Tubulointerstitial Inflammation and Scarring Predict Outcome in Lupus Nephritis

Subject inclusion and clinical data collection

This study was approved by the University of Chicago Medical Center Institutional Review Board. Subjects were eligible for inclusion if they had a renal biopsy consistent with lupus nephritis between 2001 and 2007 and sufficient material for analysis (≥six glomeruli and length of ≥0.5 cm). Subjects were excluded if they did not fulfill American College of Rheumatology (ACR) revised criteria for the classification of SLE (24) or had alternative diagnoses potentially causing glomerulonephritis. Of 70 initial subjects, 1 was excluded because ACR criteria were not fulfilled, and another excluded due to concurrent hepatitis C, resulting in 68 renal biopsies.

The clinical charts were reviewed for demographic data, medication history, degree of proteinuria and serum creatinine at time of biopsy and at most recent follow-up (or until renal failure ensued) and serological parameters at time of biopsy. Glomerular filtration rate (GFR) was estimated using the IDMS-traceable MDRD study equation in adults (25) and a modified IDMS-traceable Schwartz equation in children (26). Treatment at the time of renal biopsy was classified as major immunosuppression (prednisone or equivalent > 20 mg/day within one month of renal biopsy and/or cyclophosphamide, mycophenolate mofetil, azathioprine, or cyclosporine within three months of biopsy) versus minimal immunosuppression (prednisone or equivalent ≤ 20 mg/day within the month prior to biopsy and no major immunosuppressant as defined above within three months of biopsy.)

Renal outcome data was collected for survival analysis. End stage renal failure (ESRD) was defined as the initiation of chronic dialysis therapy or renal transplantation. The time until ESRD, or last known follow-up in the absence of renal failure, was collected by chart review.

Histologic assessments

Formalin-fixed, paraffin-embedded 2 μm sections were used for microscopic evaluation. For each biopsy, hematoxylin and eosin (H&E) stains and periodic acid-Schiff (PAS) stains were performed. Standard immunofluorescence microscopy using fluorescein isothiocyanate-conjugated antibodies for the following antigens: IgG, IgA, IgM, C3c, C1q, fibrinogen, κ and λ light chains, and albumin (DAKO, Carpinteria, CA) was applied and staining was semi-quantitatively scored from 0 to 4+. Standard electron microscopic procedures were applied using a Philips CM10 electron microscope. Interstitial inflammation was determined using monoclonal antibodies to CD45 (DAKO, Carpinteria, CA), a pan-leukocyte marker, on paraffin tissue sections. The percentage of positive cells for CD20 (DAKO), a pan-B cell marker, and anti-CD3 (Labvision, Fremont, CA), a pan-T cell marker, was also determined using the number of CD45+ cells as the denominator. Interstitial inflammation was also assessed by standard light microscopy, modified to exclude areas adjacent to tubular atrophy and interstitial fibrosis, (modified light microscopy or MLM). The extent of interstitial inflammation by each method was semiquantitatively scored without prior knowledge of the clinical outcomes from 0-4, which corresponded with 0%, <10%, 10-25%, 26-50% or >50% involvement of the tubulointerstitium.

The kidney biopsies were classified using the 2003 ISN/RPN revised lupus nephritis classification. Class III or Class IV lupus nephritis with concurrent Class V was subsumed under Class III or IV lupus nephritis. Subsequently, Class III and Class IV were combined into a proliferative nephritis category, and Class II and isolated Class V into a nonproliferative category for statistical analysis. The activity and chronicity indices were scored using the NIH system (12). To separate the effect of interstitial inflammation on the NIH activity index, a modified activity index which excluded the interstitial component was calculated by deducting the numeric score given for interstitial inflammation (0-3) from the total NIH activity index score, yielding a potential total score of 21. Similarly, the 12-point NIH chronicity index was separated into a glomerular chronicity score, and an interstitial chronicity score; each ranging between 0 and 6 (Table 1).

Statistical Analysis

Data were analyzed using STATA 10.0 (StataCorp, College Station, Texas). Initially, cross-sectional analysis of interstitial nephritis in relation to a variety of demographic, clinical and histologic characteristics was performed. Univariate statistical analysis was performed using the Fisher's Exact test and the Wilcoxon Rank Sum test. Multivariate logistic regression using decreased GFR (< 60 ml/min/BSA) was used to examine the independence of demographic, clinical and histologic findings which were significant on univariate analysis. Degree of proteinuria at time of biopsy was included in the regression analysis because heavy proteinuria has been postulated as a cause of tubular disease (27, 28). Histologic variables in the regression analyses included the intensity of interstitial inflammation as quantified by CD45 staining; interstitial inflammation by MLM; and the chronicity index broken into its two subcomponents (tubulointerstitial chronicity score and glomerular chronicity score).

Survival analysis was carried out to determine the predictive value of each measure of interstitial inflammation. Three patients were censored from the survival analysis portion of the study due to lack of follow-up renal function and 6 patients were censored due to renal failure at time of biopsy, resulting in 59 subjects for the survival analysis. Renal survival analysis was performed by the Kaplan-Meier technique and the statistical difference between survival curves was assessed with the log rank test for equality of survivor functions, using time to ESRD as the event variable. The following group variables were used in separate sequential survival analyses: severe interstitial nephritis (binary variable), the 4-tier interstitial nephritis variable (log rank trend test), serum creatinine > 1.4 mg/dL, NIH activity index > 6, African-American race, NIH chronicity index > 3, glomerular chronicity index > 0, tubulointerstitial chronicity index > 0, LN class (proliferative vs. nonproliferative), pediatric age at biopsy, degree of proteinuria (>3.5 gms/24 hours), and the presence of dsDNA antibodies. A Cox Proportional Hazards analysis was used to determine if the tubulointerstitial chronicity score and glomerular chronicity score each independently contributed to risk of renal failure. This model consisted of two covariates, the tubulointerstitial chronicity score and glomerular chronicity score, with the dependent outcome variable being renal failure. Subsequently, this model was serially adjusted for degree of proteinuria and race to determine any effects upon these covariates. A p value of < 0.05 was considered statistically significant for the above analyses.

Clinical Characteristics of Patient Cohort

We identified 68 subjects fulfilling ACR criteria for the classification of SLE (24) who had lupus nephritis on renal biopsy and had adequate stored tissue for histologic staining. The clinical characteristics of the population at time of biopsy are provided in Table 2. The population was predominantly African-American (81%) and female (85%). Median serum creatinine was 1.0 mg/dL (range 0.2-6.4), and median proteinuria was 2.5 gms/day (range 0-19.7). Sixty-five patients had data with which to estimate GFR. Fifty-five percent had Stage I or II chronic kidney disease (GFR > 60 ml/min/1.73m²), while 20% had Stage III, 11% Stage IV and 14% Stage V at time of biopsy. Seventy-five percent were receiving major immunosuppression within the month prior to biopsy. Fourteen subjects progressed to ESRD during the study, for a 24% rate of renal failure.

The distribution of nephritis class by ISN/RPS classification criteria is shown in Table 2. Evaluation of biopsies using the NIH activity index revealed moderate activity (mean 6.6 ± 5.1, range 0-19). Scoring of biopsies using the NIH chronicity score revealed substantial scarring (mean 2.7 ± 2.6, range 0-10) at time of biopsy.

Tubulointerstitial inflammation was associated with a higher serum creatinine and decreased GFR at time of biopsy

To characterize the extent of active interstitial nephritis, paraffin sections from each biopsy were stained with antibodies to the pan-leukocyte marker CD45 and then quantified on a five-point scale (0-4) (Methods) (29). All biopsies had some interstitial infiltrates on CD45 staining. Six percent of biopsies were graded as 1, 22% as 2, 40% as 3, and 32% as 4. For most subsequent analyses, the interstitial activity scores were consolidated into a binary ranking of mild (0-2) and severe (3-4) interstitial nephritis.

Sections were also stained by H&E and PAS and analyzed by MLM. By MLM, 9% scored as 0, 32% as 1, 25% as 2, 15% as 3, and 19% as 4. Interstitial inflammation, as measured by CD45, correlated well with MLM (Spearman's rho 0.63, p< 0.0001).

Interstitial inflammation by CD45 staining correlated with renal function. The median serum creatinine was 1.4 mg/dL in the severe interstitial inflammation group and 0.7 mg/dL in the mild interstitial inflammation group (Wilcoxon rank sum test p=0.0001). Of 46 subjects with severe interstitial inflammation by CD45 staining, 19 had a normal GFR (< 60 ml/minute/BSA, 27 (41%), compared to 17/19 (89%) with normal GFR amongst mild interstitial inflammation subjects (Fisher's Exact test, p<0.001 for CD45 stain, Table 3). Similar findings were seen with MLM staining. Median serum creatinine was 2.2 mg/dL in subjects with severe interstitial inflammation, and 0.8 mg/dL in subjects with mild interstitial inflammation by MLM (p=0.004). Nineteen percent of subjects with severe interstitial inflammation by MLM had a normal GFR, while 73% with mild interstitial inflammation had a normal GFR (p<0.001). T cells predominated over B cells, with a mean of 66% versus 12% of interstitial cells. While neither B or T cell percentages correlated with GFR or serum creatinine, B cells were increased in specimens with intense interstitial infiltrates (mean CD20 cells 10% in low interstitial infiltrate subjects and 16% in high interstitial infiltrate subjects, p=0.015). The relative predominance of B cells by anti-CD20 or T cells by anti-CD3 did not correlate with GFR or serum creatinine (data not shown).

Surprisingly, severe interstitial nephritis by CD45 staining was not associated with accepted measures of lupus activity such as elevated dsDNA antibodies and low serum C3 (Table 3). While severe interstitial nephritis had a higher median level of proteinuria (3.2 grams/24 hours in severe interstitial nephritis versus 1.7 grams/24 hours in mild interstitial nephritis), this difference did not achieve statistical significance (p=0.16). Severe interstitial nephritis was weakly associated with African-American ancestry (p=0.04). Amongst African-American subjects, 78% had severe interstitial inflammation, compared to 46% amongst subjects with other ancestries.

There was a poor correlation between the magnitude of tubulointerstitial and glomerular involvement

ISN/RPS Class III or IV proliferative nephritis was somewhat more common in subjects with severe interstitial nephritis when analyzed by CD45 staining, occurring in 88% of the severe interstitial nephritis group and 63% of the mild interstitial nephritis group (p = 0.04, Table 3). However, among the 13 subjects with nonproliferative glomerular lesions, 6 (46%) had severe interstitial nephritis. Thus, severe interstitial nephritis was widespread regardless of ISN/RPS lupus nephritis class.

We then examined the relationships between interstitial nephritis by CD45 staining and the NIH activity index (Table 3). This index, although weighted towards glomerular pathology, awards 3 of 24 points based on the degree of interstitial inflammation. Therefore, it was not surprising that the interstitial activity score correlated with the NIH activity index (p=0.01). However, the modified NIH activity score, which excluded tubulointerstitial inflammation, did not correlate with the interstitial activity score (p=0.18), indicating that there is not a close correlation between the severity of pathology in the glomerulus and the interstitium.

Tubulointerstitial inflammation is associated with tubulointerstitial scarring

In contrast to the NIH activity index, a significant correlation was noted between the interstitial activity score by CD45 staining and the NIH chronicity index (Spearman's rho=0.38, p=0.002). However, the NIH chronicity index gives equal weight to scarring in the glomeruli and scarring in the tubulointerstitium. To determine which component of the NIH chronicity index correlated with the interstitial activity score, we calculated separate glomerular and interstitial chronicity subcomponent scores and compared these to the interstitial activity score. As can be seen in Table 3, the interstitial activity score, as assessed by CD45 staining, strongly correlated with the interstitial component of the NIH chronicity score (rho=.44, p=0.002) and with the total NIH chronicity score (rho=0.38, p=0.002). There was a weaker association between interstitial inflammation and the glomerular chronicity score (rho=0.27, p=0.02). However, in a multivariate analysis, severe interstitial inflammation correlated significantly with the interstitial (OR 1.79 CI 1.0, 3.1, p=0.035) but not glomerular chronicity scores (OR 1.32, CI 0.6, 3.1 p=0.53).

Interstitial scarring and inflammation independently were associated with low GFR

Because a variety of factors were associated with impaired GFR on univariate analysis, multivariate logistic regression was performed using the dependent variable of diminished GFR (< 60 ml/min/BSA) at time of renal biopsy. When both the interstitial chronicity and glomerular chronicity scores were used as covariates, the interstitial chronicity score was associated with low GFR (OR 2.4, 95% C.I. 1.4, 4.3, p=0.002), while the glomerular chronicity score was not (OR 1.3, 95% C.I. 0.6, 2.6, p=0.43). Subsequent adjustment of the model serially for degree of proteinuria, and glomerular crescents/necrosis did not significantly affect these associations. These data indicate that interstitial, and not glomerular, scarring is associated with renal dysfunction.

We then added severe interstitial inflammation by CD45 staining (grade 3-4) to the regression model containing interstitial and glomerular chronicity scores. These analysis indicated that interstitial inflammation was independently associated with a low GFR (OR 5.3, 95% C.I. 1.0, 28.9, p=0.05). These data suggest that at time of biopsy interstitial inflammation, independently of renal scarring, correlates with renal dysfunction.

Severity of interstitial nephritis, but not ISN/RPS lupus nephritis class or NIH activity index, predicts renal survival

We next examined if classic histological measures, or the degree of interstitial nephritis on biopsy, identified those at risk for subsequent renal failure. As demonstrated in Figure 1A, approximately 37% of patients with severe interstitial nephritis by CD45 staining progressed to renal failure at 24 months compared to 15% of those with mild interstitial nephritis (p=0.03). The relationship between interstitial nephritis and subsequent renal failure was also apparent when interstitial nephritis was treated as a 4-tier variable (interstitial activity scores 1-4, Figure 1B). Those patients with little interstitial nephritis on renal biopsy (interstitial activity score 1) uniformly had an excellent prognosis while those with interstitial activity scores of 2 and 3 had relatively good and intermediate prognosis respectively. In contrast, nearly 40% of those with a score of 4 progressed to renal failure within 24 months. This incremental decline in renal survival was significant (p=0.02, Log-Rank Trend Test).

Because nonspecific inflammation may occur around areas of interstitial fibrosis, the survival analysis was repeated using modified light microscopy (MLM). This yielded similar results. The 5-tier interstitial activity score by MLM (interstitial activity scores 0-4, Figure 1C) also predicted renal failure risk (p=0.006, trend test). The group with an interstitial inflammation score of 4 had a much more grave renal prognosis (60% ESRD by 36 months) than the other MLM groups (scores 0-3). In contrast to CD45 staining, the other MLM groups (scores 1, 2, and 3) had survival curves that were overlapping and did not demonstrate a dose-response relationship between interstitial inflammation and ESRD (Figure 1B and 1C). These data suggest that CD45 staining better differentiates prognosis at intermediate levels of interstitial inflammation.

In contrast, the ISN/RPS lupus nephritis class and the NIH activity index had no statistically significant effect on subsequent renal survival (Figure 2, p=0.22 and p=0.80 respectively). Likewise, anti-dsDNA antibody titers, decreased C3, pediatric age and nephrotic range proteinuria were not useful in identifying patients at risk for subsequent renal failure (p all > 0.10). Therefore, among demographic data and measures of active inflammation, only the severity of interstitial nephritis provided prognostic information in our cohort and did so regardless of which histologic method was used.

The prognostic utility of the NIH chronicity score is primarily driven by tubulointerstitial scarring

A higher NIH chronicity index is a well-established poor prognostic indicator in LN.(15, 30). Analysis of our cohort confirms these previous observations. Those with a NIH chronicity score of greater than 3 were much more likely to progress to renal failure (p<0.001). We then divided the NIH chronicity score into its glomerular and tubulointerstitial subcomponents to determine which was more predictive of subsequent renal survival (Figure 3). As demonstrated in Figure 3, we observed a somewhat more significant decline in renal survival with evidence of tubular atrophy or interstitial fibrosis (score > 0, p=0.011) compared to evidence of glomerular sclerosis or fibrous crescents (score > 0, p=0.07). The tubulointerstitial chronicity score (score > 0) did not correlate with the degree of proteinuria at biopsy (p=0.66, rank sum test), arguing against a causal role of chronic proteinuria in interstitial fibrosis. A multivariate renal survival analysis, reflected a strong independent association of the interstitial component of the chronicity index with renal failure (hazards ratio 2.20, 95% C.I. 1.3, 3.6, p=0.002) and no independent association of glomerular damage (hazards ratio 1.0, 95% C.I. 0.7, 1.5, p=0.97) with renal failure. The model was unaffected if adjusted for degree of proteinuria. These data indicate that the prognostic value of the NIH chronicity index is primarily due to those components that capture tubulointerstitial scarring.