Introduction

Solid organ transplantation is the only definite treatment of end-stage organ failure but lifelong immunosuppression leads to an increased risk of viral infections and malignancies [1, 2]. One frequent complication after kidney transplantation is polyomavirus associated nephropathy, predominantly caused by the human BK polyomavirus (BKPyV). BKPyV was described in 1971 [3] as the first member of the human polyomavirus family which now comprises 14 viruses [4]. Polyomaviruses are small DNA viruses which encode two early mRNAs that are translated into the large T antigen (LTag) and the small T antigen. LTag is able to bind and block the tumor suppressor functions of p53 and pRB [5]. The oncogenic potential of polyomaviruses has been studied extensively [6, 7] and a causative role has been shown for Merkel cell polyomavirus (MCPyV) in Merkel cell carcinoma of the skin [8].

BKPyV is widely distributed with an overall IgG antibody seroprevalence rate of about 80% [9]. Primary infection usually occurs during childhood and is followed by latency of BKPyV mainly in renal epithelial cells and transitional cells of the urogenital tract, but also in tonsils, lymphocytes, blood monocytes, and brain [5, 10]. Immunosuppression can cause BKPyV reactivation with lytic replication in permissive tissues. Although lytic infection can often be asymptomatic and transient, ongoing high level virus replication leads to urothelial or renal injury. Resulting polyomavirus associated nephropathy occurs in a significant percentage of kidney transplant recipients and often leads to graft failure and transplant loss [11], BKPyV-associated hemorrhagic cystitis is a common complication after allogeneic hematopoietic stem cell transplantation [12]. For BKPyV, an association between BKPyV reactivation and an increased incidence of UC in renal transplant recipients has initially been observed [13, 14]. Up to date, <50 BKPyV-associated posttransplant UC have been reported [15, 16]. These tumors affect tissues of BKPyV latency and usually occur in kidney transplants or in the recipients' bladder. It is assumed that asymptomatic latent or productive lytic infections lead to rare genetic accidents with integration of BKPyV genomes. Tumor cells typically show strong expression of polyomavirus LTag but no productive/lytic infection [10]. BKPyV-associated tumors often display a variant morphology, e.g., micropapillary UC (MPUC), an aggressive UC variant with frequent TP53 mutations, and HER2 alterations [15, 17]. NGS based data are available for <20 BKPyV-associated UC [16, 18,19,20,21,22]. We performed a molecular study for another five cases of this rare and probably underrecognized entity.

Materials and methods

Case selection

Tissue microarrays (TMA) [23] containing 94 MPUC collected from archival material of nine cooperating centers in Germany and France were screened for polyomavirus LTag expression by immunostaining with a SV40 LTag epitope cross reacting antibody.

We also screened large series of consecutive UC, including other histopathological variants. In detail a TMA of 480 consecutive UC diagnosed in our department from 1998 until 2004 including variant UC (n = 21 MPUC, n = 11 sarcomatoid, n = 4 plasmacytoid, n = 1 lymphoepithelioma-like, n = 1 nested, n = 2 neuroendocrine) and a TMA with 199 consecutive muscle invasive UC treated with radical cystectomy and also including variant UC (n = 10 MPUC, n = 10 sarcomatoid, n = 10 neuroendocrine, n = 7 plasmacytoid, n = 2 lymphoepithelioma-like, n = 2 nested, n = 1 clear cell, n = 1 large nested) [24] was investigated by immunohistochemistry (IHC). Furthermore specific TMAs of our studies of UC with variant histology were screened with IHC, including 24 plasmocytoid, 33 nested, and 19 large nested carcinomas.

A retrospective search of the records of the departments of surgical pathology and nephropathology (University Hospital Erlangen) including routine and consult cases was performed to identify UC in solid organ transplant recipients for subsequent LTag immunostaining. Institutional review board approval (University Hospital Erlangen) was obtained for molecular analysis of archival material.

Morphology and IHC

Complete histomorphological reevaluation of hematoxylin and eosin (H&E)-stained tumor slides of the five LTag expressing UC was done according to the current TNM and WHO classification systems [25, 26] by two experienced uropathologists (AH, SB).

IHC was performed on a fully automated IHC slide staining system (BenchMark XT System, Ventana Medical Systems Inc., Tucson, AZ) using the following antibodies: p53 (DAKO, DO-7, 1:50), SV40 (Abcam, PAb416, 1:1000), and Her2 (Dako, Poly, 1:1000). Additional GATA3 (DCS, L50–823, 1:1000) and PAX8 (Cell Marque, Poly, 1:50) staining was performed for cases 3 and 4 to clarify that these tumors are UC. IHC was considered significant for LTag in the case of any nuclear staining and for TP53 in cases with aberrant staining, including nuclear diffuse immunoreactivity, or complete loss of expression. Patchy immunostaining of p53 was considered wild-type pattern [27]. HER2 IHC was evaluated according to the recommendations of the American Society of Clinical Oncology and the College of American Pathologists (ASCO/CAP) [28]. HER2-CISH analysis with the SPEC HER2/CEN17 Probe Kit (Zytovision GmbH, Bremerhaven, Germany) was performed according to the manufacturer’s instructions and evaluated as recommended in the revised ASCO/CAP guidelines for HER2 amplification in breast cancer [28].

Nucleic acid isolation

Formalin-fixed paraffin-embedded (FFPE) tissue was available for cases 1–4 and fresh frozen tissue for case 5. Tumor and nontumor areas were marked separately on H&E stained slides prior to manual dissection. Nucleic acid isolation was performed according to a previously reported protocol [29] using the Maxwell® 16 LEV Blood DNA Kit (Promega, Mannheim, Germany).

SNaPshot analysis of TERT promotor and HRAS

Mutation hot spots of the TERT promoter (at positions −146, −124, and −57 bp) and of HRAS (at codons 12, 13, and 61) were investigated using SNaPshot analysis as previously described [30, 31] with an ABI Prism 3500 Genetic Analyzer and the SNaPshotTM Multiplex Kit (Applied Biosystems, Foster City, CA, USA).

Viral detection and quantification

Real-time PCR was performed to detect BKPyV-DNA [32] and human herpesvirus 6 (HHV-6) DNA [33]. To evaluate the number of cells investigated, a real-time PCR assay for the ALB gene was performed [34]. PCR reactions were performed in a final volume of 50 µl using 10 µl DNA, 25 µl of TaqMan® Universal PCR Master Mix (Thermo Fisher Scientific), 25 pmol of each primer, and 10 pmol of each fluorescence labeled probe (Table S1). After an initial denaturation phase of 10 min at 95 °C, 45 cycles of denaturation at 95 °C for 15 sec, and a combined annealing and extension phase at 60 °C for 1 min were performed using the ABI PRISM® 7500 (TaqMan). The lower detection limit was between five and ten DNA copies per reaction in all real-time PCR assays.

NGS and sequence analysis

DNA from four FFPE- and one fresh frozen-tumor was converted into Illumina NGS libraries and sequenced on HiSeq 4000 or X-Ten (Macrogen Inc, Seoul, Republic of Korea). Paired reads (between 313 and 880 million/sample) were aligned to the human genome, and BKPyV integration sites were identified from broken paired reads mapped to human (hg19) and BKPyV genomes. For this, patient specific BKPyV consensus genomes were generated by mapping reads to the BKPyV RefSeq genome (NC_001538) and filling of gaps from this reference. Reads not mapping to human and BKPyV genomes were also mapped against the NCBI RefSeq dataset of DNA virus genomes (Fig. S1) [35]. RNA-seq of a strand specific library made from rRNA depleted (Ribo-Zero) total tumor RNA from the fresh frozen-tumor sample (case 5) was performed to test cellular and viral gene expression. All bioinformatic analyses were performed with CLC Genomics Workbench 11 and 12 (Qiagen, Aarhus, DK). Mutants were identified from de-duplicated read mappings (to hg38, coverage 10, frequency >5%) and filtered for clinical relevant variants in the ClinVar 20171029 database, focusing on a set of 50 genes most frequently involved in bladder cancer and variants therein present with at least two reads and 10% frequency and annotations “pathogenic” or “likely pathogenic”.

The BKPyV genotype was assigned by multiple sequence alignment of the BKPyV VP1 gene sequence of the patient strains and 79 VP1 sequences from the database, followed by construction of a phylogenetic tree using the Neighbor Joining method and the Jukes Cantor nucleotide substitution model, and subtypes assigned by nucleotide BLAST of patient specific consensus genomes against the 162 BKPyV genomes used by Luo et al. [36].

Results

Case identification and clinical presentation

IHC staining of 94 MPUC revealed two LTag expressing UC (cases 1 and 2). Retrospective assessment of clinical data showed that both patients were kidney transplant recipients. Subsequent screening of the records of our departments of surgical pathology and nephropathology identified 15 UC which were diagnosed after solid organ transplantation (n = 12 kidney, n = 2 heart, and n = 1 liver recipient). Among them, another 3 LTag expressing UC were found (cases 3–5).

IHC staining of additional TMAs with overall 755 conventional and variant UC (24 plasmacytoid variant UC, 33 nested variant UC, 19 large nested variant UC, 480 consecutive UC of all stages and 199 muscle-invasive UC, both including variant UC) did not show any LTag expressing UC.

Patient and tumor characteristics and clinical follow up data of the five LTag expressing cases are shown in Table 1. For patients 1 and 3, no information was available to assess a possible BKPyV reactivation. In patient 2, graft biopsy showed no polyomavirus associated nephropathy 1 month after the second kidney transplantation. Patient 4 had no clinical symptoms of BKPyV reactivation and tested repeatedly negative for BK viremia during routine checkup. Patient 5 had suffered from hemorrhagic cystitis due to BKPyV reactivation after heart transplantation.

Table 1 Clinical data of BKPyV-associated UC.
Full size table

Morphology and IHC

Clinico-pathological and morphological characteristics of the analyzed posttransplant UC are shown in Table 2. Histological examination revealed one pure MPUC (case 1), and two UC with MPUC components >90% (case 2), and 10% (case 5), respectively, the latter with a concurrent exophytic non-invasive papillary urothelial carcinoma (pTa, high-grade) component with micropapillary areas at the surface and extensive urothelial carcinoma in situ (CIS) covering the whole bladder (Fig. 1). The remaining two UC (cases 3 and 4) showed a focal conventional UC component. The unusual glandular pattern in case 3 mainly corresponded to clear cell carcinoma of Mullerian type, and there was concomitant urothelial CIS. IHC revealed GATA3 expression and at least focal coexpression of PAX8 in both the CIS and the invasive tumor component. According to the current WHO classification [26] this tumor represents urothelial carcinoma with extensive divergent differentiation with Mullerian type clear cell carcinoma. In case 4 there was extensive villoglandular pattern with nuclear expression of GATA3, whereas PAX8 was completely negative (Fig. 2).

Table 2 Morphological, immunohistochemical and molecular features of BKPyV-associated UC.
Full size table
Fig. 1: H&E staining (1st column) and immunohistochemistry (IHC) for polyomavirus large T antigen (LTag, 2nd column) and p53 (3rd column) (×200 original magnification) of cases 1 to 5  (corresponding to lanes 1 to 5, respectively).
figure 1

BKPyV-associated UC showed micropapillary growth pattern (a, d, m) with variable extent, UC with extensive divergent differentiation with Müllerian type clear cell carcinoma (g) and unusual villoglandular pattern (j). All tumors showed nuclear expression of LTag (b, e, h, k, n) and p53 overexpression (c, f, i, l, o).

Full size image
Fig. 2: Additional IHC for case 3 with CIS (insert) and case 4.
figure 2

Both cases showed nuclear staining for GATA3 (a, b). Case 3 showed additional focal PAX8 staining in a subset of invasive carcinoma and carcinoma in situ cells (c). Case 4 was negative for PAX8 (d).

Full size image

All five tumors showed nuclear staining for LTag in the invasive and CIS component while benign urothelium and surrounding benign structures were negative for LTag (Fig. 3). As a productive lytic replication would also take place in surrounding normal tissue, we assume that LTag expression most likely originated from integrated virus and not from productive infection. Aberrant expression for p53 was found in the invasive and CIS component of all five tumors. Benign urothelium showed a wild-type pattern for p53 with faint basally accentuated patchy nuclear staining. HER2 amplification was found in two tumors: case 2 (immunoscore 1+) and case 5 (immunoscore 3+).

Fig. 3: IHC for polyomavirus LTag (×400 original magnification) of CIS in case 3.
figure 3

Pagetoid spread of CIS with nuclear expression of LTag adjacent to the invasive carcinoma. Note negativity in the normal urothelial, stromal and endothelial cells.

Full size image

TERT promoter and HRAS mutation analysis by SNaPshot

Mutations of the TERT promoter were found in 3/5 LTag expressing UC (Table 2). Case 3 could not be evaluated at position −57 and we obtained a PCR-product of reduced size at −146, possibly corresponding to a deletion at this position. Mutations of the HRAS gene were not found in 4/4 evaluable cases. Case 3 was not evaluable as no PCR-products could be generated. This might be explained by a reduced DNA quality and degradation at the regions of interest.

Viral detection and NGS

All five LTag expressing UC contained BKPyV-DNA with a viral load of about two BKPyV-DNA copies per cell (0.3–4.3 copies/cell; Table 3). We also tested six posttransplant UC without LTag expression for BKPyV-DNA and none contained BKPyV-DNA.

NGS analysis of the five LTag expressing UC could be performed only from tumor tissue but not from adjacent normal tissue due to limited quality and quantity of the samples. BKPyV integration was detected in 4/5 tumors with one distinct integration site in each of tumors 1, 3, and 4 and at least two integration sites in tumor 5. The integration sites were in intergenic and intronic regions of the tumor genome and affected different human genes: LGR4, DEK, RNF144B, THADA, SMC5, KLF9, PRKAR2A and SLC6A18 (Fig. S2). Breakpoints in the BKPyV genomes were located in LTag, VP2, and VP3. All four tumors with integrated BKPyV-DNA contained intact BKPyV early coding regions as shown by NGS reads covering these regions (Fig. S2) and shown by expression of the early gene product LTag.

RNA-seq was performed from fresh frozen tissue for tumor 5 and showed strong transcription of the early coding regions for small Tag and LTag but scarcely of late structural BKPyV genes (Fig. S3). RNA-seq could not be performed for cases 1–4 due to the low quantity and quality of the FFPE material.

In cases 1 and 3, the integrated BKPyV genomes had small deletions within the in the P-block [22] of the noncoding control region (NCCR) (Fig. S4). The five BKPyV isolates represented three different genotypes (Fig. S5).

The only tumor without detectable BKPyV integration (case 2) was found to contain HHV-6 DNA with an average viral load of about one HHV-6 DNA copy per cell (Fig. S6).

Analysis of the 50 most common cancer related genes altered in UC [37] showed only few alterations per case and no mutations in the TP53 tumor suppressor gene nor in the HRAS gene. Most of the detected alterations have been described in the ClinVar database and ranked as benign/likely benign or of uncertain significance (e.g., TTN: p.Val10240Phe, p.Val9503Ile, p.Glu5963Gly). Few pathogenic or likely pathogenic mutations were detected in cases 2, 3, and 4 (Table 3).

Table 3 Results of next generation sequencing.
Full size table

Discussion

Herein, we characterize five LTag expressing UC which were diagnosed between 2 and 18 years after kidney (n = 4) or heart (n = 1) transplantation. The first two cases were detected among 94 MPUC, and only retrospective analysis showed a posttransplant state. This observation triggered a retrospective screening of our records which yielded 15 posttransplant UC, with LTag expression present in 3/15 (20%). Others reported a somewhat higher LTag prevalence in UC in kidney transplant (31%) [38] or solid organ recipients (44%) [16]. BKPyV-associated UC can arise in graft and non-graft sites due to reactivation of latent BKPyV of donor or recipient origin. Case 3 developed in the renal pelvis of the kidney graft, while the other four UC affected the urinary bladder.

All five LTag expressing tumors were high-grade tumors and had at least partial variant morphology. Three were MPUC or showed a focal MPUC component. MPUC account for 0.7–6% of UC and are associated with a poor prognosis [39]. Two tumors showed extensive and unusual glandular patterns, one of them representing UC with extensive divergent differentiation with Mullerian type clear cell carcinoma, the other with villoglandular morphology. Since variant and unusual morphologic patterns have been reported in BKPyV-associated UC [15, 16, 40], we additionally investigated rare variant UC other than MPUC for LTag expression. Among 76 plasmacytoid, nested and large nested variant UC we did not find any positive cases. Additional search in two TMAs containing tissue cores of altogether 679 urothelial carcinomas including rare variant UCs (MPUC, sarcomatoid, plasmacytoid, neuroendocrine, lymphoepithelioma-like, nested, large nested, and clear cell) no LTag expressing cases were found, which confirms the low incidence of LTag expressing UC.

In the five LTag positive UC, IHC staining for LTag was restricted to tumor tissue. As a productive lytic replication would also take place in surrounding normal tissue, LTag expression most likely originated from integrated virus and not from productive infection.

Mutations in the TERT promoter were present in 3/5 cases, they are the most frequent alterations in UC and frequently found in MPUC [41, 42]. As TERT promoter mutations are detectable irrespective of stage and grade they might reflect early events in UC development. TERT promotor mutations could result in increased telomerase expression stabilizing tumor cell genomes and might facilitate transformation of cells with BKPyV-DNA integration. We found two MPUC with HER2 amplification which has been reported in 5–10% of advanced UC and in up to 42% of MPUC [43,44,45,46,47]. Assessment of the HER2 status can detect a subset of tumors that may benefit from anti-HER2 targeted therapies which are currently tested in clinical trials.

BKPyV-DNA was detectable in all five LTag expressing UC but was not present in six posttransplant UC without LTag expression. NGS showed chromosomal integration of BKPyV-DNA in 4/5 UC affecting four human genes, two of which (THADA, LGR4) have been implicated in cancer development [48, 49]. This study significantly enlarges the limited number of UC with proven BKPyV integration. Others have reported intronic, exonic and intergenic integration of BKPyV in the tumor genome with variant breakpoints in the viral genome [15, 16, 18, 19, 21, 22, 50]. In a similar manner, unique and random chromosomal integration sites of polyomavirus MCPyV are present in Merkel cell carcinoma [51]. Polyomavirus associated tumorigenesis is associated with unregulated LTag expression and subsequent p53 binding and inactivation [52, 53], while in UC without viral integration p53 inactivation is usually caused by mutations of the TP53 gene. According to this proposed mechanism, all five UC in our study revealed aberrant p53 expression by IHC but no TP53 gene mutations by NGS. In addition, NGS detected only one to five alterations per tumor within the 50 most common cancer related genes altered in UC [37]. There were very few pathogenic or likely pathogenic mutations and most of the affected genes (PCLO, PANK2, RYR1, FLG, TTN, and ATM) were also reported as mutated in a pan cancer analysis based on TCGA data [54], which might reflect common biological processes in many cancers. As human embryonic kidney cells transfected with BKPyV-DNA were reported to grow only in the presence of a mutated and activated HRAS oncogene [55], we specifically searched for HRAS mutations in the LTag expressing UC, but detected none.

The paucity of pathogenic genomic mutations suggests that carcinogenesis occurs via unregulated LTag expression in BKPyV-associated UC. Sirohi et al. [16] reported a lower number of deleterious genomic mutations in LTag expressing UC compared with UC without LTag expression.

We detected a break point in the viral LTag in one case, but NGS data suggested that BKPyV integrated as a concatemer. Therefore, a full functional set of BKPyV genes was present in the tumor facilitating the expression of LTag. Insertional breakpoints within the BKPyV large or small T antigen genes have also been reported by others in UC which nevertheless expressed LTag, indicating that BKPyV integration was also in a concatemeric form [16, 21]. We could perform transcriptional analysis only for case 5 where RNA-seq showed strong transcription of the early region genes for small and LTag. There was nearly no transcription of the late structural genes suggesting very limited or absent lytic replication of BKPyV.

In two UC, the integrated BKPyV genomes had small deletions within the viral NCCR, which have been attributed a role in deregulation of LTag expression [10]. However, two other UC with BKPyV integration and one UC without detectable BKPyV integration did not have such mutations.

In case 2, we failed to detect BKPyV integration. FFPE material from this tumor was limited and 10 years old, consequently DNA was of relatively low quality. We obtained low genome coverage and the lowest BKPyV read numbers from this tumor, therefore we might have missed BKPyV integration. Interestingly, we detected about one copy HHV-6 DNA per tumor cell, suggesting inherited chromosomally integrated HHV-6 (ciHHV-6). We had no additional tissue available from this patient to test this hypothesis. HHV-6 can integrate near the subtelomeric/telomeric junction of chromosomes and can be genetically transmitted, which occurs in 0.6–2% of the general population [56]. HHV-6 integration is a proposed mechanism to establish viral latency but no association with human cancer has been observed. Our search in the TCGA supplemental data revealed no UC with HHV-6 integration among 412 muscle invasive UC [37].

In summary, we describe five posttransplant UC with distinct morphology and LTag expression. Unique and random BKPyV integration was detectable in four tumor genomes. The proposed mechanism of tumorigenesis is BKPyV lytic reactivation under immunosuppression leading to erroneous chromosomal BKPyV integration with unregulated LTag expression and subsequent p53 inactivation. IHC screening for LTag should be performed in posttransplant UC to identify patients with BKPyV-associated UC. Some of these patients might benefit from cessation of immunosuppression as reported in individual cases [21, 22] or from therapy with checkpoint inhibitors as recently shown for Merkel cell carcinoma [57, 58].