membrane.

Developmental Switch of Kidney-Resident Macrophages Driven by C1Q, CD163 Across Species

A major limitation of the abovementioned approach is that it uses average transcriptional levels of the analyzed tissues and does not discriminate among potentially divergent transcriptional responses of individual cell types. Hence, to better delineate the significance of the commonly enriched genes in ATN and AIN, we then analyzed the 17 outlined common enriched genes (Fig. 1B) using publicly available single-cell RNA sequencing (scRNA-Seq) platforms as described in materials and methods. Such analyses revealed robust AKI-induced enrichment of multiple genes whose expression has been described as a hallmark of kidney-resident MΦ, including those encoding C1Q subunits (C1QA and C1QB) and CD163 (11, 12). Indeed, we found that these genes were significantly upregulated in our NanoString analysis of ATN and AIN groups (Fig. 1E). Notably, although the NanoString panel used in this study did not specifically target C1QC gene expression, we performed a similar analysis as shown in Supplemental Fig. S2 to examine the expression levels of C1QC. As expected, the enrichment and expression patterns of C1QC were almost identical under all examined conditions to those of C1QA and C1QB (Supplemental Fig. S2). Using the Kidney Precision Medicine Project data resource, we compared scRNA-Seq of healthy individuals versus those diagnosed with AKI and found enrichment of C1QA, C1QB, and CD163 (Fig. 2A). Note, Fig. 2A depicts the monocyte/MΦ subpopulation that was cropped from the reference UMAP for better illustration. The full reference UMAP is shown in Supplemental Fig. S4. Next, we used another publicly available transcriptional resource generated by Dr. Susztak’s team (see materials and methods for more details). This approach recapitulated our findings when comparing samples obtained from healthy individuals and those with kidney disease (Fig. 2C). Note, Fig. 2C depicts the monocyte/MΦ subpopulation that was cropped from the reference UMAP for better illustration. The full reference UMAP is shown in Supplemental Fig. S3 and is also presented in the original publication (10). Given the recently emerging body of evidence that suggests kidney-resident MΦ population undergoes developmental reprogramming during AKI, an emerging pivotal repair phase of inflammation, we next sought to investigate how the aforementioned enriched genes fit into a developmental state in mouse and human kidneys. Indeed, we demonstrate that developing kidneys across species (human and mouse) show a striking similarity in the enrichment of C1QA, C1QB, and CD163 (Fig. 2, B and D).

Next, we asked whether these results would be applicable to establish mouse models of AKI. We addressed this question by analyzing another transcriptional data set platform generated by Dr. Humphreys’ team (details in materials and methods) and found that genes encoding the C1Q subunits and CD163 are enriched in MΦ populations following AKI in a mouse model of ischemia-reperfusion-induced injury (Fig. 3). Moreover, a similar pattern emerged when examining a human allograft-rejecting specimen (Fig. 4).

Figure 3.

Mouse model of ischemia-reperfusion-induced AKI reveals similar developmental reprogramming of monocyte/MΦ. Panels demonstrate dynamic induction of labeled genes at different time points following injury. AKI, acute kidney injury.

Figure 4.

Expression of C1Q subunits (C1QA, C1QB, C1QC) and CD163 in rejecting human allograft kidney. Arrows indicate monocyte/MΦ expression of designated genes rejecting allograft kidneys.

DISCUSSION

Despite the well-established negative impact of AKI and rigorous efforts to better define, identify, and implement targeted therapies, the overall approach to treatment of AKI continues to principally encompass supportive measures. This enduring challenge is primarily due to the heterogeneous nature of insults that activate many independent and overlapping molecular pathways. Consequently, it is evident that the identification of common mechanisms that mediate the pathogenesis of AKI, independent of etiology and engaged pathophysiological pathways, is of paramount importance and could lead to the identification of novel therapeutic targets. In this study, we sought to address this challenge with the coalescing objective of identifying common genes and inflammatory pathways that are modulated during AKI, irrespective of species, etiology, and histological manifestations. Using the 255-gene NanoString inflammation panel, we examined interrelated inflammatory pathways activated during ATN and AIN using biopsies from human specimens with a sole diagnosis of ATN, AIN, and TBM based on renal pathologists’ comments. We demonstrate a distinct set of inflammation-associated genes that are significantly modulated in both ATN and AIN conditions. Our findings deliver two key observations. 1) expression of genes encoding C1Q subunits (C1QA and C1QB) and CD163 emerged as a reliable marker of kidney-resident MΦ as evidenced by their expression during fetal kidney development across species. 2) Our results unravel a fundamental developmental reprogramming with the expansion of C1Q⁺, CD163⁺ MΦ that is independent of the etiology of AKI and conserved across mouse and human species.

Although numerous etiologies have been implicated, almost all forms of kidney diseases engage an immune component with MΦ playing a central role (3). MΦ comprise a unique, heterogeneous cell population with a high degree of phenotype plasticity. They play key roles in response to a wide spectrum of environmental signals to either contribute to the functional status of parenchymal cells or, in the event of tissue injury, coordinate tissue repair and regeneration (13, 14). Such versatility enables gene expression patterns of MΦ to rapidly adapt in response to the dynamics of the microenvironment. MΦ expansion is a common finding in both preclinical AKI models and human biopsies with a direct correlation between the degree of MΦ infiltration and the severity of kidney injury (15, 16). There is compelling evidence that indicates a crucial role for MΦ in instigation and propagation of kidney disease and hence modulating MΦ phenotype has emerged as a potential novel therapeutic target. Maintaining tissue homeostasis is the cardinal functionality of tissue-resident MΦ (11). Molecular landscapes that are generated during organ injury and their similarities to the organ developmental phase have gained considerable attention recently. It is proposed that the developmental switch of resident MΦ is a hallmark of engagement of molecular machinery with the overall objective of constructing the necessary signals for repair and resolution of inflammation (17).

An evolutionary conserved factor, C1Q, is involved in regulating many physiological and immunomodulatory processes. These include, but are not limited to, activation of the classical complement pathway, orchestration of the response to various inflammatory stimuli, regulation of cell differentiation, and removal of apoptotic bodies, among others (18, 19). The paramount homeostatic function of C1Q is evident in its deficiency state, where lack of self-tolerance is coupled with the development of autoimmunity and susceptibility to infections (18, 19). Moreover, the proangiogenic and prorepair properties of C1Q implicate it as an important determinant of outcomes following AKI (20).

As a marker of a subset of tissue-resident MΦ, CD163 has been extensively studied in the context of endocytosis and clearance of hemoglobin-haptoglobin complexes. Completing this task is shown to exert anti-inflammatory properties, which are mediated in part by the induction of heme oxygenese-1 (21, 22). Notwithstanding this protective task, CD163 also serves as an innate immune effector by recognizing bacteria and viruses. CD163⁺ cells are known to exert anti-inflammatory effects and are being investigated as potential therapeutic targets given their abundance at the site of inflammation (23, 24). These observations underscore the critical role of CD163⁺ cells during tissue injury and ensuing repair efforts and further validate their prospective as a potential therapeutic target (25).

The lack of significant clinical breakthroughs has increased the scrutiny surrounding the design, conduct, and reproducibility of preclinical models of AKI. Several strategies have been proposed to better formulate preclinical AKI models that efficaciously exhibit key features of the human AKI microenvironment enabling identification of germane therapeutic targets (26, 27). Our findings demonstrate a reassuring set of results that mimic and validate key aspects of the molecular underpinnings that occur during human AKI (Fig. 3). These results underscore the invaluable contribution of preclinical AKI models to expand our understanding of the dynamics associated with instigation of AKI and ensuing activation of a cascade of molecular events during the resolution and repair phase of inflammation.

Taken together, we demonstrate an explicit developmental reprogramming of MΦ during AKI that is independent of AKI etiology. The consistent pattern of our results across species suggests a conserved evolutionary strategy that may be pivotal for the reconstitution of tissue homeostasis following injury and may provide novel therapeutic targets for a clinical syndrome that is immensely heterogeneous in terms of etiology and engagement of various inflammatory pathways.

DATA AVAILABILITY

Data are available upon request from the authors.

SUPPLEMENTAL DATA

Supplemental Figs. S1–S4 and Tables S1–S3: https://doi.org/10.17632/x4nmtwwwxh.1.

Raw and normalized NanoString data are uploaded on Mendeley Data and Digital Commons Data website: https://doi.org/10.17632/m7738bc5rg.1.

GRANTS

This work was supported by National Institutes of Health (NIH) Grant DK134402 (to A.Z.); M.M. was supported in part by the NIH-funded PKD Research Resource Consortium (U54DK126087) and grants from the Office of Research and Development, Medical Research Service, Department of Veterans Affairs (I01BX004232 and I01BX006266). Results here are in part based upon data generated by the Kidney Precision Medicine Project: DK133081, DK133091, DK133092, DK133093, DK133095, DK1330971, DK114866, DK114908, DK133090, DK133113, DK133766, DK133768, DK114907, DK114920, DK114923, DK114933, and DK114886. This work was funded in part by the UAB-UCSD O’Brien Center for Acute Kidney Injury Research (NIH Grant U54DK137307).

DISCLOSURES

M.M. reports grants and consulting fees outside the submitted work from Otsuka Pharmaceuticals, Sanofi, Palladio Biosciences, Reata, Natera, Chinook Therapeutics, Goldilocks Therapeutics, Carraway Therapeutics, and Vertex Pharmaceuticals.

AUTHOR CONTRIBUTIONS

A.A. and A.Z. conceived and designed research; F.R. performed experiments; M.M., E.M., J.C., X.C., and A.Z. analyzed data; M.M., E.M., F.R., J.C., and X.C. interpreted results of experiments; M.M. and E.M. prepared figures; E.M. and A.Z. drafted manuscript; M.M., E.M., F.R., J.C., X.C., A.A., and A.Z. edited and revised manuscript; M.M., A.A., and A.Z. approved final version of manuscript.