Cellular senescence in kidney biopsies is associated with tubular dysfunction and predicts CKD progression in childhood cancer patients with karyomegalic interstitial nephropathy

Karyomegalic interstitial nephropathy (KIN) has been reported as an incidental finding in patients with childhood cancer treated with ifosfamide. It is defined by the presence of tubular epithelial cells (TECs) with enlarged, irregular, and hyperchromatic nuclei. Cellular senescence has been proposed to be involved in kidney fibrosis in hereditary KIN patients. We report that KIN could be diagnosed 7–32 months after childhood cancer diagnosis in 6/6 consecutive patients biopsied for progressive chronic kidney disease (CKD) of unknown cause between 2018 and 2021. The morphometry of nuclear size distribution and markers for DNA damage (γH2AX), cell‐cycle arrest (p21+, Ki67−), and nuclear lamina decay (loss of lamin B1), identified karyomegaly and senescence features in TECs. Polyploidy was assessed by chromosome fluorescence in situ hybridization (FISH). In all six patients the number of p21‐positive TECs far exceeded the typically small numbers of truly karyomegalic cells, and p21‐positive TECs contained less lysozyme, testifying to defective resorption, which explains the consistently observed low‐molecular‐weight (LMW) proteinuria. In addition, polyploidy of TEC was observed to correlate with loss of lysozyme staining. Importantly, in the five patients with the largest nuclei, the percentage of p21‐positive TECs tightly correlated with estimated glomerular filtration rate loss between biopsy and last follow‐up (R2 = 0.93, p < 0.01). We conclude that cellular senescence is associated with tubular dysfunction and predicts CKD progression in childhood cancer patients with KIN and appears to be a prevalent cause of otherwise unexplained CKD and LMW proteinuria in children treated with DNA‐damaging and cell stress‐inducing therapy including ifosfamide. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
Chronic kidney disease (CKD) in childhood cancer patients typically shows a tubular injury phenotype with high urine output and low-molecular-weight (LMW) proteinuria.Karyomegalic interstitial nephritis/nephropathy (KIN) has been described as a rare cause of CKD in ifosfamide-treated patients [1].KIN is defined by tubular cell degeneration featuring abnormally large irregular and hyperchromatic nuclei [2,3].It has been proposed that cellular senescence due to compromised DNA damage repair might contribute to kidney fibrosis in hereditary KIN patients with mutations in the FANCD2/FANCIassociated nuclease 1 (FAN1) gene and in humans and experimental animals exposed to ochratoxin A, busulfan, or pyrrolizidine alkaloids [2,4,5].
To explore the prevalence of ifosfamide induced early KIN and the role of cellular senescence, we analyzed the biopsies and clinical histories of six consecutive patients from the pediatric nephrology department of the Wilhelmina Children's Hospital, University Medical Center (UMC) Utrecht, the Netherlands, between 2018 and 2021 with unexplained and progressive loss of kidney function after ifosfamide treatment for pediatric cancers.

Patients and biopsies
All children undergoing a kidney biopsy at the Wilhelmina Children's Hospital, UMC Utrecht, between 2018 and 2021 for progressive decline of kidney function after ifosfamide treatment were included.The biopsies were taken 7-32 months (median 20 months) after initiation of ifosfamide treatment for various childhood cancers at the Princess Máxima Center for Pediatric Oncology.Biopsies were reviewed by two expert renal pathologists (TN and RG) using standard light microscopy, immunofluorescence, and electron microscopy, unless otherwise specified.Additional staining and morphometry were performed (see following discussion).For comparison, six protocol kidney allograft biopsies without any pathological findings taken 3 months after kidney transplantation were included.Clinical data were retrieved from the electronic patient files.
All patients or their legal representatives consented to undergo the medical procedures and additional staining to review the specific features of KIN.
Fluorescence in situ hybridization (FISH) was performed and imaged as previously described using the Prenatal 13, 18, and 21 Enumeration probe kit (CytoCell, Begbroke, UK) [6].Scoring was accomplished manually in at least 300 cells.Nuclei with more than two FISH signals for any of the three tested chromosomes were scored as polysomic.

Morphometry and image analysis
Whole slides were obtained with NanoZoomer scanners (Hamamatsu, Japan), and selected images were acquired by taking snapshots in ImageScope.Nuclear size was measured using QuPath version 0.2.3 [7], which automatically detected nuclear outlines and calculated the nuclear surface of total available cortical area in three H&E sections per biopsy.The number of p21-positive cells was automatically counted with QuPath (positive cell detection) in one entire cross section of the available cortex in the biopsy.The cortical surface area staining for lysozyme was quantified using the QuPath pixel classifier.For quantifying p21 and Ki67, double staining fluorescent images of the entire biopsy cross section were automatically quantified using QuPath.γH2AX was manually counted for the entire cross section of the biopsy as the foci were too small to detect automatically.

Statistical analyses
Cell distribution analysis, percentile calculations, and statistics (Mann-Whitney test, simple linear regression, and the Wilson/Brown method for receiver operating characteristic curve [ROC] curves) were performed on exported measurements in R (cran.r-project.org,last accessed October 2019) and GraphPad Prism version 8.3.0 (GraphPad Inc., San Diego, CA, USA).

Case reports
Clinical information is summarized in Table 1 and Figure 1A, and a detailed individual patient history is described in supplementary material, Appendix A. One patient had neuroblastoma, three had Ewing sarcoma, one embryonal rhabdomyosarcoma, and one Wilms tumor.Age at diagnosis ranged from 3.2 to 16.2 years.
All patients had a normal kidney function at diagnosis of malignancy.Cumulative ifosfamide doses were between 12 and 100 g/m 2 .In addition, one patient received cisplatin, and three patients received cyclophosphamide.Four patients (two Ewing sarcoma, one neuroblastoma, and one Wilms tumor) had received radiotherapy including the kidney region, with a mean kidney dose range of 0.2-26 Gray.At latest follow-up (22-69 months after diagnosis of malignancy), kidney function declined in all patients, reaching CKD stage 2 (n = 3), Stage 3 (n = 2), or stage 4 (n = 1).All had LMW proteinuria.Blood pressure was normal in all patients.The clinical picture was consistently that of tubulointerstitial nephropathy.Two patients received steroids, for inflammation observed in the biopsy.Kidney function initially improved in case 1 but progressed to CKD stage 4 despite steroid treatment.In case 3, kidney function stabilized at CKD stage 2. Of the four patients that did not receive steroid treatment, cases 2 and 4 stabilized (at CKD stage 2 and 3a), while cases 5 and 6 progressed to CKD Stage 3b and 2 at latest follow-up.

Identification of KIN by automated assessment of nuclear size and correlation with kidney function
Tubular epithelial cells (TECs) with enlarged nuclei are the hallmark of KIN [8].The initial diagnosis of KIN in five of our six patients was evident at standard microscopic assessment, but the biopsy of one patient (case 6) initially appeared normal on routine evaluation (Figure 1B).To validate these findings, nuclear sizes were measured using QuPath.Although the mean nuclear size in KIN and protocol biopsies was comparable (18.83 and 19.87 μm 2 , respectively), there was a clear difference between groups regarding the size of the largest nuclei (Figure 2B,C), with a small subpopulation in all six biopsies showing abnormal enlargement.ROC analysis revealed that all six biopsies could be distinguished from protocol biopsies by comparing the 99.9th percentile of largest nuclei (supplementary material, Figure S1), while case 3 was still misclassified when comparing the 99th percentile, and discriminative value was completely lost when comparing the 10% largest nuclei.The 99.9th percentile of nuclear size trended to associate with urinary β2-microglobulin/creatinine ratio (Figure 2G) but did not correlate with decline in estimated glomerular filtration rate (eGFR) or increase in serum creatinine (Figure 2E,F).Thus, morphometry of nuclear size distribution was a valuable tool in identifying KIN, but the relevance of such small proportions of karyomegalic cells might be questioned.However, in addition, decline in eGFR was neither correlated with glomerulosclerosis (R 2 = 0.35, p = 0.2, data not shown) or tubulointerstitial fibrosis (R 2 = 0.32, p = 0.2, data not shown).

Features of cellular senescence in and beyond karyomegalic cells
Cellular senescence has been implicated in KIN caused by impaired DNA damage response due to a FAN1 mutation [4].We tested for FAN1 gene abnormalities in our first patient, but none were found.We looked for cellular senescence features in the patient's kidney biopsies and found that most of the enlarged nuclei in the six biopsies were Ki67-negative and p21-positive, consistent with cell-cycle arrest, the key feature of cellular senescence (Figure 3A-C) [9].In addition, a small subset of nuclei contained γH2AX foci, indicating unresolved doublestrand DNA breaks (Figure 3D,E), and showed decreased expression of the nuclear envelope protein lamin B1 (Figure 3F and supplementary material, Figure S2), both indicative of cellular senescence [9,10].Surprisingly, the loss of lamin B1 and increased expression of p21 were not confined to the karyomegalic cells but were also seen in morphologically normal appearing TECs.
Nuclei positive for p21 were abundant throughout the (proximal) nephron (observed in S1, S2, and S3 segments) and, to a lesser extent, the distal nephron (supplementary material, Figure S3).
Since endocycling has been associated with the expression of senescence features in TECs, we also performed FISH with centromere probes for three different

Discussion
The major finding of this study is that tubular dysfunction, associated with features of cellular senescence in TECs, appears to be an important aspect of early CKD development in childhood cancer patients.This phenotype has been described as KIN [2,3,5].With morphometry and p21 staining, we could readily diagnose this condition in kidney biopsies from all six consecutive pediatric oncology patients treated with ifosfamide and presenting with polyuria, LMW proteinuria, and with loss of GFR.
KIN has been linked to FAN1 mutations, suggesting a role for defective DNA damage repair and ensuing cellular senescence, but this hypothesis has not been substantiated to date [4].
Several studies have reported KIN in patients treated for childhood cancer as a rare event related to ifosfamide treatment, a nephrotoxic drug that is known to cause tubulopathy, reduced glomerular filtration, and renal Fanconi syndrome, but details on clinical features and pathology are lacking [1,[13][14][15][16][17][18].Despite the small sample size, our observations suggest that tubular injury after DNA-damaging and cell stress-inducing therapy, including ifosfamide, is responsible for the clinical picture and suggest that KIN and cellular senescence should be considered as the underlying cause of early CKD in childhood cancer patients.All patients received additional nephrotoxic therapies, such as chemotherapeutic agents (cisplatin in one patient and cyclophosphamide in three patients) and radiotherapy to the kidney region (limited to four patients), so kidney damage may not be attributable to ifosfamide alone.The observations that case 3 with the highest dose of ifosfamide showed the lowest number of karyomegalic cells and stabilization of renal function and that case 1 with the worst renal function decline received ifosfamide, cisplatin, and radiotherapy in the kidney region suggest an additional role of other nephrotoxic agents in the pathophysiological mechanism of KIN in childhood cancer.Multiple stress responses seem important in the development of this particular form of nephrotoxicity, as described recently in cisplatin toxicity animal models [19].Nevertheless, tubular karyomegaly, the most specific feature of KIN, has thus far only been reported after ifosfamide and cyclophosphamide therapy, and not for other chemotherapeutic agents or radiotherapy [1,13,14,20].
All six biopsies in the current study contained TECs with large, irregular, and hyperchromatic nuclei.Tubulointerstitial fibrosis was observed in four patients, of whom three had little (<5%) fibrosis and only one had prominent (20%) fibrosis.Only two biopsies (including the one with prominent fibrosis) also showed mild to moderate interstitial inflammatory infiltrate.Previous reports of ifosfamide-induced KIN cases reported moderate to severe fibrosis with a mild to severe lymphocytic infiltrate [1,13,14].Although KIN is classically referred to as interstitial nephritis [3], the (near) absence of inflammation in our cases suggests that the abbreviation "KIN" might be better used to refer to nephropathy, rather than nephritis.the p value and R 2 increase when the biopsy with the smallest 99.9th percentile of nuclear size is excluded (R 2 = 0.99, p = 0.07).For two patients no urinary β2-microglobulin/creatinine ratio was available at time of biopsy.

462
SN Knoppert, MG Keijzer-Veen et al The karyomegalic TECs in all six biopsies had senescence features, including unresolved DNA damage, cell-cycle arrest, and nuclear lamina alterations.Strikingly, these senescence features were also observed in far larger numbers of morphologically normal cells.Senescence is a prolonged antiproliferative cell state that is typically induced when DNA repair and apoptosis fail and is associated with maladaptive repair and progressive fibrosis in CKD [21][22][23].Defective DNA damage repair was previously linked to genetic KIN caused by FAN1 mutations [4].Furthermore, high rates of DNA damage have been observed in pediatric oncology patients treated with ifosfamide, and low rates of cellular proliferation [24], indicative of cell-cycle arrest, are observed in ifosfamide-induced KIN [1,14].Our data underscore that senescence is a prominent feature in KIN.Furthermore, these senescent cells can contribute to progressive loss of kidney function through senescence-associated secretory phenotype (SASP) factors that drive progressive tubulointerstitial fibrosis [25].
Recently, endocycling in TECs has been identified as a survival mechanism of TEC after acute kidney injury [11,26].This process results in polyploidy, which can have deleterious effects contributing to cellular senescence [26].In the KIN biopsies we found significant numbers of polyploid cells, which correlated with reduced lysozyme expression, suggesting that also in this condition endocycling may contribute to tubular cell dysfunction.
Although KIN is considered a rare condition, it may be a more prevalent cause of reduced kidney function among childhood cancer patients than previously suggested.Of note, our study only refers to children in which a biopsy was performed because of kidney function decline of unknown cause was detected.However, development of some degree of (even subclinical) kidney dysfunction is not a rare event in childhood cancer patients but can easily be overlooked [27][28][29].Subtle changes in proteinuria resulting from tubular dysfunction may go unnoticed in standard urinalysis if one does not measure LMW proteins.Hence, KIN/cellular senescence might be even more prevalent than realized thus far.Therefore, we would suggest that standard urinalysis, protein-to-creatinine ratio, and detection of urinary LMW protein, glucose and phosphate should be pursued as a standard of care in high-risk patients to optimize surveillance of kidney toxicity [30].
The importance of recognizing KIN as a separate entity in children suspected of CKD, early after childhood cancer treatment, lies in the feasibility of early intervention to avoid later sequelae.The use of steroids may have prevented further kidney function decline in one of our patients, but steroids were only administered in two patients with interstitial inflammation.Although alternative therapies are currently not available, our data may provide perspectives for treatment of KIN with novel, so-called senomorphic and senolytic therapies [9].
In conclusion, cellular senescence due to unresolved DNA damage might be the pathogenesis of KIN.KIN can develop during or after childhood cancer treatment.This urges the implementation of surveillance of kidney function as standard of care, as early recognition of this condition may benefit from the development of novel interventions targeting senescent cells.

Figure 1 .
Figure 1.Kidney function decline and nuclear size.(A) Change in eGFR from diagnosis of malignancy (first point in time) to biopsy (time = 0) to last moment of follow-up (last point in time).All six patients had a decline in eGFR from moment of diagnosis of malignancy until last moment of follow-up (last point).After biopsy three patients showed progressive decline of eGFR, while the other three patients showed improvement of eGFR during follow-up.(B) Representative pictures of H&E stain for all six biopsies.The six biopsies showed a variation in inflammation and nuclear size.On visual observation the nuclei ranged from bizarre, irregularly enlarged nuclei to almost inconspicuously enlarged nuclei.

Figure 2 .
Figure 2. Nuclear size distribution and correlation with kidney function and LMW proteinuria.(A) Nuclear outlines are detected using QuPath, calculating nuclear surface area.(B) Comparison of nuclear size distribution of KIN biopsies with protocol kidney graft biopsies (aggregate of six biopsies per group).The dotted vertical line indicates the mean, which is comparable for both groups.The small lines on the X-axis represent single nuclei.(C and D) Comparison of 99th percentile and 99.9th percentile of nuclear size in individual KIN biopsies and protocol kidney graft biopsies.Although a significant difference is seen when comparing the 99th percentile of nuclear size between the two groups, one of the biopsies in the KIN groups (case 3, indicated by red dot) shows remarkable overlap with the control group.When focusing on an even smaller subpopulation of nuclei (99.9th percentile), the overlap is lost.(E-G) Scatter diagram with best-fit line of nuclear size (99.9thpercentile) in relation to kidney function and urinary β2-microglobulin/creatinine ratio.No correlation or trend is seen between the size of the 99.9th percentile of nuclei and change in eGFR or serum creatinine (E, R 2 = 0.12, p = 0.5; F, R 2 = 0.26, p = 0.3).A nonsignificant trend is seen between nuclear size (99.9thpercentile) and urinary β2-microglobulin/creatinine ratio (uβ2m/Cr).For two patients no urinary β2-microglobulin/creatinine ratio was available at time of biopsy (G, R 2 = 0.83, p = 0.09).In all graphs the red dot represents the KIN biopsy with the smallest average nuclear size in the 99th and 99.9th percentiles.

Figure 3 .
Figure 3. Senescence phenotype in KIN.(A) Representative pictures of Ki67 and p21 double stain.(B) Cell cycle was evidenced by the presence of Ki67 costaining in only 0.3% of p21-positive cells.(C) Representative picture of a single Ki67 and p21 double positive nucleus (arrow).(D) Representative image of γH2AX positive cell.(E) A small subset of nuclei contains γH2AX foci, indicating unresolved double-stranded DNA breaks.(F) Decreased staining for nuclear lamin B1 in both enlarged (left arrow) and regular sized (right arrow) nuclei.

Figure 4 .
Figure 4. p21 expression in relation to kidney function and tubular reabsorption.(A-C) Representative pictures and quantification of p21-positive cells in (A) protocol kidney graft biopsies and (B) KIN biopsies.KIN biopsies showed significantly more p21-positive cells compared to protocol kidney graft biopsies (C) ( p = 0.002).(D-F) Scatter diagram with best-fit line of p21-positive cell percentage in relation to kidney function and urinary β2-microglobulin/creatinine ratio.The red dot indicates the biopsy with the lowest 99.9th percentile in nuclear size.(D) p21 and change in eGFR show a nonsignificant negative trend (A, R 2 = 0.43, p = 0.2) but when the biopsy with the smallest nuclei is excluded a clear correlation is present between p21 expression and a decrease in kidney function (R 2 = 0.93, p = 0.008).(E) p21 and change in serum creatinine show a positive correlation (R 2 = 0.83, p = 0.01) in the KIN biopsies.This correlation is maintained when to biopsy with the smallest 99.9th percentile of nuclear size is excluded (R 2 = 0.79, p = 0.04).(F) p21 and urinary β2-microglobulin/creatinine ratio (uβ2m/Cr) show a positive correlation (R 2 = 0.93, p = 0.04), although this correlation is lost when the biopsy with the smallest 99.9th percentile of nuclear size is excluded (R 2 = 0.96, p = 0.1).For two patients no urinary β2-microglobulin/creatinine ratio was available at time of biopsy.

Figure 5 .
Figure 5. Lysozyme and p21 stain and LMW proteinuria.(A) Double stain for p21 and lysozyme.Tubules and single cells expressing p21 showed reduced staining for lysozyme.(B) Scatter diagram with best-fit line of percentage of p21-positive cells in relation to percentage of lysozyme positive surface area in biopsy cortex.Lysozyme and p21 show a nonsignificant negative trend (R 2 = 0.63, p = 0.06); the p value increases when the biopsy with the smallest 99.9th percentile of nuclear size is excluded (R 2 = 0.64, p = 0.1).(C) Scatter diagram with bestfit line of relation between urinary β2-microglobulin/creatinine ratio (uβ2m/Cr) and lysozyme positive surface area (R 2 = 0.88, p = 0.06);the p value and R 2 increase when the biopsy with the smallest 99.9th percentile of nuclear size is excluded (R 2 = 0.99, p = 0.07).For two patients no urinary β2-microglobulin/creatinine ratio was available at time of biopsy.

Table 1 .
Baseline and follow-up characteristics of cases.