Introduction

Renal cancers are estimated to be diagnosed in 73,750 people in the USA in 2020, and to contribute to 14,830 deaths1. The 2018 GLOBOCAN report on the global incidence of renal cancers has described 403,262 patients with renal cancer, and 175,098 deaths caused by renal cancers worldwide2,3. Approximately 90% of all diagnosed renal parenchymal malignancies are renal cell carcinomas (RCCs)4.

A rare transformation called sarcomatoid dedifferentiation can occur in most RCC histological subtypes and portends an especially poor prognosis. RCCs in which sarcomatoid dedifferentiation has occurred are commonly referred to as sarcomatoid RCCs (sRCCs), and patients with sRCCs often present with advanced or metastatic disease and rarely survive >1 year5,6,7. Sarcomatoid features are present in approximately 4–5% of all RCCs8,9,10,11. However, this proportion can range from 1 to 29% depending on the primary histology and the reporting study7,9,10,11,12,13. Although infrequently diagnosed in the localized setting (~20–40% of all sRCCs)14,15, approximately 20% of patients with metastatic RCC harbour sarcomatoid dedifferentiation15. Patients commonly present with an sRCC at between 54 and 63 years of age6,7,13,14,16,17, and the male-to-female ratio ranges from 1.3:1 to 2:1 (refs6,13,14,18,19). Similar to all other RCCs, the mechanisms underlying this gender difference remain unclear. However, possibilities include historical gender differences in occupational exposure or smoking habits, or the influence of sex hormones on tumour biology20.

The natural history and prognosis of sRCCs are poor, as approximately 60–80% of patients present with advanced or late-stage disease14,15. Median survival is approximately 6–13 months and a higher percentage of sarcomatoid dedifferentiation on histology has been reported to confer a worse prognosis6,7,11,21,22. Retrospective reports have shown reduced overall survival and cancer-specific survival in patients with increased sarcomatoid features11,21. Compared with patients without sarcomatoid features, patients with ≥25% sarcomatoid features had ~30% increased overall risk of dying of any cause (HR 2.07, P = 0.048)11 and those with ≥30% sarcomatoid features had a 52% increased risk of dying because of an RCC (HR 1.52, P = 0.018), with each 10% increase in sarcomatoid features associated with a 6% increased risk of death due to an RCC (P = 0.028)21. Furthermore, patients with sRCC have worse survival than patients with an RCC without sarcomatoid features (non-sarcomatoid RCC) at every disease stage23. In a stage-for-stage comparison between a clear cell RCC and an sRCC, cancer-specific mortality was 1.4, 0.9 and 0.8 times higher in patients with sRCC for stages 1–2, 3 and 4, respectively23. Periods of survival >1 year are possible but are typically observed in patients diagnosed with early-stage disease (T1–T2)6. Nevertheless, even in the setting of localized disease treated with surgery, most patients with a non-metastatic sRCC experience recurrence within 2 years24. By contrast, >90% of patients with a non-sarcomatoid localized RCC treated with surgery remain recurrence-free 5 years after surgery25.

Survival of patients with an sRCC has shown little improvement over the past 2 decades. In the largest multi-institutional epidemiological study on sRCCs to date, 5-year survival between 2010 and 2015 was 23.5%14. In previously reported studies using the same national multi-institutional database between 2000 and 2009, 5-year survival was 27–37%23,26. This lack of survival improvement highlights the need for more effective treatment options and is likely a result of limited research availability, as sRCC is a rare form of RCC that is often at an advanced stage by the time of presentation.

In this Review, we discuss and consolidate the current knowledge on sRCCs. We first explore the origins of sRCC nomenclature to assist in the understanding of current naming conventions. We then review the common presentation of sRCCs and modern techniques for pathological diagnosis. We summarize contemporary knowledge on the molecular biology of sRCC, including next-generation genomic sequencing investigations and their findings. Finally, we discuss treatment options including surgery, radiotherapy and systemic therapy and summarize ongoing studies and recommendations.

History of sRCC

In 1943, Weisel et al.27 described renal tumours with sarcoma-like appearance. Pathologists later reported subsets of renal sarcomas with characteristics of RCCs that were clinically more aggressive than conventional RCCs28. The term “carcinosarcoma of the kidney” was used to distinguish these variants from both renal soft-tissue sarcoma and conventional RCCs28. In 1968, Farrow and colleagues29 coined the term “sarcomatoid renal cell carcinoma” to describe a renal malignancy closely resembling a sarcoma with pleomorphic spindle-cell and giant-cell morphology. The authors wrote that sarcomatoid RCC seemed to be “a complex tumour of an altogether different type” and that its appearance was similar to “that of a double neoplasm, one component of which is a [renal cell] carcinoma of the adult type and the other an undifferentiated pleomorphic [tumour] which appears morphologically sarcomatous”29. Subsequently, sRCC was considered to be a separate histology as its clinical behaviour was more aggressive than that of a conventional RCC29. However, the labelling of an sRCC as a distinct histological entity came under scrutiny, as pathologists recognized that sarcomatoid features could be found in conjunction with most histological subtypes of RCC. Work by Thoenes and colleagues published in 1986 resulted in the Mainz classification of renal tumours, echoing contemporary sentiments that the “sarcomatous” variant “in principle may be related to most cell types” and proposed sRCC as an end-stage dedifferentiated tumour30. By 1997, the American Joint Committee on Cancer (AJCC)31, Union Internationale Contre Le Cancer (UICC)31 and the Heidelberg classification of renal cell tumours32 all formally removed sRCC as a distinct histology and assessed it to be a feature that could be present within any histology of RCC. Following these events, pathologists met to standardize criteria for diagnosing and reporting sRCCs. In 2012, the International Society of Urological Pathology (ISUP) Consensus Conference established that sarcomatoid transformation could be designated within an RCC if it contained atypical spindle cells and resembled “any form of sarcoma”33. No minimum amount or percentage of sarcomatoid dedifferentiation is required to establish the diagnosis of sRCC33. The 2012 Consensus Conference also updated the definition of grade to include RCC with sarcomatoid dedifferentiation in the category of WHO–ISUP grade 4 and mandated that the background epithelial RCC histology must be reported (e.g. papillary RCC with sarcomatoid features). Finally, it was determined that RCCs that contained ~100% sarcomatoid dedifferentiation with no recognizable epithelial RCC component should be labelled unclassified RCCs (uRCCs), WHO-ISUP grade 4. These 2012 reporting recommendations were officially endorsed in 2016 by the World Health Organization in their genitourinary classification guidelines34, aiding in the standardization of reporting techniques for sRCC. Now, although sRCC is not considered its own histology within RCC, the terms sarcomatoid features, sarcomatoid elements, sarcomatoid dedifferentiation, sarcomatoid differentiation, sarcomatoid component and sarcomatoid renal cell carcinoma are often used interchangeably to describe the same entity, namely sRCC.

Presentation

Clinical presentation

The presentation of sRCCs varies and depends on the stage of disease at diagnosis. Most patients present with locally advanced or metastatic disease and, therefore, approximately 90% are symptomatic at presentation7,10,17,19. Signs and symptoms are routinely non-specific and can consist of flank or abdominal pain (present in 51% of symptomatic patients), haematuria (22–34.7%), weight loss (18–22.6%), fatigue (15%), fever (6–10.6%), night sweats (6–12.6%) and cough and/or dyspnoea (6%)7,10,19,26. The most common sites for distant metastasis, in decreasing order of prevalence, include the lungs (34.6–71.0%), bone (13.0–44.0%), lymph nodes (25%), liver (12.6–23.0%) and brain (5.1–16.0%)14,15,17,24,35,36.

Radiographic presentation

To date, reliable preoperative imaging methods capable of identifying sarcomatoid dedifferentiation within a renal tumour do not exist15,37. Specific biomarkers for sRCCs have yet to be described and biopsy accurately detects sarcomatoid features within tumours at a low rate, ~7.5%19. Retrospective studies have attempted to elucidate features on CT and MRI imaging that would help diagnose sRCCs preoperatively. Unfortunately, these data are sparse and in the past 5 years only one study has investigated sarcomatoid features in CT imaging — the most commonly ordered imaging modality used in the radiographic assessment of renal masses.

A retrospective case–control study assessing preoperative CT images with texture analysis in sRCC (n = 20) versus ccRCC (n = 25), found that sRCCs were on average larger than ccRCCs, measuring 7.7 cm compared with 5.0 cm (P = 0.003), respectively38. Other features more commonly associated with sRCCs than ccRCCs included peritumoural neovascularity (P = 0.001), larger peritumoural vessels (P < 0.001) and a more heterogeneous texture analysis (P < 0.001). However, these findings are non-specific38.

On preoperative MRI, sRCC has been reported to have similar non-specific and subtle differences from ccRCC on T2-weighted images, such as more irregular and infiltrative morphology39,40. One study assessed 11 patients with pathologically confirmed ccRCCs with sarcomatoid dedifferentiation. Patients underwent preoperative MRI, the results of which were studied to identify differences between the sarcomatoid and epithelial tumour components. Using normal renal cortex (defined as T2 low-intensity signal) as a baseline, the investigators found that sRCCs had a more hypointense signal than ccRCCs, which had an isointense-to-high signal intensity40,41. Building on their previous study, the researchers applied a scoring system between 1 and 5 based on the presence of low-intensity signals within the tumour in a cohort of renal masses (10 sRCCs and 131 non-sarcomatoid masses). With a positive predictive value of 56% and negative predictive value of 99%, low-intensity areas in renal tumours of 1–3 cm were considered probable for sRCCs and tumours with low-intensity areas of >3 cm, multiple >1-cm low-intensity areas, or disruption of the pseudocapsule were considered definite sRCC findings41. This classification study41 has served as the basis for relative thresholds used in subsequent work42 to calculate the percentage volume of sarcomatoid involvement in RCC tumours on MRI. Sarcomatoid percentage estimated by MRI showed a positive Pearson correlation with the post-nephrectomy histology (R = 0.782, P < 0.001)42. However, further analysis revealed poor agreement with the histological examination, as MRI estimations underestimated percentage sarcomatoid values by at least 19%42.

PET–CT with 18F-fluorodeoxyglucose (18F-FDG) for detection of sRCCs has been previously described in four separate case reports43,44,45,46 to assess for recurrence, metastasis or initial staging43. However, in larger conventional RCC series 18F-FDG PET/CT has not demonstrated clinical usefulness47, and current professional practice guidelines from the American Urological Association48, European Society for Medical Oncology49 and the European Association of Urology50 do not routinely recommend its use for diagnosing or staging RCC. Currently, the utility of 18F-FDG-PET/CT in the diagnosis and management of sRCC is questionable. Most patients with sRCC present with gross metastatic disease, which can be effectively viewed using CT and MRI. Although studies are ongoing, the ability of CT and MRI to preoperatively identify sRCCs or the percentage of sarcomatoid involvement within a tumour remains unreliable for clinical decision-making at this time and at best underestimates the true extent of sarcomatoid disease42. Further research to identify radiographic features of sRCC is needed.

Pathology

sRCC is not a distinct histological subtype of RCCs but is a pattern of dedifferentiation characterized by the loss of epithelial features that are classic to RCCs34. Sarcomatoid dedifferentiation can be present within most histologies of RCC31,33,34, although it is most commonly seen in ccRCCs and chromophobe RCCs10,35. However, current incidence data are primarily based on smaller series (<150 patients)10,11,16, which require validation in larger datasets.

Histological diagnosis

sRCCs are often large (median ~10 cm) and can appear as dense grey or white areas within the tumour architecture that usually reveal a firm and fleshy cut surface when dissected6,18,51 (Fig. 1). The microscopic features of sRCCs often include a mixture of both epithelial and sarcomatoid components (Fig. 2). Unlike classic RCCs, the sarcomatoid component does not have recognizable epithelial components10 and histologically appears similar to sarcomas with pleomorphic and spindle cells with high cellularity and atypia52. These regions of sarcomatoid dedifferentiation can be heterogeneous or uniform. Approximately 90% of sRCC cases have coagulative necrosis, and roughly 30% contain some form of microvascular invasion6,7,10. Rhabdoid features are reported in approximately 15% of sRCC tumours53. According to the 2016 WHO guidelines, the presence of any amount of sarcomatoid dedifferentiation is sufficient for diagnosis of RCC with sarcomatoid features33,34.

Fig. 1: Gross sections of an sRCC after radical nephrectomy.
figure 1

Sarcomatoid renal cell carcinoma (sRCC) components are often large and can appear as dense grey or white areas within the tumour architecture and typically reveal a firm and fleshy cut surface when dissected. Arrow shows fleshy area that corresponds to sarcomatoid transformation and asterisk marks yellow and friable area of the tumour that corresponds to lower grade clear cell renal cell carcinoma.

Full size image
Fig. 2: Histopathology of sRCC.
figure 2

Part a (×40 magnification) shows a representative area of a sarcomatoid renal cell carcinoma (sRCC) with a well-differentiated clear cell carcinoma (ccRCC) component (asterisk) and a dedifferentiated sarcomatoid component (arrow). Magnified areas of the ccRCC (part b, ×200 magnification) and dedifferentiated sarcomatoid components (part c, ×200 magnification) are also shown. The ccRCC comprises cells with optically clear cytoplasm organized in alveolar/acinar architectural patterns, whereas the sarcomatoid component exhibits spindled cells associated with a dense lymphocytic infiltrate. Immunohistochemistry for carbonic anhydrase IX192 (a HIF-1α target gene that shows diffuse membranous localization in ccRCCs, which have increased HIF-1α signalling secondary to alterations of VHL) shows an area of transformation (part d, ×40 magnification), with strong cell membrane-localized expression in the well-differentiated ccRCC areas (asterisk) and gradual loss of expression in the more poorly differentiated sarcomatoid areas (arrow).

Full size image

Sarcomatoid dedifferentiation has been described in roughly 5–8% of ccRCCs, 8–9% of chromophobe RCCs, 2–3% of papillary RCCs, 2–3% of uRCCs and up to 29% of collecting-duct RCCs6,10. Overall, the most common histology seen with sarcomatoid features is ccRCC, which comprises ~80% of all RCC cases54. Reports have suggested that chromophobe RCCs, which comprise ~5–10% of all RCC cases55, might harbour a higher rate of sarcomatoid dedifferentiation than ccRCCs10,56, although the higher prevalence of ccRCCs made it difficult to validate this finding. Sarcomatoid change has also been reported in acquired cystic disease RCCs57 and hereditary leiomyomatosis RCC syndrome-associated RCCs58,59.

Diagnostic sampling of the primary tumour is recommended, as the well-differentiated epithelial component and dedifferentiated sarcomatoid component might only be focally present, which could lead to underdiagnosis. Immunohistochemical markers, including markers of renal histogenesis, can help to resolve this diagnostic dilemma60. In the absence of a low-grade epithelial component such markers are useful, as a tumour with spindled morphology has a wide differential diagnosis that includes sarcomatoid urothelial carcinoma, angiomyolipoma, dedifferentiated liposarcoma, sarcomatoid adrenocortical carcinoma, and mesenchymal neoplasms such as solitary fibrous tumour and synovial sarcoma, and others6. Thus, a diagnosis of sRCC relies heavily on the use of ancillary testing to establish renal histogenesis in the form of immunohistochemistry for transcription factors such as PAX2 and PAX8, which are both nephric-lineage transcription factors required for establishing renal-lineage cells and kidney formation and, therefore, a useful maker for renal epithelial tumours61,62. Additionally, markers of epithelial differentiation including pan-cytokeratins and/or epithelia membrane antigen can be employed, which are often focally expressed in the sarcomatoid component of RCCs but are negative in sarcomas. Finally, markers specific to underlying epithelial subtypes, such as carbonic anhydrase IX for clear cell RCCs, can also be employed63 (Fig. 2).

Role of biopsy

The use of percutaneous renal biopsy for suspicious masses has increased since the early 2000s64. Although percutaneous biopsies have shown satisfactory correlation with histological classification at the time of nephrectomy, determining the presence of sarcomatoid dedifferentiation can be difficult37,65. Identifying sarcomatoid features is limited primarily by the amount of tissue obtained by core biopsy and by the number and spatial distribution of biopsy cores. Sarcomatoid features often comprise <50% of sRCC tumours, which are commonly large heterogeneous masses15. In a study of 199 patients with an sRCC, biopsies of the primary tumour or metastatic site prior to nephrectomy or metastasectomy detected the presence of sarcomatoid features in 7.5% of patients19. Thus, the sensitivity of preoperative biopsies for sRCCs is low and is likely non-diagnostic for these features. Multi-quadrant biopsies have shown improved sensitivity in diagnosing sRCCs66. In a study comparing standard single biopsy with multi-quadrant biopsy (biopsies from at least four separate solid enhancing areas in the tumour), 23 of 96 large RCCs studied on post-nephrectomy final pathology were sRCCs. Sensitivity for identifying sarcomatoid features was higher using the multi-quadrant technique than the standard biopsy technique (86.7% versus 25.0%, P = 0.0062)66. To date, the most reliable method for diagnosing sRCC is based on post-surgical final pathological review by a dedicated genitourinary pathologist.

Metastases

Metastases of sRCCs can exhibit sarcomatoid features similar to those in the primary tumour. Information on the differences between sarcomatoid and non-sarcomatoid metastasis from sRCC is limited. However, patients with >30% sarcomatoid dedifferentiation within their primary tumour frequently have sarcomatoid features in their metastases9.

Tumour biology

An sRCC is composed of two separate cell types, the sarcomatoid (mesenchymal) component, and the RCC (epithelial) component. The mechanism by which sarcomatoid dedifferentiation arises within RCCs is not clearly understood; however, evidence suggests that the sarcomatoid component might originate from a common cell-of-origin, resulting in cells that lose their epithelial characteristics and gain mesenchymal characteristics in a process known as epithelial–mesenchymal transition (EMT)5,67,68,69,70. EMT is an essential transformation process in early development to generate layers of specialized tissue. However, EMT can also contribute to tumorigenesis71. EMT is regulated by a group of transcription factors including Snail, Zeb and Twist, which cause hallmark downregulation of epithelial markers such as E-cadherin and upregulation of mesenchymal markers such as N-cadherin60,72 (Fig. 3). E-cadherin is a cell membrane protein important in cell–cell adhesion and is attached to the cytoskeleton via the protein β-catenin. During EMT, β-catenin translocates to the nucleus where it acts as a transcription factor for Snail, a transcriptional repressor of E-cadherin71,73,74,75,76. As a result, epithelial cells lose their phenotypic features (such as expression of E-cadherin) and subsequently gain mesenchymal characteristics, leading to an increased ability to metastasize77, which might contribute to the aggressive features of sRCCs60 (Fig. 3). Evidence of EMT in sRCC was reported in a study of 21 tumours containing matched epithelial and sarcomatoid components immunohistochemically examined for known markers of EMT68,78. In the sarcomatoid components, expression of E-cadherin was reduced and expression of N-cadherin, Snail and β-catenin was increased68. Notably, E-cadherin and N-cadherin were both found to be highly expressed in the RCC (epithelial) components; however, the physical localization of cadherin in the cell varied. In the sarcomatoid component, N-cadherin was predominantly present in the cytoplasm and, in the epithelial component, N-cadherin was primarily localized to the cell membrane. Subcellular localization to the cell membrane is consistent with epithelial cell–cell adhesion, whereas the movement of N-cadherin into the cytoplasm promotes cell motility79,80. The authors of this study on EMT in sRCCs68 conclude that these results provide strong evidence for sRCC as an example of EMT and support the leading theory that sRCC originates from a precursor epithelial component that has undergone EMT. Indeed, sRCC is considered a classic example of a mesenchymal phenotype rationalized by EMT, with multiple studies supporting this theory5,67,68,69,70,78,81,82. Moreover, in vitro investigations into sRCC cell lines have shown that the mesenchymal markers vimentin and N-cadherin are strongly expressed in tumour cell cytoplasm, with minimal expression of intracytoplasmic E-cadherin, providing further evidence that sRCC biology includes EMT82.

Fig. 3: Signalling pathways involved in EMT reported in sRCC.
figure 3

A sarcomatoid renal cell carcinoma (sRCC) is composed of two separate cell types, the sarcomatoid (mesenchymal) component, and the RCC (epithelial) component. The mechanism by which sarcomatoid dedifferentiation arises within RCC is not clearly understood; however, there is evidence that the sarcomatoid component may originate from a common cell-of-origin, resulting in cells that lose their epithelial characteristics and gain mesenchymal characteristics through a process called epithelial–mesenchymal transition (EMT). EMT can occur via multiple pathways including TNF, TGFβ, Wnt, MAPK and PI3K/AKT signalling to regulate expression of Snail, Zeb and Twist. Activation of these transcription factors results in the downregulation of epithelial markers (E-cadherin) and upregulation of mesenchymal markers (N-cadherin). E-cadherin is a cell membrane protein that is important in cell–cell adhesion; however, during EMT these intercellular tight junctions (E-cadherin) break down and transform the cell into a more mesenchymal phenotype that increases the likelihood of tumour cell metastasis.

Full size image

In addition to genomic alterations, EMT can be related to epigenetic regulatory mechanisms such as methylation of gene promoters, histone modifications or microRNA-induced expression changes83,84,85,86. Studies on epigenetic regulatory mechanisms specific to sRCCs are not yet available, although these mechanisms have been reported in RCCs83,84,87. In a 2010 study that used clinical tumour samples and xenograft models to assess genomics and promoter methylation in the establishment of RCC metastasis83, variations in methylation contributed to the expression of pro-metastatic mesenchymal genes in non-metastatic RCC cells. In non-metastatic RCCs, epigenetically silenced genes were identified via demethylation using DNA methyltransferase inhibitor 5′-AZA, which leads to upregulation of pro-mesenchymal genes such as S100A488,89 — a known fibroblast marker that mediates metastasis and EMT88. Without 5′-AZA treatment, pro-mesenchymal S100A4 remained methylated (downregulated) in non-metastatic RCC cells83. By contrast, S100A4 in metastatic RCC cells was hypomethylated and thus spontaneously upregulated at baseline. These results suggest that epigenetic mechanisms can lead to a mesenchymal phenotype in RCCs that enables tumour progression. Indeed, the impact of promoter methylation on EMT is not unprecedented, as altered methylation of gene promoters has resulted in EMT tumour progression in breast cancer86. Such epigenetic mechanisms may contribute to the aggressive mesenchymal phenotype68,78 observed in sRCCs. Evidence of these epigenetic mechanisms was identified in a 2013 study that characterized an sRCC cell line for cancer modelling82, in which sRCC cells were found to be positive for cytoplasmic N-cadherin, vimentin and S100A4. Notably, S100A4 was present in the cytoplasm in 52.7% of sRCC cells on cytofluorometric analysis, and vimentin and N-cadherin were expressed in the cytoplasm in 99.8% and 81.2% of sRCC cells, respectively. S100A4 is an example of a gene integral to EMT (and metastasis) in RCC that is regulated through methylation modulation. Thus, it might also be possible that the same epigenetic mechanisms could be exploited for the methylation or demethylation of corresponding promoter sequences of interest as a future treatment in sRCC.

Framework for genomic investigation

Before the advent of next-generation sequencing only a few studies examined genomic aberrations present in sRCC. Four notable studies63,68,69,90 have created a foundation for modern genomic analysis of sRCC.

In 1995, Oda et al.90 compared matched epithelial and sarcomatoid components in 14 tumour samples and used PCR to assess the mutational status of TP53 and HRAS, both of which are known to be associated with cancer progression. Although no HRAS mutations were observed, 78.6% of tumours had TP53 mutations within their sarcomatoid component, and only 14.3% of tumours had TP53 mutations in their epithelial component. Furthermore, p53 overexpression was observed on qualitative immunohistochemical analysis in sarcomatoid components compared with their corresponding epithelial components. The authors concluded that TP53 mutation likely results in overexpression of mutated p53 and is critical for sarcomatoid transformation90.

In 2005, Jones et al.69 conducted a study of 22 patients with sRCC cases to assess the molecular and clonal relationship between matched ccRCC epithelial and sarcomatoid components in laser-microdissected tumours. X chromosome inactivation and loss of heterozygosity (LOH) at loci previously implicated in RCCs (3p14, 7q31, 8p21, 9p21, and 17p13) were analysed69. Patterns of allelic loss were variable in both epithelial and sarcomatoid components, highlighting substantial genetic heterogeneity within individual RCCs69. In 13 of 14 female patients, the same pattern of X chromosome inactivation was observed in matched epithelial and sarcomatoid components, supporting a clonal cell-of-origin theory69.

In 2007, Tickoo et al.63 assessed the expression of hypoxia inducible factor-1α (HIF1α), glucose transporter 1 (GLUT1), carbonic anhydrase IX (CAIX), and vascular endothelial growth factor (VEGF) in 34 RCCs with sarcomatoid dedifferentiation (22 ccRCCs and 12 non-ccRCCs) using immunohistochemistry. Loss of the von Hippel–Lindau (VHL) tumour suppressor gene is a defining feature of ccRCCs91,92. The VHL protein assists in the activation of a E3-ubiquitin complex responsible for degrading HIF. In the presence of oxygen normotension, HIF is marked for ubiquitylation. In cells with low oxygen tension or loss of functional VHL (as seen in ccRCCs), HIF escapes degradation, and activates downstream targets that assist in tumorigenesis, including VEGF and GLUT1 (ref.63). In the study by Tickoo et al.63, sarcomatoid dedifferentiation in ccRCCs was associated with higher levels of HIF than sarcomatoid dedifferentiation in non-ccRCCs, suggesting that sarcomatoid components maintain some similarity to their primary histology.

In 2011, Conant et al.68 examined known markers of EMT in matched epithelial and sarcomatoid components in 21 sRCCs. Within the sarcomatoid component, a loss of epithelial markers and a gain of mesenchymal markers were observed, further supporting EMT as a mechanism for transformation to sRCC68.

Next-generation sequencing

The use of next-generation sequencing to evaluate the genomics of sarcomatoid transformation has provided additional evidence of the common cell-of-origin theory93,94,95,96,97.

In 2016, Bi et al.93 used whole-exome sequencing to examine sarcomatoid dedifferentiation in 21 ccRCCs with epithelial, sarcomatoid and normal kidney components. Two hypermutated tumours suspected of being biologically different were excluded from analysis. Across the remaining 19 tumours, 41.7% of somatic single-nucleotide variants (SSNVs) were shared between epithelial and sarcomatoid components. The most frequently shared SSNVs were in VHL, PBRM1, PTEN and SETD2. TP53 was the most commonly mutated gene in sarcomatoid components: mutations were present in the sarcomatoid element of 31.6% of tumours, and not present in any of the matched epithelial components. In addition, some sarcomatoid components had increased expression of cancer driver genes such as BAP1 (10.5%) and ARID1A (15.7%), which were mutually exclusive with TP53 and with each other. Furthermore, several novel SSNVs were reported in the sarcomatoid component, including mutations in FAT1, FAT2 and FAT3, TSG101, LRIF1, RQCD1 and PTK7 (ref.93). Overall, the sarcomatoid component had a higher mutational burden than the epithelial component and sarcomatoid-specific LOH on chromosomes 1p, 9, 10, 14, 17p, 18 and 22 (ref.93). This study was one of the first to suggest — using three major findings — that sarcomatoid components in sRCCs originated from a pre-existing epithelial component through a process of dedifferentiation. First, the authors showed that the most frequently mutated genes shared between the epithelial and sarcomatoid components were those commonly mutated in ccRCCs. Second, the frequency of SSNVs in known cancer driver genes was greater than five times higher in the sarcomatoid versus the epithelial regions. Last, mutation of TP53 was more common in sarcomatoid elements than in non-sarcomatoid elements, suggesting a specific role for TP53 in the development of sarcomatoid elements, in addition to recurrent mutations and/or segments of LOH affecting other known cancer genes. These findings lend support to a pathogenic sequence theory in which somatic mutations that occur in ccRCC drive dedifferentiation to a sarcomatoid state.

Malouf et al.94 also sought to identify genomic alterations in sRCC in a 2016 study. In addition to ccRCC, papillary, collecting duct and uRCC primary histologies were included for analysis. In their study, 26 patients with sRCCs in three unique cohorts underwent genomic profiling. The first cohort comprised three sarcomatoid ccRCCs from three patients with matched microdissected epithelial and sarcomatoid components who underwent targeted sequencing using a panel of 236 frequently mutated cancer-related genes and 37 introns frequently rearranged in cancer. Two tumours showed identical mutational profiles between their epithelial and sarcomatoid components. However, the third tumour had distinct inactivating mutations of PTEN and TP53 between the epithelial and sarcomatoid components, and JAK2 was amplified in the sarcomatoid component. In the second cohort, 26 sRCC tumours from 26 patients (23 patients plus the previous 3) associated with various primary histologies were analysed. The most frequently mutated gene was TP53 (in 42.3% of tumours), followed by CDKN2A (26.9%) and NF2 (19.2%). In the final cohort, whole-exome sequencing was performed on four non-microdissected tumours from four patients with sarcomatoid ccRCCs. Median mutation rate was lower in these four sarcomatoid ccRCCs than in The Cancer Genome Atlas downloaded dataset of 446 ccRCCs98 — 37.5 vs 49 mutations per case, respectively. Additionally, Sanger sequencing of VHL and TP53 in multiple regions from the primary tumours of two patients did not show intratumoural heterogeneity for these two genes.

A subsequent study that included 252 sRCCs from a total dataset of 2,636 RCCs supported the findings of Malouf et al.94 by showing increased prevalence of JAK2 (9p24.1) amplifications in sRCCs: JAK2 amplifications were present in 5.95% of sRCCs compared with 0.6% of all RCC tumours (P < 0.001)99. Additionally, co-amplified genes at the 9p24.1 locus included PDL1 and PDL2, and these amplification events correlated with immunohistochemistry-based programmed cell death 1 ligand 1 (PDL1) expression (mean H score: 222/300, n = 10). Amplification of PDL1 and PDL2 was often absent in low-grade epithelial components and restricted to the sarcomatoid tumour component99 (Fig. 4). Gupta et al.99 documented the presence of constitutive PDL1 expression primarily in high-grade components, although the exact frequency was not documented owing to the study design (scoring of tissue microarrays of high-grade RCCs). However, these findings were supported by the observations of Kawakami et al.100

Fig. 4: PDL1 expression in sRCC.
figure 4

Representative haematoxylin-and-eosin-stained images (left) of a 9p24.1-amplified clear cell renal cell carcinoma (ccRCC) with sarcomatoid transformation are depicted alongside corresponding immunostaining for PDL1 (right). Staining for PDL1 is absent in areas with a clear cell component and shows constitutive expression in areas with a sarcomatoid component, which is higher grade than the clear cell component. Constitutive expression of PDL1 in higher grade sarcomatoid components implies an underlying molecular event such as an amplification in JAK2, PDL1 and PDL2 at the 9p24.1 locus. Patients with this expression pattern have the potential for an enhanced response to immune checkpoint inhibitors owing to the increased expression of PDL1. Parts a and b show adjacent clear cell and sarcomatoid areas (×40 magnification). Parts c and d show a representative area with clear cell morphology (×200 magnification). Parts e and f show a representative area with sarcomatoid morphology (×200 magnification). Reprinted from ref.99, Springer Nature Limited.

Full size image

Wang et al. assessed genomic and molecular characteristics of 40 frozen sRCCs with papillary, chromophobe and clear cell histologies through unsupervised clustering analysis of copy number and transcriptional data, and showed that sarcomatoid dedifferentiation segregated according to primary histology (clear cell, chromophobe, papillary) rather than to epithelial or sarcomatoid morphological components96. This finding suggests that sRCCs can be classified by primary histology instead of being labelled as a broad disease category of sRCCs. Whole-exome, single-nucleotide polymorphism characterization and RNA sequencing, showed that the overall mutational burden of sRCCs was similar to that of RCCs96. Gene mutations shared across all sRCC samples included VHL, C10orf113, BAP1, TMEM97, CALML3, IL15 and, most notably, PTEN, TP53, NF2 and RELN. When analysing specific histologies, PTEN was more frequently mutated in sarcomatoid ccRCCs than in ccRCCs, TP53 mutations were elevated in sarcomatoid chromophobes and ccRCCs compared with their non-sarcomatoid counterparts, and NF2 mutations were associated with sarcomatoid papillary RCCs. RELN mutations observed across all sRCC histologies were notable because RELN encodes Reelin, an extracellular matrix protease that regulates microtubule function and cell mobility101. Reelin inhibits TGFβ-1-induced cell migration102, which might play a role in metastasis and tumour aggressiveness. However, overexpression of TGFβ-1 can reciprocally suppress Reelin expression via the transcription factor Snail, a known activator of EMT102. Thus, in the event of Reelin loss (i.e. RELN mutation), upregulation of TGFβ-1-induced cell migration and increased expression of mesenchymal markers are observed102. Wang et al.96 reported upregulation of TGFβ-1 signalling across all sRCCs, which could be explained by similar elevation of RELN mutations across all histologies. This finding suggests that Reelin might be an anti-metastasis target for future drug development, although further investigation is necessary. Additionally, the study by Wang et al.96 was one of the few to report patient outcomes in relation to genomics and ccRCCs. Mutations in TP53, PTEN and RELN were associated with sarcomatoid change in ccRCCs (P = 1.64 × 10−6, OR = 6.53); however, in sarcomatoid ccRCCs, overall mutational frequencies of TP53, PTEN and RELN were lower than for VHL (P = 0.035), a hallmark of ccRCCs. These results suggest that mutations in TP53, PTEN and RELN occurred later in tumorigenesis than mutations in VHL. Notably, loss of VHL or deletion of 3p21–25 genes (VHL, PBRM1, SETD2 and BAP1) was associated with increased overall survival, and mutations in TP53, PTEN, RELN were associated with decreased overall survival96. Similarly, mutations in TP53, PTEN, RELN, BAP1 and SETD2 were associated with sarcomatoid change, and mutations in VHL and PBRM1 negatively predicted sarcomatoid change.

In an unrelated molecular analysis of 62 aggressive primary uRCC specimens, sarcomatoid architectural patterns were found in a subset of uRCCs with NF2 loss. These results raise the possibility that aggressive uRCCs share a common molecular aetiology with sRCCs involving NF2 loss. Interestingly, NF2 is known to be a key effector of the Hippo–YAP pathway, which affects tissue growth, proliferation and differentiation as well as cell migration103. Dysregulation of this pathway has been linked to uncoordinated cell growth and malignancy formation, and the key pathway effectors YAP and TAZ often accumulate in the nucleus to drive proliferation104,105. Wild-type NF2 promotes a signalling cascade that results in phosphorylation of the transcription factors YAP and TAZ, resulting in inactivated cytoplasmic accumulation. However, mutated NF2 will deregulate this cascade, causing YAP and TAZ to remain unphosphorylated, with the ability to translocate into the cell nucleus and promote cell proliferation106. The postulated role of the Hippo-YAP pathway in sarcomatoid dedifferentiation is supported by the promotion of EMT by the key effectors of the pathway — YAP1 (refs107,108,109) and TAZ110,111. YAP1 and TAZ are transcriptional regulators and are both associated with EMT signatures in solid tumours through upregulation of known EMT genes such as vimentin, fibronectin, SLUG and ZEB1107.

In a 2020 study112, Malouf and colleagues leveraged findings from previous sRCC genomic investigations that had shown mutations in Hippo–YAP pathway effectors, such as NF2 (refs94,96,105) and FAT2 (ref.93), to guide a dedicated study investigating Hippo–YAP pathway regulation in sRCCs. Using 49 clear cell sRCCs, targeted sequencing was performed in microdissected sarcomatoid and epithelial components and, together with non-microdissected sRCCs, compared with 268 non-sarcomatoid RCCs. The authors separately assessed the effects of NF2 knockout and reconstitution on sRCC proliferation both in vitro, using an NF2-mutant sRCC cell line, and in vivo, using male immunocompromised mouse xenografts. Results of targeted sequencing in 50 samples from 27 microdissected sRCC patient tumours showed mutations in VHL in 72%, SETD2 in 40%, PBRM1 in 34% and BAP1 in 26%112. In the 22 non-microdissected sRCC cases, mutations were observed in VHL (68%), and TP53 (27%). Mutational burdens observed within the microdissected and non-microdissected sRCCs were in agreement with previously reported sRCC genomic evaluation studies93,94,96. Notably, in addition to the previously mentioned mutational findings, Hippo–YAP pathway mutational alterations (NF2, FAT1, LATS1, LATS2, YAP1 and TAZ) were observed in 20% of the 49 sRCC cases that underwent targeted sequencing. By contrast, 5% of the 268 non-sarcomatoid RCC patients had Hippo–YAP pathway mutations. Thus, the frequency of Hippo–YAP pathway mutations in sRCCs was significantly higher than in non-sarcomatoid RCCs (P = 0.001). Moreover, the authors then show that YAP1 knockout and NF2 reconstitution inhibit proliferation and invasion in an NF2-mutant sRCC cell line both in vitro and in vivo. In vitro, knockdown of Hippo–YAP effector YAP1 resulted in suppressed cell proliferation and invasion, as well as tumour growth, and induced morphological cellular change. In vivo, YAP1 suppression reduced tumour growth in NF2-mutant xenografts in male immunocompromised NOD-SCID IL2Rg−/− (NSG) mice. Overall, increased Hippo–YAP pathway alterations in sRCCs and the in vitro and in vivo observations of Malouf and colleagues suggest that interference with Hippo–YAP pathway function might disrupt sRCC tumour growth and could be a novel therapeutic target that warrants future study.

Expression of immune checkpoint markers

The rapid expansion of immunotherapies approved for RCCs has generated a desire to identify biomarkers indicative of treatment response, and PDL1 and programmed cell death 1 (PD1) expression have emerged as candidate biomarkers113,114. PDL1 is a cell surface protein that binds to PD1 on activated T-lymphocytes and decreases their anti-tumour activity115. Thus, tumour cells often upregulate PDL1 expression as a way of avoiding immune surveillance116,117. Tumour expression of PDL1 is associated with improved responses to PD1 and PDL1 blocking agents such as nivolumab or atezolizumab, compared with targeted agents such as sunitinib and everolimus113,118,119. In conventional RCCs, PDL1 is a poor prognostic marker that is associated with high-grade tumours and tumour necrosis120,121. The prognostic potential of PDL1 and PDL2 was explored in a study that immunohistochemically evaluated 425 RCC samples and correlated the expression of these proteins with patient clinicopathological features114. Overall, PDL1 expression was seen in 9.4% of samples and PDL2 in 49.6%114. In ccRCCs, PDL1 expression was associated with adverse prognostic features, including higher WHO–ISUP grade, necrosis and sarcomatoid transformation (all P < 0.001)114. Furthermore, ccRCC PDL1 and PDL2 expression were both associated with shorter progression-free survival (P < 0.001 and P = 0.033, respectively) and shorter cancer-specific survival (P < 0.001 and P = 0.010, respectively)114. In non-ccRCCs, PDL1 positivity was associated with higher tumour stage and grade when 10.9% of tumour cells and 56.4% of tumour infiltrating mononuclear cells were PDL1 positive121.

Expression of PD1 and PDL1 has been examined in sRCCs122 (Supplementary Table 1). PD1 expression was observed in 25 of 26 sRCCs and PDL1 expression was observed in 14 tumours. By comparison, of 29 non-sarcomatoid ccRCC samples, 18 expressed PD1 and 5 expressed PDL1. Dual expression of PD1 and PDL1 was observed in 13 cases in the sRCC group compared with 1 in the non-sarcomatoid ccRCC group122. Similarly, in an immunohistochemical analysis of PDL1, PD1, CD4 and CD8 in 118 sRCC specimens and 92 non-sarcomatoid RCC (clear cell) specimens, sRCCs had higher PDL1 expression and higher PD1+ CD8+ cell density than grade 4 ccRCCs100. Moreover, 41% of sarcomatoid components within these tumours had an adaptive immune response phenotype (PDL1+, tumour-infiltrating-lymphocyte+)100.

In a 2019 study99, PDL1 expression was assessed in 398 high-grade RCCs, including 127 sRCCs, using H scores (the product of the percentage of tumour cells with PDL1 expression and intensity of staining graded on a scale of 1–3, maximum score = 300). A large number of sRCCs (~27.6%) showed high PDL1 expression (H score ≥ 50). Most of these cases exhibited adaptive patterns of PDL1 expression, but those with JAK2, PDL1 and PDL2 (9p24.1) amplifications showed constitutive patterns of PDL1 protein expression. Constitutive patterns of PDL1 expression seen in a subset of patients with sarcomatoid RCCs, in which every single tumour cell shows high levels of PDL1 expression, imply an underlying molecular event such as amplification of JAK2, PDL1 and PDL2 at the 9p24.1 locus. These patients may have enhanced or exceptional response to immunotherapy. However, analysis of outcomes based on PDL1 expression in this study did not reveal a significant prognostic effect, which was likely reflective of advanced disease. The WHO and International Society of Urologic Pathology (ISUP) classify sRCC as grade 4 disease and, accordingly, the majority of sRCC patients presented with advanced or high-stage disease99. Overall, these results add to an emerging body of data regarding the sRCC tumour immune microenvironment, which warrants further investigation into the possible benefits of PD1 and PDL1 immune checkpoint blockade therapy.

Treatment

Surgical management

In the setting of localized non-sarcomatoid RCCs, nephrectomy is a curative procedure. However, outcomes are less encouraging in patients with localized sRCCs21,24. Approximately 77–80% of patients who receive nephrectomy with curative intent for localized sRCC recur within 5–26 months7,24. Bulky disease is usually present at initial presentation, commonly requiring radical nephrectomy for complete resection.

In addition to large primary tumours, approximately 60–80% of patients with an sRCC present with metastatic disease7,14,26. In these patients, cytoreductive nephrectomy can precede systemic treatment. In retrospective series investigating the surgical treatment of metastatic RCCs, cytoreductive nephrectomy before initiation of systemic therapy resulted in improved survival compared with systemic therapy alone123,124,125,126,127. In a 2014 study of 189 patients with sRCCs, median survival was 10.2 months in patients who underwent cytoreductive nephrectomy compared with 5.5 months in those who did not128. Indeed, in the largest epidemiological study of sRCCs to date, cytoreductive nephrectomy performed in patients with sRCC with good performance status had an observable survival benefit, albeit minor, compared with non-surgically treated patients14. In this study in 472 patients with metastatic sRCCs, the 1-, 3- and 5-year disease-specific survival for those who underwent nephrectomy was 33.7%, 10.8% and 6.2%, respectively, compared with 11.5%, 1.9% and 0% in patients who did not. Overall, median disease-specific survival in those who underwent cytoreductive nephrectomy versus those who did not was 7 months (interquartile range (IQR) 3–17 months) versus 4 months (IQR 2–7 months)14. Moreover, multivariate cox proportional hazards modelling showed that cytoreductive nephrectomy was significantly and independently associated with improved disease-specific survival (HR 0.53, 95% CI 0.43–0.66, P < 0.001 (ref.14). Nevertheless, the role of cytoreductive nephrectomy in patients with sRCCs remains unclear, as retrospective data from existing reports are conflicting and no randomized controlled study exists to determine benefit. Although the aforementioned studies have shown data supporting the use of cytoreductive nephrectomy in sRCC patients, surgical management of metastatic sRCCs in the form of nephrectomy or metastasectomy has been alternatively argued to show variable benefit to survival and potentially delay initiation of systemic therapy owing to postoperative recovery129. In 419 patients who underwent cytoreductive nephrectomy, 62 patients with sRCCs had a median survival of 4.9 months compared with 17.7 months in those without sarcomatoid features15. Furthermore, contrary to non-sarcomatoid RCCs, those with sRCC were shown to have no survival benefit from post-nephrectomy metastasectomy130. A major difficulty of this debate is that, given the low rates of sRCC detection on preoperative imaging and biopsy15,19,37, most patients are not known to have an sRCC until after the nephrectomy is performed. Nevertheless, cytoreductive nephrectomy in selected patients with sRCCs might have palliative benefits, such as decreasing local symptoms of bulky disease and gross haematuria131.

Radiotherapy

RCC is largely considered a radio-resistant tumour132. The use of conventional radiotherapy as a primary treatment modality for RCCs was initially hindered by preclinical evidence suggesting inherent radio-resistance133 and implementation was further impeded by an absence of demonstrable benefit in clinical studies134,135. Accordingly, radiotherapy has been used sparingly in the treatment of RCCs for the past 50 years136. Similarly, sRCC responses to radiotherapy have been underwhelming. In a study in 408 patients with non-metastatic sRCCs, overall or disease-specific survival at 1, 3 and 5 years in those who received adjuvant radiotherapy was not significantly different from those who underwent surgical treatment alone137. Currently, radiotherapy is indicated mainly as a palliative measure for patients with metastatic disease or recurrent local tumour growth132,136. In the palliative setting, metastatic burden, metastasis accessibility, resectability and patient performance status are important harm–benefit management considerations138. Radiotherapy provides patients with a localized non-invasive therapy option that can alleviate pain, decrease neurological symptoms and improve haematuria from symptomatic metastases in otherwise non-surgical candidates132,139. Work is ongoing to assess the role of stereotactic ablative radiotherapy in RCCs, which delivers a higher (≥8 Gy) dose of radiation to the tumour than conventional radiotherapy (≤2 Gy)140,141. Data on new forms of radioablative therapy for the treatment of primary RCCs, such as proton beam therapy, are limited. A 2017 case report about a patient with an inoperable RCC, owing to morbid obesity and multiple comorbidities, is the only published instance of proton therapy for a primary RCC. Although a decline in glomerular filtration rate was observed from 34 ml/min/1.73 m2 to 29 ml/min/1.73 m2, no clinical symptoms of late radiation-induced toxic effects were observed and the patient had remained asymptomatic by the last follow-up point of 1 year142. This is a hypothesis-generating report, but it remains a single-institution case report and further research is required to determine whether proton therapy is appropriate in both RCCs and sRCCs. To date, no study has assessed the outcomes of patients with sRCCs treated exclusively with radiotherapy, and the available findings indicate that outcomes would presumably not be promising.

Systemic therapy

Systemic treatment for sRCC has been predominantly ineffective with few therapies producing durable responses35,143,144. Combinations of cytotoxic therapies, targeted therapies and immunotherapies are being used with varying effects. Selection of chemotherapeutic agents was initially driven in part by the reported benefits of doxorubicin and ifosfamide in sarcomas145. In 1987, Sella et al.36 performed one of the earliest studies assessing systemic therapy in patients with sRCCs. The study reports outcomes for 44 patients after doxorubicin chemotherapy, non-doxorubicin chemotherapy, hormonal therapy (medroxyprogesterone 17-acetate and androgen therapy), and interferon-α as separate groups. Overall median survival was 6–12 months. Notably, two patients had complete responses with doxorubicin and 4 patients treated with interferon-α had a median survival of 41 months36. These findings spurred future studies to focus on chemotherapy and immunotherapy as potential treatment options moving forward.

Chemotherapy

Chemotherapy efficacy in sRCCs (Table 1) has been explored over the past two decades in four notable studies146,147,148,149. First, a retrospective analysis of 8 patients with sRCCs treated with doxorubicin-based therapy and observed a median survival of 20–60 months in 3 patients146. Second, a multi-institutional phase II trial in 23 patients with sRCCs of combination doxorubicin and ifosfamide reported no clinical responses: median time to progression was 2.2 months and overall median survival was <4 months147. Third, a study in 10 patients with sRCC receiving combination antimetabolite gemcitabine and topoisomerase inhibitor doxorubicin observed two complete responses and one partial response148. The 2 patients who had complete responses remained disease-free at 6 years and 8 years150. Finally, a prospective phase II clinical trial evaluated the efficacy of combination doxorubicin and gemcitabine in patients with previously untreated sRCCs149. The Eastern Cooperative Oncology Group (ECOG) 8802 trial reported responses to treatment in 6 (16%) patients (5 partial responses and 1 complete response), and stable disease in 10 (26%) patients. Median overall survival (OS) was 8.8 months, and median progression-free survival (PFS) was 3.5 months149. Treatment response based on percentage sarcomatoid from available data (19 tumours from 19 patients) showed 7 tumours with 0–49% sarcomatoid features, 1 tumour with 50–74% sarcomatoid features and 11 tumours with 75–100% sarcomatoid features. The data on progression, stable disease, partial response and complete response from the ECOG 8802 trial, although descriptive and not subject to formal statistics, suggest that patients whose tumours had a high percentage (>75%) of sarcomatoid component might benefit from cytotoxic therapy, irrespective of the underlying subtype149.

Table 1 Chemotherapy for the treatment of sRCCs
Full size table

Overall, the use of cytotoxic chemotherapy in sRCCs has yielded poor outcomes. Overall survival has never reached >9 months and progression of disease has occurred within<4 months in 17.4–35% of patients (Table 1). Moreover, phase II sRCC chemotherapy trials resulted in PFS of 2.2–3.5 months, and OS of 3.9–8.8 months147,149. Overall, cytotoxic agents have failed to prove an effective therapy option for patients with sRCCs.

Targeted therapy

During the era of targeted therapies (2006–2015) (Table 2) the number of studies exploring the effectiveness of these interventions in sRCC increased. One study examined the efficacy of the VEGF-inhibitor sorafenib in 15 patients who had progressed on combination doxorubicin and gemcitabine151. Before receiving sorafenib, no responses had been observed on combination chemotherapy and median time to progression was 6.6 months. Patients who received subsequent sorafenib had a mean time to progression of 10.9 months151. In a retrospective study assessing 43 patients with sRCCs treated with VEGF-targeted agents (sunitinib, sorafenib, or bevacizumab), a 19% partial response rate and a median PFS and OS of 5.3 months and 11.8 months, respectively, was observed143. Six patients with ccRCC histology and <20% sarcomatoid dedifferentiation in their tumour achieved partial responses143. In a larger retrospective series, VEGF inhibitor therapy (sunitinib, sorafenib, axitinib, pazopanib, tivozanib or bevacizumab) in 230 patients with metastatic sRCCs compared with 2,056 patients with non-sarcomatoid RCCs was explored152. Over 93% of patients received anti-VEGF agents as first-line therapy and objective responses were slightly less frequent in sRCCs than in non-sarcomatoid RCC (20% vs 26%). Median PFS and OS were 4.5 months and 10.4 months in sRCCs and 7.8 months and 22.5 months in non-sarcomatoid RCCs, respectively152. Another retrospective analysis of 23 patients with sarcomatoid ccRCCs treated with mTOR inhibitors (temsirolimus or everolimus) reported a 13% partial response rate and 30% stable disease rate144. Median PFS was 3.5 months and median OS was 8.2 months. None of the patients experienced a complete response and overall the authors deemed patient response to therapy to be poor144. A separate study examined survival in patients with sRCCs who underwent nephrectomy and were treated with systemic therapy in the cytokine era (1987–2005) and the targeted therapy era (2006–2015)19. The cytokine era cohort, which assessed IL-2, IFNα and chemotherapeutic systemic agents, comprised 122 patients and the targeted therapy era cohort, which assessed sunitinib, sorafenib, bevacizumab, axitinib, pazopanib, erlotinib, nivolumab, ipilimumab, temsirolimus or everolimus systemic agents, comprised 77 patients. The authors noted a 12-month OS benefit in patients with sRCC treated in the targeted therapy era (P = 0.011). However, this improvement in OS was not durable and disappeared at 3–5 years to become indistinguishable from that of patients treated in the cytokine era. Notably, the OS improvement occurred in patients at intermediate risk but not in patients with poor-risk disease19. Rationale for this finding is likely linked to the aggressive and treatment-resistant nature of sRCCs and perhaps further compounded by the added poor prognosis of poor-risk features. Indeed, at the time of this study, poor-risk patients had not experienced an improvement in survival in ~20 years, a finding in agreement with similar reports14,19. The authors concluded that survival in patients with sRCCs was not substantially augmented by the addition of either IL-2 or IFNα, or targeted therapy19. A similar study reviewed outcomes after systemic therapy in 63 patients with metastatic sRCCs, stratifying by first-line agent: anti-angiogenesis-targeted therapy (sunitinib, sorafenib, temsirolimus or everolimus; n = 34), cytokine therapy (IFNα, IL-2, or combination; n = 20) and chemotherapy (geldanamycin, cetuximab, bortezomib or gemcitabine; n = 9)153. Of 63 patients, the retrospective analysis found that 5 patients had an objective response (1 cytokine, 4 sunitinib), and the overall cohort median PFS and OS were 3 months and 10 months, respectively. Median PFS was 4.4 months in sunitinib-treated patients versus 2 months for all other patients (P = 0.03)153.

Table 2 Targeted therapy for treatment of sRCC
Full size table

Targeted therapy studies have reported overall partial response rates of 11.1–19%, with progression of disease observed in 33–57% of patients within 1 year (Table 2). Rapid recurrences and poor response rates have curtailed enthusiasm for the use of targeted therapies in sRCCs, as studies have shown little to modest improvement in sRCC survival outcomes.

Combination therapy

The effectiveness of combination chemotherapy plus targeted therapy has been explored in clinical trials in patients with sRCCs154,155 (Table 3). A phase II trial evaluated combination gemcitabine plus sunitinib in sRCCs and poor-risk RCCs154. In 39 patients with sRCCs, objective response rate (ORR) was 26%, with a median time to progression of 5 months and a median OS of 10 months. Patients whose tumours had >10% sarcomatoid dedifferentiation had greater clinical benefit than patients with ≤10% sarcomatoid dedifferentiation (P = 0.04). A retrospective study analysed combination gemcitabine, capecitabine and bevacizumab response in 10 patients with metastatic RCC156. PFS and OS in patients with sRCCs was 3.9 months and 9 months, respectively, compared with 6.1 months and 10.9 months in patients with non-sarcomatoid RCCs156. In addition, a phase II trial of capecitabine, gemcitabine and bevacizumab in 34 patients with metastatic sRCCs was conducted to explore the use of combination chemotherapy with targeted therapy agents155. Overall objective response was 20% (5 partial, 1 complete), and the disease control rate was 73%. Median time to treatment failure was 4.2 months, median PFS was 5.5 months and median OS was 12 months155. Preliminary results of the phase II ECOG 1808 trial of sunitinib with or without gemcitabine in patients with sRCCs were presented at the 2016 American Society of Clinical Oncology annual meeting157. Of 87 patients, 47 were randomly assigned to sunitinib plus gemcitabine and 40 to sunitinib only. Response rates for assessable patients were 20% (7/35) for the sunitinib plus gemcitabine arm and 11.1% (4/36) for the sunitinib arm. Median PFS and OS were 5.29 months and 9.43 months for sunitinib plus gemcitabine, and 2.99 months and 7.59 months for sunitinib157.

Table 3 Combination chemotherapy with targeted-therapy or immunotherapy for treatment of sRCC
Full size table

Combination therapy for sRCCs has yielded objective, partial and complete response rates ranging from 0% to 26%, 12.8% to 23%, and 2.6% to 3.3%, respectively. When quantitatively assessing the available data on combination therapy, response rates in sRCCs remain low even when combining two effective conventional RCC therapies.

High-dose IL-2

The rapid development of immunotherapies has added a new dimension of potential systemic therapy options for patients with sRCCs. A 2017 study reported outcomes of 21 patients with metastatic sRCCs who received high-dose IL-2 following nephrectomy158. Overall response rate was 10%, and 5% of patients experienced complete response. Localized disease was associated with improved responses to high-dose IL-2, and median PFS and OS were 7.9 months and 30.5 months, respectively158. The authors conclude that high-dose IL-2 offers modest overall response rates similar to responses seen at the time in other immunotherapy studies, which had complete response ranges of 6–9.3% and partial responses of 8.3–18%35,159,160,161. However, since this publication, high-dose IL-2 therapy in RCCs has fallen out of favour and has been largely replaced with newer immune checkpoint blockade agents owing to the limited overall response rates, high rate of treatment-related deaths (~4-6%)161 and poor tolerability162,163 in the form of high rates of grade 3 and grade 4 adverse events164,165 associated with IL-2 therapy.

Immune checkpoint blockade therapy

Immune checkpoint inhibitors have seen the most progress compared with other systemic therapies for sRCCs. Tumours with either constitutive or high levels of adaptive PDL1 expression patterns and high levels of tumour-infiltrating lymphocytes (TILs) are the most likely to respond to immune checkpoint blockade166. Previous studies indicate that sRCCs have higher expression of PDL1 on tumour cells and PD1 on TILs than non-sarcomatoid RCCs122,167. Furthermore, concomitant expression is reported in up to 50% of sRCCs of any subtype, compared with 3% of non-sarcomatoid RCCs122, suggesting that the PD1–PDL1 axis has an active role in sRCC. Additionally, in a study of 118 sRCCs, ~40% of sarcomatoid components harboured both PDL1 expression and TILs, compared with 8% of the corresponding epithelial components and only 1% of the non-sarcomatoid RCC control group100. These results suggest a pattern of immune resistance in sRCCs, making investigations into anti-PDL1 and PD1 agents a sensible choice for therapeutic exploitation.

Taken together, four case reports have documented a rapid response to PD1 inhibition by nivolumab in patients with sRCCs. A 2017 report involved a patient with clinically localized chromophobe sRCC initially treated with nephrectomy who, 7 years later, developed local disease recurrence168. The patient was treated with adjuvant combination doxorubicin and ifosfamide with minimal effect and underwent palliative debulking surgery followed by six cycles of second-line nivolumab therapy, which resulted in a partial response168. A 2015 report described a patient with a metastatic papillary sRCC with rapid disease progression on carboplatin–gemcitabine, sunitinib and gemcitabine given in sequence169. At 3 weeks following a single dose of nivolumab, clinical, biological and radiological responses were observed169. Similarly, in a 2018 report, a complete response to the PD1 checkpoint inhibitor nivolumab was reported in a 67-year-old man with clear cell sRCC who initially experienced disease progression after a 4-month treatment of combination sunitinib plus gemcitabine170. At 1 month after starting nivolumab the patient experienced partial resolution of metastatic sites and at 6 months of treatment PET–CT showed complete resolution of all known metastatic disease. Nivolumab was continued for 2 years and PET–CT at that time showed no evidence of recurrence170. In a 2019 report, 2 patients with sRCCs with documented amplifications of PDL1 at 9p24.1 were reported to have a dramatic response to immunotherapy99. The first patient received pembrolizumab as a second-line therapy. Pretreatment imaging demonstrated diffuse metastatic disease, and on-treatment imaging showed stable disease at 16 months. The second patient received atezolizumab as a first-line therapy. Imaging prior to immunotherapy documented metastases to the lungs, liver, bone, thoracic adenopathy and nephrectomy bed. At 3 months post-treatment, imaging showed a decrease in the size and number of pulmonary metastases, adenopathy, and nephrectomy bed involvement, with radiological evidence of stable disease at 14 months. A combination of PDL1 amplification, coupled with JAK2-activation-dependent upregulation of PDL1 and adaptive and/or induced expression leads to extremely high levels of PDL1 expression in tumours. These patients might, therefore, show an enhanced response to immune checkpoint blockade therapy.

Prospective immune checkpoint blockade studies and sRCCs

Five prospective studies have investigated different forms of immune checkpoint blockade therapy in sRCCs, which continues to show the strongest improvement in outcomes compared with any previous systemic therapy for this disease. Most of the trial results have now been published in abstract form (Table 4), with completed publications anticipated.

Table 4 Prospective data assessing response of sRCC to immune checkpoint blockade
Full size table

The safety and clinical activity of anti-PDL1 atezolizumab in metastatic RCCs were evaluated in a phase Ia trial that included patients with sarcomatoid features171. Patients with WHO/ISUP grade 4 RCC and/or sarcomatoid features (n = 18) were separately assessed for ORR, PFS and OS. ORR was 33% for patients with sarcomatoid dedifferentiation and 25% for patients with grade 4 disease. Median PFS and OS for patients with grade 4 RCC and/or sarcomatoid features was 4.2 months and 26.2 months, respectively171. Median PFS and OS was 5.6 months and 28.9 months for the entire cohort, respectively171.

A post hoc exploratory analysis of the CheckMate214 trial172, which originally assessed the anti-CTLA4 agent ipilimumab plus the anti-PDL1 agent nivolumab compared with single-agent anti-VEGF sunitinib in patients with intermediate-risk or poor-risk RCC, identified patients with sarcomatoid histology and measured efficacy and safety outcomes in this subpopulation173. Of 139 patients with sRCC (74 receiving ipilimumab plus nivolumab and 65 receiving sunitinib), the median follow-up duration was 47.7 months and ORR after ipilimumab plus nivolumab or sunitinib administration was 60.8% vs 23.1% compared with 41.9% vs 29.4% in the remaining participants, respectively. Overall PDL1 expression was higher in patients with sRCCs than in other patients (51% vs 27.5%), and patients with sRCCs with ≥1% PDL1 expression treated with ipilimumab plus nivolumab had a longer median OS than those treated with sunitinib173. Overall, patients with sRCCs treated with ipilimumab plus nivolumab had improved median PFS (26.5 vs 5.1 months, P = 0.0093), complete response rates (18.9% vs 3.1%), partial response rates (41.9% vs 20.0%) and median OS (not reached vs 14.2 months, P = 0.0155) compared with those treated with sunitinib173,174. The 42-month overall survival probability was 50.1% (95% CI 37.9–61.2) with ipilimumab plus nivolumab versus 22.6% (95% CI 13.3–33.4) with sunitinib. Based on these results, the Society for Immunotherapy of Cancer recommended ipilimumab plus nivolumab combination immunotherapy as a first-line treatment option for patients with sRCCs175.

The KEYNOTE-426 trial176 assessed anti-PD1 pembrolizumab plus anti-VEGF axitinib as a first-line therapy for patients with metastatic RCCs. In 2019, abstracted outcomes for a subgroup of 105 patients with sRCCs were presented177. The PDL1 combined positive score ≥1 (calculated by dividing the number of PDL1-positive tumour cells, macrophages and lymphocytes by the total number of cells and multiplying the result by 100) for patients with sRCCs was 74.5–79.6%. Of 105 patients with sRCCs, 51 were randomly assigned to receive pembrolizumab plus axitinib and 54 to sunitinib. ORR was 58.8% versus 31.5% and the complete response rate was 11.8% versus 0% in the pembrolizumab plus axitinib versus sunitinib arms, respectively. Percentage of tumour shrinkage was greater for pembrolizumab plus axitinib than for sunitinib: 80% vs 50% of patients experienced a ≥30% decrease change from baseline in target lesions according to RECIST v1.1 (ref.178) criteria, respectively. Furthermore, PFS at 12 months was 57% in the pembrolizumab plus axitinib arm compared with 26% in the sunitinib arm, respectively. Overall, the PFS HR was 0.54 (95% CI 0.29–1.00) and the OS HR was 0.58 (95% CI 0.21–1.59) for pembrolizumab plus axitinib. The authors concluded that pembrolizumab plus axitinib has substantial activity in sRCCs177.

Abstracted post hoc data regarding sRCC outcomes from the phase III JAVELIN-Renal 101 trial179 were presented at the 2019 annual European Society of Medical Oncology in Barcelona, Spain. This trial originally assessed the anti-PDL1 checkpoint inhibitor avelumab plus axitinib versus sunitinib in previously untreated patients with advanced RCCs. The primary end points were PFS and OS in patients with PDL1-positive tumours. Of 886 patients enrolled in the trial, 442 were assigned to receive avelumab plus axitinib and 444 to receive sunitinib. Among the 560 patients with PDL1-positive tumours, median PFS was 13.8 months with avelumab plus axitinib, compared with 7.2 months with sunitinib (disease progression or death HR 0.61, 95% CI 0.47–0.79; P < 0.001). ORR in patients with PDL1-positive tumours was 55.2% with avelumab plus axitinib and 25.5% with sunitinib. Overall, PFS was longer in the avelumab plus axitinib arm as a first-line agent in PDL1-positive tumours. In the post hoc sRCC analysis179, 108 patients with sRCCs were identified. Of these patients, 47 were randomly assigned avelumab plus axitinib, whereas 61 received sunitinib. The avelumab plus axitinib arm achieved an improvement in ORR (46.8%, 95% CI 32.1–61.9; OR 3.249 95% CI 1.300–8.236) versus sunitinib (21.3%, 95% CI 11.9–33.7), and 2 patients receiving avelumab plus axitinib had a complete response compared with none receiving sunitinib. Additionally, treatment with combination agents resulted in a 3-month improvement in PFS (HR 0.57, 95% CI 0.325–1.003), as well as a 2.4-month improvement in duration of response for patients with sRCC when compared with sunitinib180. Overall, these results suggest that sRCCs may benefit from an immunotherapy plus VEGF-targeted therapy combination. Continued reporting beyond 24 months is highly anticipated and will further add to the growing body of data on the use and efficacy of immune checkpoint blockade in patients with sRCCs.

A pre-specified subgroup analysis181 of patients with sRCC was conducted based on the phase III randomized IMmotion151 trial182, which was designed to assess atezolizumab plus bevacizumab versus sunitinib in previously untreated patients with advanced or metastatic RCC. The original trial results showed prolonged PFS for patients receiving atezolizumab plus bevacizumab within the intent-to-treat population (11.2 months vs 7.7 months, HR 0.74, 95% CI 0.57–0.96, P = 0.0217). The subgroup analysis assessed the effectiveness of atezolizumab plus bevacizumab versus sunitinib in patients with sRCCs. Of 142 patients with sRCC, 68 received atezolizumab plus bevacizumab and 74 received sunitinib. Patients with sRCC who received atezolizumab plus bevacizumab had longer PFS (8.3 months vs 5.3 months, HR 0.52, 95% CI 0.34–0.79) and higher ORR (49% vs 14%) than those who received sunitinib. Median overall survival in the sRCC atezolizumab plus bevacizumab group versus sunitinib group was 21.7 months versus 15.4 months, respectively. Notably, more patients with sRCCs who received atezolizumab plus bevacizumab achieved complete response (10% vs 3%). Results from this subgroup analysis suggest improved clinical efficacy with atezolizumab plus bevacizumab compared with sunitinib for patients with sRCCs, as shown by improved PFS, OS and ORR.

In summary, immune checkpoint inhibitors are a promising therapeutic strategy in the treatment of sRCCs. The available data support a higher expression of PDL1 in sRCC tumour cells and improved outcomes after immune checkpoint blockade. Indeed, immune checkpoint inhibitors have shown increased ORR compared with sunitinib (~50% vs 20%, RR 2.15, P < 0.00001) overall, with ~40% decrease in progression (HR 0.56, P < 0.0001) and overall mortality risk (HR 0.56, P = 0.001)183. As new data become available, the PD1–PDL1 signalling axis will likely be an important therapeutic target to explore in the future.

Ongoing clinical trials

Currently, one clinical trial184 is evaluating systemic therapy applications specifically for patients with sRCC (Supplementary Table 2), although recruitment for this trial is currently suspended because of the COVID-19 pandemic. This single-arm, phase I trial is designed to assess the feasibility and safety of combination avelumab and gemcitabine, for use in patients with metastatic sRCCs. As a phase I clinical trial, the primary outcome for this study is the number of adverse events experienced by patients. Previous reports indicate that sRCCs have higher expression of PDL1 on tumour cells than non-sarcomatoid RCCs122, making avelumab a practical choice for trial investigation. Gemcitabine has previously been paired with a number of systemic agents — either other chemotherapies or targeted therapies, such as gemcitabine–doxorubicin, gemcitabine–sunitinib, and gemcitabine–capecitabine-bevacizumab — with varying and generally limited success148,149,151,154,155,156,169. However, gemcitabine in combination with an anti-PDL1 agent has not previously been explored in sRCCs.

Notably, a currently recruiting phase III trial185 will assess first-line therapy with the anti-PDL1 agent nivolumab combined with ipilimumab. Standard treatment of nivolumab and ipilimumab followed by nivolumab alone will be compared with cabozantinib and nivolumab in patients with untreated metastatic RCCs. Patients with metastatic disease who have not received systemic therapy for RCCs are eligible for inclusion: the primary outcome is overall survival, and the secondary outcomes include PFS, objective response and complete response. Although not specific to sRCC, this trial will include patients with metastatic sRCCs. At the conclusion of this trial, a sub-analysis of patients with metastatic sRCCs would help to confirm or refute current speculation. Specifically, studies have reported sRCCs to have higher expression of PD1 and PDL1 (refs99,122,168,169,174) and a high infiltration of regulatory T (Treg) cells, which constitutively express CTLA4; expression of both is associated with poor prognosis in RCC, indicating an immuno-exhaustive state186,187,188,189. Thus, treatment of patients with sRCCs with a combination of an anti-PDL1 agent and an anti-CTLA4 agent seems potentially promising. No trial to date has completed enrolment and specifically assessed this combination in sRCCs and, therefore, the data generated from this study are highly anticipated to help to address this speculation.

Management strategies in sRCCs

An sRCC is an aggressive variant of an RCC, and most patients present with large primary tumours and synchronous metastases. Unfortunately, regardless of treatment choice, 5-year survival for patients with sRCCs has remained steady at 23.5–33% for the past three decades14,23,26 despite improved surgical techniques and systemic therapies. Currently, no reliable method of identifying sRCC preoperatively exists. Owing to the heterogeneous distribution of sarcomatoid regions within an sRCC tumour, successful identification with biopsy only occurs in ~7.5% of cases19. Indeed, although localized sRCC has worse outcomes than localized non-sarcomatoid RCC, durable responses have been shown following nephrectomy in the localized setting24. Thus, for localized disease, surgical resection remains the standard of care24.

However, the majority of sRCC patients present with advanced or metastatic disease for which the role of nephrectomy is controversial. A 2009 study based on cytokine systemic therapy era data previously suggested that patients with metastatic sRCCs first receive a trial of systemic therapy and are only offered cytoreductive nephrectomy if a response was demonstrated15. A larger 2019 study advocated that cytoreductive nephrectomy in good performance status patients14 is appropriate. Taking an individualized patient-centred approach when deciding whether or not to surgically intervene in sRCC patients is imperative and caution is recommended.

To date, a search for an effective systemic therapy for sRCC is still ongoing. Cytotoxic and targeted therapies have both shown poor outcomes as single agents or in combinations. However, the immunotherapy era has seen a renaissance of new drugs that show encouraging levels of response in sRCCs. Multiple contemporary studies show that sRCCs have increased co-expression of PD1 on tumour-infiltrating lymphocytes and PDL1 on sRCC tumour cells compared with conventional subtypes of RCC99,100,122, which suggests that immune checkpoint blockade therapies might be of increased benefit. Post hoc analyses from clinical trials have shown increased benefit of immunotherapy in combination with targeted therapy agents such as anti-VEGF therapy, and have achieved all-time high objective response rates in advanced or metastatic disease of 55.2–58.8%177,179. Furthermore, anti-CTLA4 systemic agents such as ipilimumab might be beneficial against sRCC tumours in conjunction with anti-PD1 and/or PDL1 therapy190: results from a post hoc analysis of the CheckMate214 trial172 show an objective response of 56.7% vs 19.2%, and a complete response in 18.3% vs 0% of patients after ipilimumab plus nivolumab versus sunitinib, respectively174. Contemporary data have shown increased infiltration of Treg cells, which constitutively express CTLA4 (ref.191), into sRCCs compared with conventional RCCs, providing further evidence of the potential benefits of combination immunotherapy treatment.

The growing number and availability of immune checkpoint inhibitors such as anti-CTLA4, anti-PD1 and anti-PDL1 systemic agents in the setting of compelling data supporting higher expression of PD1 and PDL1 and greater infiltration of Treg cells in sRCCs than in non-sarcomatoid RCCs suggests that this signalling axis could be a critical therapeutic target. Future clinical trials should focus on therapies such as anti-CTLA4, anti-PDl, anti-PDL1 and combinations with targeted therapy agents that have been shown to augment the cytotoxic tumour immune microenvironment99,122,168,169,173,186,187,188,189 with the aim of improving progression and survival outcomes in patients with sRCCs.

Although findings from ongoing and future studies are awaited, current data from immune checkpoint inhibitor trials suggest that treatment for patients with metastatic RCCs and confirmed sRCCs on renal biopsy should begin with immune checkpoint inhibitor combination therapy. Indeed, improved PFS and ORR with combination therapy have been observed in subgroup analyses of sRCC patients within the randomized controlled IMmotion151 trial assessing atezolizumab plus bevacizumab versus sunitinib in advanced or metastatic RCCs181 The use of cytoreductive nephrectomy in metastatic patients with sRCCs should be considered, though with a high threshold for surgical eligibility, as conflicting data on survival benefit are available, with most recent population-based data suggesting a possible benefit in good-risk group patients.

Conclusions

The paucity of data on sRCCs highlights the need for continued research into the biology, diagnostics and effective treatment options for patients with this disease, as responses to conventional therapies have been underwhelming and leave few options for those diagnosed with sRCC. Cytoreductive nephrectomy for patients with sRCC with metastatic disease is controversial, but literature as recent as 2019 (ref.14) advocates for resection at any stage in patients with good performance status. Reports of higher expression of PD1 and PDL1 in sRCCs than in non-sarcomatoid RCCs has generated growing interest in immune checkpoint blockade therapies in combination with other systemic agents, and at least two clinical trials are underway to directly explore this relationship. Ultimately, large collaborative efforts are needed to improve our understanding of sRCCs in order to devise more effective treatments for this lethal disease.