Introduction

MiT family translocation renal cell carcinoma (MiT-RCC), incorporated into the World Health Organization classification in 2016 [1], harbors translocations involving the TFE3 or TFEB gene. The TFE3 gene is located on Xp11, and tumors harboring a TFE3 gene rearrangement were originally referred to as Xp11 translocation RCC. The three most common translocation partners are ASPL (ASPSCR1), PRCC, and SFPQ (PSF), which are located on chromosomes 17q25, 1q21, and 1p34, respectively [2, 3]. Since then, additional partner genes have been identified including CLTC, PARP14, RBM10, NONO, MED15 [4], DVL2 [5], and RBMX [6]. The TFEB gene is located on chromosome 6. TFEB is overexpressed in most MiT-RCC involving TFEB due to a t(6;11) translocation leading to the fusion of TFEB with MALAT1. In addition, TFEB has several other fusion partners including KHDRBS2, COL21A1, CADM2, CLTC, EWSR1, and ACTB [7]. More recently, RCC with TFEB amplification has been described as a unique subtype of RCC associated with poor outcome [8].

MiT-RCC was initially thought to be a childhood tumor characterized by papillary architecture composed of epithelioid clear cells with clear to eosinophilic cytoplasm [3]. There is now increasing evidence indicating a wide spectrum of morphologies occurs in MiT-RCC [4], making the differential diagnosis challenging. Immunohistochemical stains for markers such as cathepsin K are helpful to a certain degree but are not always reliable [9]. MiT-RCC and RCC with TFEB amplification are distinct entities, and recognizing these as a specific category of RCC is important [10,11,12,13]. Therefore, accurate diagnosis is crucial because variable clinical behavior and therapy are observed with MiT-RCC [2, 4, 14].

Fluorescent in-situ hybridization (FISH) using break-apart probes for TFE3 and TFEB is the gold standard for the diagnosis of MiT-RCC tumors [10]. However, FISH cannot identify fusion partners and more importantly, FISH is unreliable for rare MiT-RCC cases that have small intrachromosomal gene inversions [6, 15, 16]. Therefore, alternative assays are needed.

The Archer FusionPlex assay utilizes anchored multiplex PCR (AMP) technology [17]. AMP technology combined with next-generation sequencing (NGS) can detect gene fusion events in a partner-agnostic manner and is compatible with low nucleic acid input from formalin-fixed paraffin-embedded (FFPE) tissues [17]. The objective of this study is to evaluate the performance of the Archer FusionPlex assay for detection of MiT-RCC with translocations or intrachromosomal inversions. In addition, we evaluated whether elevated TFEB expression accurately predicts TFEB amplification.

Materials and methods

Tissue samples

A total of 68 FFPE cases including 49 cases of RCCs with known TFE3 and TFEB FISH status (Table 1) and 19 cases of sarcomas and carcinomas harboring non-TFE3/TFEB gene fusion events (see Supplementary Table 1) were retrieved from the archives of Departments of Pathology at the University of Verona, Pederzoli Hospital, Peschiera del Garda, Verona, Italy, and The University of Alabama at Birmingham, Birmingham, AL, USA. One hematoxylin and eosin (H&E) stained slide and five 10 µm thick unstained slides were cut from each selected FFPE block. H&E slides were reviewed for adequacy and tumor cellularity by a board certified pathologist. Seraseq Fusion RNA Mix v4 and Seraseq FFPE tumor fusion RNA v4 reference materials (SeraCare, Milford, MA) with 18 known gene fusion events were used as controls. This study was approved by the Institutional Review Board at the both institutions in accordance with the Helsinki Declaration.

Table 1 Fusion panel results of all 49 renal cell carcinoma cases with known FISH status.
Full size table

Nucleic acid extraction

FFPE tissue samples were scraped off the slides into microfuge tubes and tissue lysis buffer with proteinase K (440 µL) was added. Tissue was deparaffinized using Acoustic Paraffin Emulsification on the Covaris ME220 Adaptive Focused-ultrasonicator (Covaris, Inc. Woburn, MA). Total nucleic acid (DNA and RNA) was extracted from the lysate using the Covaris truXTRAC FFPE total Nucleic Acid Plus kit - Column (Cat. No. 520252; Covaris, Inc.), following the manufacturer’s instruction [18]. The extracted DNA was not used for this study. RNA quantity was assessed using the Qubit fluorometer (Life Technologies, Carlsbad, CA).

AMP-based targeted NGS

Target enriched cDNA libraries were prepared with a clinically validated, custom-designed Archer FusionPlex panel targeting 94 target genes, as per the manufacturer’s instructions (ArcherDX, Inc. Boulder, CO) [17, 19, 20]. The quality of amplifiable RNA was determined by the presequencing quality control cycle threshold (PreSeq QC CT), which was identified using real-time reverse transcriptase qPCR in duplicate for each sample. The assay utilizes AMP technology, which employs unidirectional gene-specific primers (GSPs) that enrich both known and unknown fusion gene partners [17]. All libraries were purified and quantified using KAPA Biosystems qPCR kit for Illumina (KAPA Biosystems, Wilmington, MA) and then sequenced on the Illumina NextSeq instrument (Illumina, San Diego, CA). Paired-end sequence data were generated by concatenation of lane level FASTQ files with Illumina’s Local Run Manager and uploaded to the Archer analysis unlimited software (version 6.2.7). Quality of sequencing was evaluated using Fusion QC score, which reflects the average number of unique reads per GSP2 and the average number of unique start sites per GSP2. Gene fusion events were analyzed using Archer analysis unlimited software (version 6.2.7) and JBrowse genome browser [20].

Analysis of TFEB mRNA expression

RNA expression values for all GSP2 primers of TFE3 and TFEB were captured as “rna_expression_visualization.tsv” files using Archer analysis software. Cases with Fusion QC score less than 10 were excluded. Average of all mRNA values of GSP2 primers for TFEB and TFE3 was calculated and compared between different tumor types.

FISH analysis

FISH on FFPE sections was performed with dual-color break-apart probes for TEF3 on chromosome Xp11.2 and TFEB on 6p21 (Cytotest Inc., Rockville, MD) as previously described [21, 22]. Centromeric chromosome 6 probe (CEP6) was used as the control probe (Vysis-Abbott, Rome, Italy). Scoring was performed by an experienced pathologist. At least 100 nonoverlapping, neoplastic nuclei were counted. Rearrangement, designated as FISH positive, was defined as red and green signals separated more than 2 signal diameters. The FISH-positive cutoff was 25%. Amplification was defined by the ratio of the mean copy number of TFEB signals to the mean copy number of CEP6 signals ≥ 10.

Statistical analysis

When applicable, the results were evaluated using Student’s t test. A p value equal to or less than 0.05 was considered statistically significant.

Results

RNA quality and sequencing quality

We evaluated the performance of the Archer FusionPlex Panel using Seraseq Fusion RNA Mix v4 and Seraseq FFPE tumor fusion RNA v4 control materials. The Seraseq Fusion RNA Mix is not derived from FFPE and is therefore of higher quality compared to the Seraseq FFPE tumor fusion RNA v4, which is similar to a routine FFPE sample. The Seraseq Fusion RNA control was analyzed on five different runs. PreSeq CT values were between 20 and 22 cycles, and Fusion QC values, defined as the average unique start sites per GSP2 control, were over 200, indicating excellent RNA and sequencing quality (Supplementary Fig. 1A). All expected fusion transcripts were detected across all runs. The Seraseq FFPE tumor fusion control was analyzed across 14 different runs. PreSeq CT and Fusion QC values of this control sample varied widely. Fusion QC inversely correlated with PreSeq QC CT values (y = −52.64X + 1508; r2 = 0.748) (Supplementary Fig. 1A). All 18 fusions were detected when Fusion QC was ≥100 (11/14 samples). However, the number of expected fusions that were detected dropped as the Fusion QC values decreased, with ROS1 and MET dropping out most frequently (Supplementary Fig. 1B). Even though Fusion QC ≥ 10 is considered “PASS” per the manufacturer’s recommendation, our data indicated that false negative results are observed when Fusion QC < 100 [23].

Concordance between FISH and fusion panel results

Concordance between the Archer Fusion panel and TFE3/TFEB FISH was 80% (39/49) for the 49 RCC cases (Table 1). Figure 1A shows Fusion QC values plotted against PreSeq CT values for these cases. Fusion QC inversely correlated with PreSeq QC CT values (Y = –18.10X + 598; R2 = 0.838; p < 0.0001). Sequencing quality metrics were important for TFE3/TFEB fusion detection. When Fusion QC value was ≥100, concordance of TFE3 and TFEB to FISH was 100% (Table 2). Concordance was 43% for TFE3 (8/17) and 91% (10/11) for TFEB when Fusion QC value was <100. Interestingly, even when QC < 10, concordance was 100% for TFEB (n = 4) indicating that the detection of gene fusions events is not consistent across all genes in the panel.

Fig. 1: Effect of age of FFPE block and PreSeq QC on Fusion QC and assay performance.
figure 1

A Correlation between PreSeq CT value and Fusion QC value in RCC cases (some cases were run in duplicate). Blue open square: FISH-positive case in which fusion gene transcript was detected. Red triangle: FISH-positive case in which fusion gene transcript was not detected. Green star: FISH-negative case. Linear regression equation Y = −18.10X + 598 (R2 = 0.838; p < 0.0001). B Fusion QC distribution in samples with newer (0–2 years) and older (3 years or older) FFPE tissue blocks. Each black dot (circle and square) represents a case with concordant FISH results and red dots represent the case with discordant results. Lines indicate mean ± SEM.

Full size image
Table 2 Concordance with FISH results in different Fusion QC values.
Full size table

Since Fusion QC and PreSeq QC values likely reflect the quality of the RNA, we correlated the age of the FFPE specimen with both the Fusion QC and the detection of TFE3 and TFEB gene rearrangements. Comparing FFPE samples less than 2 years old with samples older than 2 years, both Fusion QC (107.4 ± 11.2 versus 27.02 ± 4.93, respectively) and detection of TFE3/TFEB fusion events (95% versus 64%, respectively) were superior in samples less than 2 years old (Fig. 1B and Supplementary Table 2). In fact, all samples with Fusion QC ≥ 100 are less than 2 years old (Supplementary Table 3). Given that the majority of FFPE samples submitted for routine clinical testing are less than 2 years old, these data indicate that the Archer Fusion panel is reliable for clinical testing of both TFE3 and TFEB gene rearrangements.

We analyzed input RNA ranging from 250 (manufacture’s recommendation) to 20 ng and obtained sufficient numbers of sequence reads for fusion detection as long as tumor percentage is ≥20% (data not shown). These results are consistent with a previous study by Sussman et al. [24]. The limit of detection (LOD) of the Archer Fusion panel was determined by mixing RNA from two fusion positive cases at different ratios (Fig. 2). The mixing experiments were performed using TFE3 and TFEB cases with high Fusion QC scores to help mitigate any potential influence that poor amplification may have on determining the LOD. The mixing experiments were designed based on the premise that mixing FFPE-derived RNAs from MiT-RCCs with different fusion events is equivalent to a dilution study using RNA extracted from normal tissue. In addition, mixing studies are cost effective because you can determine the LOD for both targets in the same experiment. Mixing studies demonstrated TFE3 and TFEB fusion transcripts were detected with 5% variant fraction (Fig. 2).

Fig. 2: The mixing experiments using TFE3 and TFEB rearrange to determine the limit of detection (LOD).
figure 2

RNA from a case with RMB10-TFE3 (A, B) and a case with ACTB-TFEB (C, D) were mixed at different ratio and the number (A, C) and the percentage (B, D) of fusion transcript reads were plotted against the percentage of tumor. The number and the percentage of fusion transcript reads at each data point is shown in (E).

Full size image

The Archer Fusion panel identified intrachromosomal inversions in two cases (TFE3-RBM10 and NONO-TFE3) (Figs. 3 and 4; cases 3 and 5, respectively, in Table 1). The FISH patterns on these cases were unusual. Case 3 was a female whose tumor exhibited ~30% FISH closely split signals which is close to the 25% threshold for positivity (Fig. 4E). Case 5 was a male whose tumor exhibited a unusual split pattern with one green signal and one fusion signal (1G1F) (Fig. 4F). The distance between green and fusion signals appears to be 2–3 signal diameters.

Fig. 3: Schematic diagrams of fusion transcripts by intrachromosomal inversion.
figure 3

Exx represents exon xx of each gene. Thick black arrows indicate oncogenic fusion transcripts. A RBM10 → TFE3 (breakpoint: chrX:47041725,chrX:48895639). B NONO → TFE3 (breakpoint: chrX:70517788,chrX:48891766). C Schematic representation of chromosome X with TFE3, RBM10, and NONO.

Full size image
Fig. 4: Morphology (H&E) and fluorescent in-situ (FISH) images of RCCs with intrachromosomal inversions.
figure 4

H&E of RCCs with RBM10→TFE3 (case 3) shows a solid tumor with distinctive biphasic morphology, composed of larger epithelioid cells with clear and eosinophilic cytoplasm and smaller cells occasionally clustering around basement membrane material (A, B). TFE3 break-apart FISH shows 1R1G1F pattern with closely split signals (E). H&E of RCCs with NONO→TFE3 (case 5) shows a tumor with predominant papillary architecture lined by cells with clear cytoplasm with relatively uniform and round nuclei. Perivascular hyalinization is also apparent. Focal psammoma bodies are occasionally observed (C, D). TFE3 break-apart FISH shows predominant pattern of one green and one fusion (1G1F) (F). White arrows indicate green signals, red arrow indicates red signal, and black arrows indicate fusion signals (yellow).

Full size image

Evaluation of TFEB expression

We calculated the mRNA expression levels of TFEB and TFE3 and determined the TFEB/TFE3 mRNA expression ratio for cases with Fusion QC (≥10) (Fig. 5 and Table 3; individual results are shown in Table 1). TFEB mRNA expression and TFEB/TFE3 ratio were significantly higher in RCCs with TFEB fusion compared to RCCs with TFE3 fusion, RCCs with no fusion, or nonrenal tumors. Similarly, RCCs with TFEB amplification had significantly higher TFEB expression levels and TFEB/TFE3 ratios compared to cases with TFE3 fusion events and cases without fusion or amplification events (Table 3). There was no statistically significant difference in TFE3 mRNA levels between the groups. Using a cutoff value of 0.5 for the TFEB/TFE3 ratio, all TFEB altered tumors (11/11) were above 0.5, and 96% (25/26) of the tumors without TFEB alteration were below 0.5. The overall concordance of the 0.5 expression ratio with TFEB alteration is 97% (36/37) with 100% sensitivity, 96% specificity, 91% positive predictive value, and 100% negative predictive value. These data suggest that TFEB expression levels and the TFEB/TFE3 expression ratio are very accurate for predicting RCC cases harboring TFEB amplification that require confirmatory testing.

Fig. 5: Evaluation of TFEB expression levels in different types of tumors.
figure 5

Expression levels of TFEB mRNA (A) and TFEB/TFE3 mRNA expression ratio (B) in renal tumors with TFE3 fusion, TFEB fusion, TFEB amplification (Amp), no fusion, and tumors from other organs (carcinomas and sarcomas). N indicates number of cases and p value was calculated by Student’s t test.

Full size image
Table 3 Summary of RNA expression levels of TFE3, TFEB, and TFE3/TFEB ratio in different tumor types.
Full size table

Discussion

It has been more than two decades since the Xp11 translocation RCC was first reported [25,26,27,28]. Clinical and histopathological features of MiT-RCCs are more diverse than originally described [4, 29], and the accurate diagnosis of these tumors is challenging. Currently, there is no standardized work-up for MiT-RCC. TFE3 immunohistochemistry (IHC) for screening has been proposed [29, 30]. However, IHC can be problematic due to low specificity and inconsistent performance [29]. FISH is currently the gold standard for diagnosis but has several drawbacks. The fusion partner of TFE3 and TFEB is not identified by FISH. The lack of a consensus threshold for interpreting FISH-positive nuclei [29] and the technical challenge of interpretation—particularly for cases with TFE3 intrachromosomal inversions, are also limitations for FISH analysis [15, 31]. By analyzing a large cohort of MiT-RCC with a variety of TFE3 or TFEB gene alterations, we demonstrated that the Archer FusionPlex assay reliably identifies TFE3 and TFEB gene fusion events—including intrachromosomal TFE3 inversions and TFEB gene amplification.

Our study demonstrates that fusion detection is not uniform across all genes in the panel. Fusion genes involving TFEB were detected in 92% of all cases including those with low Fusion QC, whereas detection of TFE3 fusions was 100% when Fusion QC is >100 but falls to 40–50% when Fusion QC < 100 [23]. In support of nonuniform detection, we observed that identification of ROS1 fusions and Met exon14 skipping, which are both targets of the Archer Fusion panel, correlates strongly with Fusion QC score, whereas detection of ALK1-EML4 occurs over a range of Fusion QC scores including Fusion QC < 40 (Supplementary Fig. 1 and data not shown). Fusion QC values indicate confidence in the assay’s performance and demonstrate the critical importance of preanalytical assessment for quality control and quality assurance purposes. When no fusion transcript is detected and Fusion QC is less than 100, the results should be interpreted with caution. Clinical labs need to have policies for alternative testing to confirm the negative results in this scenario.

The reason for the differences in target detection is not clear but likely reflects variation in PCR amplification efficiency for the different GSPs in the panel [32]. When input RNA becomes the limiting factor, as reflected by a decrease in Fusion QC score, the differences in GSP efficiency becomes apparent. The importance of RNA quality and sequencing metrics for successful fusion detection has not been well characterized. A few studies have shown correlation between the PreSeq QC CT value and Fusion QC [23, 24], consistent with our study. Garcia et al. demonstrated that average read depth is significantly lower in cases where fusion events were not detected [32]. We observed robust detection and very little dropout of TFE3 and TFEB gene alterations using RNA from cases less than 2 years old. Taken together, certain target genes in the Archer Fusion panel are more susceptible to dropout. Fusion QC scores can predict these false negatives, but it is important to empirically determine which targets are susceptible. Using RNA from cases less than 2 years old largely avoids this pitfall.

Our study also demonstrated that the TFEB/TFE3 expression ratio was significantly higher in RCCs with TFEB amplification as compared to tumors with TFE3 fusion or no alterations. Consequently, the TFEB/TFE3 expression ratio calculated using the RNASeq data is highly useful for predicting cases with TFEB amplification. A simple TFEB/TFE3 ratio threshold of 0.5 results in 97% concordance to FISH results with no false negatives. Since TFEB amplification is relatively uncommon, the number of the cases with TFEB amplification that we were able to analyze was low. In addition, we did not have any cases exhibiting gain of TFEB/chromosome 6p (low levels of increased copy number rather than amplification) in our cohort. Those cases may represent chromosome 6 polysomy and seen in a variety of high-grade cancers [8, 12], but may be challenging to distinguish from the case with true amplification. Further prospective validation of this ratio for predicting TFEB amplification using various types of RCCs including cases with chromosome 6p gain is warranted. Interestingly, TFEB expression levels in TFEB amplified RCC are not as high as those in TFEB rearranged RCCs. These results are consistent with previous work, where TFEB gene expression levels were evaluated by digital droplet PCR [33]. Our work underscores the value of applying this same principle for other tumors with gene amplification, such as liposarcoma with MDM2 amplification, when running Archer FusionPlex sarcoma panel. However, expression levels for different gene amplifications events will vary, and appropriate validation studies are needed rather than generalizing a cutoff value from this study.

In summary, our study demonstrates that the Archer FusionPlex assay enables accurate molecular diagnosis of all MiT-RCCs in a convenient, single workflow by reliably identifying TFE3 and TFEB fusions including intrachromosomal inversions, as well as TFEB amplification. The assay performs well down to 20 ng of input RNA allowing testing of challenging or limited samples, such as core needle biopsies. Age of the specimen, RNA quality, and sequencing metrics are important parameters for successful detection of TFE3 and TFEB gene alterations. Caution is needed when sequencing is suboptimal, and alternative testing strategies should be considered in fusion-negative cases where Fusion QC values are <100 to avoid false negative results.