Introduction

Leiomyosarcoma (LMS) is one of the most common soft tissue sarcoma (STS) subtypes, accounting for 10–20% of cases [1]. The disease arises from the smooth muscle cell lineage and therefore can affect various anatomical sites. However, LMS commonly develops in the uterus, the abdomen, retroperitoneum and extremities [1]. LMS can also arise from the smooth muscle layer of the vasculature, mainly affecting the inferior vena cava [2]. Due to the distinct clinical features of uterine LMS, the disease is currently classified as uterine LMS and non-uterine LMS. Along with anatomical site heterogeneity, LMS displays a range of histomorphological variants. While conventional LMS displays a spindle cell histology resembling smooth muscle tissue, other variants exhibit epithelioid or myxoid appearance [3]. In addition, dedifferentiated LMS which is characterised by reduced expression or loss of smooth muscle markers has been described and is associated with worse prognosis [3,4,5]. For localised LMS, the standard clinical management relies on wide surgical excision with clear margins. In certain situations, (neo)adjuvant radiation and chemotherapy can be considered. However, there is a high risk of recurrence in LMS [6, 7], and the treatment options for advanced/ metastatic LMS are limited and rely on chemotherapy with doxorubicin in combination with either ifosfamide or dacarbazine as a first-line treatment. Retrospective data suggest that ifosfamide may not be as active in LMS compared to other STS subtypes. Patients who experience disease progression on first-line therapy can be considered for other systemic agents including trabectedin [8] or pazopanib [8, 9]. The combination of gemcitabine and docetaxel, although not recommended as first-line due to increased toxicity [10], is considered for patients with disease progression on doxorubicin-based first-line treatment [8]. More recently, doxorubicin combined with trabectedin in first-line treatment of LMS has shown promising results in the LMS-04 phase 3 trial (NCT02997358), as the combination treatment significantly prolonged progression-free and overall survival of patients compared to doxorubicin monotherapy [11]. However, the clinical benefit of current treatment options is still very limited, particularly for patients with advanced disease [12]. Studies on the molecular biology of LMS have reported key genomic and proteomic features, revealing significant heterogeneity in disease biology and identifying potential new therapeutic avenues. Here, we review the current molecular understanding of key genetic features and inter- and intra-patient heterogeneity in LMS and their implications for clinical management.

Molecular heterogeneity in LMS

Molecular biology of LMS and inter-patient heterogeneity

Common genetic alterations

Both uterine and non-uterine LMS are considered to be sarcomas with complex karyotypes [13, 14]. Large-scale genome sequencing studies have identified TP53, RB1 and PTEN as the most altered genes in LMS [15, 16••, 17, 18, 19•, 20•]. While the exact frequencies of these alterations vary across study cohorts, there is a high level of concordance between paired primary and recurrent LMS samples indicating that these genetic alterations are likely to be early initiating events in disease development [15, 21••]. In addition, loss of function mutations in ATRX are common (16–24% of all LMS reported to have deleterious mutations in ATRX [16••, 20•, 22, 23]) and enriched in uterine LMS cases [16••, 17]. In addition to alterations in specific genes, widespread somatic copy number alterations and whole genome doubling events have been reported in LMS [16••, 21••]. Consistent with most STS subtypes which have an overall low tumour mutational burden (TMB) compared to other solid tumours such as lung cancer or melanoma [24], TMB in the majority of LMS cases is low with a median of < 5 mutations/megabase pairs (Mbp) compared to median of 10 mutations/Mbp in other solid tumours [19•]. However, there is also some TMB heterogeneity within LMS where a subset of uterine LMS tumours (15%) have been reported to harbour increased levels of tumour mutational burden (TMB > 5 mutations/Mbp) [19•].

Molecular heterogeneity and LMS subtypes

Various transcriptomic studies have shown that there are several molecular subtypes within LMS that harbour distinct biology and clinicopathological features [19•, 20•, 21••, 25•, 26•, 27,28,29,30]. These molecular subtypes appear to be conserved in paired primary and relapse patient specimens indicating that they are likely to be an intrinsic feature of the disease [21••, 25•, 28]. Anderson et al. showed that although genomic structural rearrangements including kataegis and chromothripsis varied considerably between primary and metastatic specimens, recurrent tumours shared the transcriptional subtype and > 60% of clonal substitutions and indel mutations with its primary tumours [21••]. However, there is currently no consensus definition of LMS molecular subtypes as the results from the different studies are not always consistent [27]. That said, transcriptomic studies describing molecular subtypes in LMS consistently showed three molecular subtypes [20•, 25•, 26•, 28]. Although there are some discrepancies in the molecular and clinicopathological factors defining these LMS subtypes among different studies, there appear to be some molecular features, namely anatomical site distribution, immune cell composition and smooth muscle differentiation that are associated with the three molecular subtypes. The anatomical site of the disease seems to contribute to the molecular stratification of LMS. This is particularly apparent in uterine LMS as various studies have identified a distinct LMS molecular subtype enriched in uterine cases [21••, 26•, 28]. Additionally, transcriptomics and proteomics studies have consistently identified a subset of dedifferentiated LMS with significantly reduced smooth muscle differentiation markers and myogenic-related signalling [21••, 26•, 31••].

In addition to anatomical site and myogenic markers, studies have reported an LMS subtype that is enriched in immune-related signalling and immune cell infiltration [19•, 20•, 21••, 28]. This immune-enriched LMS molecular subtype has been described to have increased natural killer (NK) and mast cell infiltration [19•], higher macrophage infiltration [21••, 28] and T cells [28]. In other research using immune deconvolution methods on transcriptomic data, Petitprez et al. identified five sarcoma immune clusters (SIC A-E) in two independent cohorts of STS including LMS patients [32•]. The authors showed that although the majority of LMS patients have lower scores for immune signatures (SIC A and B), some LMS cases indeed displayed an “immune hot” phenotype and were grouped with the SIC E cluster characterised by highest immune scores. Further, a study investigating the immune heterogeneity in LMS identified three different immune consensus clusters based on immune deconvolution of transcriptomics data [33••]. The study showed that about 15% of LMS samples form an immune hot subtype. The authors also reported that these LMS immune clusters are associated with SICs defined by Petitprez et al. as well as other immune features including higher CD8 + immune cell infiltration. However, many of the studies described above rely on single institutional series which are retrospective in nature and susceptible to selection bias. In order for molecular subtyping to have clinical utility moving forward, it is important that findings need to be validated in independent cohorts using similar data acquisition and analytical approaches.

Intra-tumour heterogeneity and clinical implications

Intra-tumour heterogeneity (ITH) is an important consideration in tumour progression and treatment response. ITH is a general term that reflects the genetic and phenotypic heterogeneity of cancer cell populations within the tumour as well as the different microenvironmental elements and their spatial and temporal distribution across the tumour over the course of its evolution [34]. Various studies on common cancers have reported an association between genomic and transcriptional ITH and inferior clinical outcomes [35,36,37]. However, the landscape of ITH and its clinical relevance is less well-characterised in STS including LMS. This section summarises the current evidence of ITH in LMS and its prognostic and therapeutic relevance. Anderson et al. undertook multi-regional sampling of LMS tumours and performed genomic and transcriptomic profiling to investigate evolutionary patterns in LMS [21••]. The authors used phylogenetic reconstruction of multiple regions taken from the same tumour and showed that later widespread chromosomal rearrangements and kataegis events resulted in distinct tumour subpopulations in tumour regions that are only a few centimetres apart. In addition, bulk sequencing of paired primary and metastatic regions showed early origins of metastasis and seeding of metastatic clones 10–30 years before diagnosis [21••], consistent with previous reports from more common cancer types [38, 39].

In addition to genetic heterogeneity of tumour cells, stromal components including distribution of tumour infiltrating immune cells also contribute to ITH. The immune cell infiltrate has been of particular interest in sarcomas as this is a potential predictive indicator for response to immunotherapeutic agents [40]. However, the spatial distribution of immune cells has been reported to display considerable heterogeneity. In a study investigating the use of tumour microarrays in LMS, Lee et al. showed heterogeneous distribution of tumour infiltrating lymphocytes across tissue microarrays from the same tumour sample in some LMS cases [41]. A study by Feng et al. using different immune deconvolution methods and multiple LMS tumour regions showed distinct immune signatures of samples taken from the same tumour. Some LMS patient samples displayed both “immune hot” and “immune cold” phenotype when multiple regions were profiled [33••]. Other studies using multiplex immunohistochemistry also showed similar findings for different immune cell populations. For example, Manzoni et al. showed heterogeneity in spatial distribution of tumour infiltrating lymphocytes as well as myeloid cells including macrophages in uterine LMS cases [42]. However, more work is needed to further understand the landscape of ITH and evolutionary patterns in LMS as intra-tumour genetic diversification and heterogeneity in spatial distribution of immune cell populations have important clinical implications in terms of biopsy sampling and the implementation of personalised medicine.

Personalised medicine avenues in LMS

The use of a personalised treatment strategy can help improve clinical outcomes for cancer patients [43]. The application of patient stratification based on molecular subclassification has shown prognostic and predictive clinical value in more common cancer types such as lung and breast cancer [44]. Developing a similar personalised treatment paradigm for LMS is attractive, and over the past decade, several distinct biomarker-matched molecular vulnerabilities have been identified. Here, we discuss some of the promising therapeutic avenues available for LMS patients based on the different molecular features of the disease. Key response predictive biomarkers described in LMS are summarised in Table 1.

Table 1 Putative predictive biomarkers in leiomyosarcoma
Full size table

Tyrosine kinase inhibitors

Pazopanib is a multitarget tyrosine kinase inhibitor (TKI) and is thought to mediate its anticancer effect through the inhibition of a range of TKI involved in angiogenesis and oncogenic signalling [58]. Pazopanib has been approved by the Food and Drug Administration (FDA) for selected advanced STS patients including LMS following results from the multicentre phase III PALLETTE study which showed improved progression-free survival in advanced non-adipocytic STS patients treated with pazopanib compared to placebo control [9]. There was, however, no improvement in overall survival, and responses vary considerably among patients, indicating the need for a predictive biomarker to help select patients for pazopanib treatment. One example of such biomarker development was undertaken by Heilig et al. who defined a pazopanib efficacy predictor (PEP) score using genomics and transcriptomics profiling on tumour tissue samples prior to pazopanib treatment [45••]. The PEP score was developed based on the mRNA expression of three tyrosine kinase genes (NTRK3, IGF1R and KDR) which was significantly associated with progression-free survival in a training dataset (n = 62) as well as a validation cohort (n = 43). Furthermore, the score was not associated with clinical outcome in pazopanib-naïve comparison cohorts suggesting its predictive, rather than prognostic value. Other efforts investigating biomarkers of pazopanib response in STS have shown that mutations in TP53, PD-L1 expression, PDGFRA expression, and FGFR1 expression also significantly correlate with progression-free survival outcome in pazopanib-treated patients [46, 47, 59, 60]. However, many of these findings are limited to small cohort sizes and often lack external validation.

Targeting homologous recombination deficiency in LMS

Genomics and transcriptomics studies using LMS patient cohorts have demonstrated frequent alterations in crucial components of homologous recombination repair (HRR) of DNA double-strand breaks. These include deleterious single base substitutions or genomic alterations affecting key genes in this pathway such as BRCA1, BRCA2, RAD51, ATM, CHECK1, CHECK2, XRCC1, XRCC3, PTEN and FANCA1 and FANCA2 [17, 20•, 21••, 61], suggesting homologous recombination deficiency (HRD) and a “BRCAness” phenotype. HRD is associated with increased sensitivity to DNA double-strand break-inducing agents such as poly (ADP-ribose) polymerase (PARP) inhibitors [62]. Olaparib, an FDA-approved PARP inhibitor for some ovarian, breast, pancreatic and prostate cancers, has been assessed in various clinical studies including LMS patients, particularly in the uterine LMS setting [50•, 51, 52•, 63]. Recent clinical trials have utilised the combination of olaparib together with other chemotherapeutic agents [48•, 55, 64] to further impair the ability of cancer cells to repair DNA damage and ultimately lead to apoptosis [65]. Ingham et al. assessed the combination of olaparib and temozolomide in a phase 2 clinical trial comprised of 22 advanced uterine LMS patients (NCT03880019). The study showed an overall objective response rate of 27% [55]. However, the authors reported considerable myelosuppression leading to dose reduction and toxicity. On the other hand, a phase 1b clinical trial (TOMAS) investigated the combination of olaparib with trabectedin, which binds to minor groove of DNA causing single- and double-strand breaks [48•]. The trial included 55 bone and soft tissue sarcoma patients and showed that the combination was safe and tolerable. The phase 2 multicentre TOMAS2 study compared the combination of olaparib and trabectedin to single agent trabectedin (NCT03838744) [64] and reported potential benefit of the combination treatment with 20% of patients in the combination arm showing durable response for over a year.

Due to heterogeneity in responses to PARP inhibition, there is a need to identify response-predictive biomarkers to enhance patient stratification. Increased PARP1 basal expression was associated with improved response to PARP inhibition in the TOMAS trial [48•]. In a preclinical study that preceded TOMAS, a synergistic effect of combining trabectedin and olaparib in sarcoma cell lines and mouse models was associated with PARP1 basal expression. Further evaluation using gene silencing and overexpression experiments confirmed a functional relevance of PARP1 expression in predicting treatment response [49•]. Additionally, prolonged clinical responses to olaparib have been reported in advanced LMS patients with BRCA1/2 mutations [50•, 51, 52•]. However, the correlation between BRCA1/2 mutational status and response is unclear as some studies reported no correlation with outcome [48•]. Other putative biomarkers to predict response to PARP inhibition include RAD51. In the clinical trial conducted by Ingham et al. (NCT03880019) discussed above, the absence of RAD51 foci in patient samples was assessed as a biomarker for HRD. Patients with absent RAD51 foci had prolonged median progression-free survival compared to homologous recombination proficient patients [55]. Other studies assessed the use of HRD scores and polygenic mutational signatures as a putative predictive biomarker for PARP inhibitor response [53, 54]. However, it should be noted that all these biomarkers need to be further validated in independent cohorts and assessed prospectively to better understand their predictive relevance.

Alternative lengthening of telomeres (ALT)

Alternative lengthening of telomeres (ALT) is a mechanism utilised by ~ 10% of cancers to maintain telomere length and therefore achieve replicative immortality in a telomerase-independent fashion [66]. Using different methods to assess ALT status, various studies have reported a high proportion of ALT-positive LMS tumours [20•, 67,68,69]. In a study by Chudasama et al., DNA C-circles, an ALT-specific biomarker, were detected in 78% of LMS patients (n = 49) [20•]. ALT is commonly associated with ATRX alterations which have been associated with shorter progression-free survival in uterine LMS patients [61]. However, the high frequency of ALT in LMS cannot be explained by ATRX alterations alone as some ALT-positive tumours—assessed by DNA C-circles—did not have ATRX alterations, and alterations of other telomere maintenance genes including RBL2 and SP100 showed significant association with positive ALT status [20•]. Targeting the molecular players utilised in ALT may be a promising therapeutic strategy in LMS. For example, targeting the ATR kinase is thought to lead to synthetic lethality by inducing DNA double-strand breaks at telomeres. Cells with ATRX mutations demonstrated increased sensitivity to ATR inhibition [56]. A study using a panel of soft tissue sarcoma cell lines including 3 LMS cell lines showed that targeting ATR with the ATR inhibitor VE-822 resulted in a synergistic effect when combined with gemcitabine [70]. However, this effect was shown to be ALT independent, and thus, more studies and future clinical trials are needed to better evaluate the therapeutic vulnerabilities of ALT-positive tumours.

Targeting the LMS immune microenvironment

Results from clinical trials and retrospective cohort studies assessing the use of immune checkpoint inhibitors (ICIs) in LMS have thus far been disappointing, with no to modest efficacy observed in patients. This remains true when using multiple ICIs or ICIs in combination with chemotherapy or targeted therapies (immunotherapy trials in LMS reviewed in ref [71]). Recent studies exploring the immune biology of LMS have identified certain immune features that may help select patients for immunotherapeutic strategies [32•, 33••, 57••]. For example, in the study by Petitprez et al. discussed earlier, the authors evaluated specimens from the SARC028 clinical trial and demonstrated that cases with a SIC E signature had an improved objective response rate (ORR) to pembrolizumab compared to patients with the other SIC signatures [32•]. In addition, using fluorescent multiplexed immunohistochemistry, the same study showed enrichment in tertiary lymphoid structures (TLS) in SIC E patients as 82% of these patients were identified to have one or more TLS. Although the LMS cases assessed in this study cohort were very limited (n = 6) and none were SIC E tumours, the study nonetheless demonstrates the potential utility of TLS as an immune biomarker of SIC status and response to pembrolizumab in STS. These results are also consistent with analysis from the PEMBROSARC trial which showed the presence of TLS as a potential predictor of response to pembrolizumab [57••]. The study reported a 6-month non-progression rate of 40% in TLS-positive patients (n = 30 including 4 LMS cases), compared to 4.9% in TLS-negative patients in the TLS-unselected cohort (n = 41 including 13 LMS cases).

Programmed death-ligand 1 (PD-L1) is an immune checkpoint molecule expressed on a range of normal, tumour and immune cells. PD-L1 expression has been associated with poor prognosis in STS including LMS [72], but its role as a response predictive biomarker to anti-PD-1 and anti-PD-L1 ICI in LMS is unclear. PD-L1 expression has been shown to be associated with response to pembrolizumab in some STS histological subtypes [73], but the predictive value of PD-L1 has not been robustly evaluated in STS, and more research in this area is required.

Novel immune targets: targeting the macrophages

Most clinical trials assessing the efficacy of immunotherapy in STS have focused on targeting lymphocyte-based immune checkpoint inhibitors, mainly targeting programmed cell death 1 (PD-1) or its ligand (PD-L1) and cytotoxic T-lymphocyte associated protein 4 (CTLA-4) molecules. On the other hand, studies have shown increased macrophage infiltration, particularly CD163 + macrophages in the immune microenvironment of STS and in particular LMS [74, 75]. Consistent with the immunosuppressive nature of CD163 + macrophages, some reports have shown an association between increased CD163 + infiltration in LMS and worse prognosis [74] which is in line with studies in other cancers [76, 77]. Thus, macrophage-directed therapeutics may form a promising strategy in the treatment of LMS. Targeting the immune checkpoint protein CD47 and its receptor SIRPa has shown exciting results in a range of solid and haematological malignancies [78]. In LMS cell lines co-cultured with peripheral blood mononuclear cells (PBMC)-derived macrophages, treatment with anti-CD47 monoclonal antibodies resulted in increased phagocytic capacity [79]. Results from a phase 1/2 clinical trial investigating the combination of doxorubicin together with the recombinant protein TTI-621 which acts as a decoy receptor for SIRPa are being evaluated in metastatic and high-grade LMS (NCT04996004) [80].

In addition to CD47, recent studies investigated more novel macrophage-directed therapeutic strategies including the CD40/CD40L as well as CSF1/CSF1R axes. CD40 is a surface molecule expressed by macrophages, and its activation leads to enhanced antigen presentation and indirect activation of T cells. A phase II clinical trial assessing the safety and efficacy of targeting CD40 with the CD40 agonist APX005M in advanced STS patients including LMS, is under evaluation (NCT03719430) [81]. CSF1R is predominantly expressed by monocytes including macrophages, and macrophages with active CSF1R-mediated signalling are associated with pro-tumoural phenotype. In other cancer types, the inhibition of this pathway through targeting CSF1R increases macrophage polarisation towards proinflammatory phenotype and therefore increases antitumour activity [82]. CSF1R inhibition led to reprogramming of tumour-associated macrophages and boosted antitumour T cell responses in cancers with high macrophage infiltration in pancreatic cancer models [83]. An ongoing phase 1b trial is assessing the safety and efficacy of the combination of CSF1R inhibition and PD-1 inhibition in high-grade sarcomas including 7 LMS cases (NCT04242238) [84]. Results from these trials combined with more in-depth study of the immune landscape of LMS will provide exciting future therapeutic opportunities.

Future perspectives and role of international collaborations

The implementation of personalised medicine strategies in LMS is still in its infancy. More work is needed to evaluate predictive biomarkers prospectively in clinical trials and to understand their functional and mechanistic role in preclinical studies. Due to the rare and heterogeneous nature of LMS, national and international collaborations are needed to address key biological and clinical questions. These include efforts to achieve consensus molecular definitions of LMS molecular subtypes and establish molecular biomarkers of therapeutic relevance [27]. International consortia such as the LMS SPORE (https://www.rogelcancercenter.org/leiomyosarcoma-spore) and the Sarcoma Accelerator Consortium (https://sarcomaaccelerator.org.uk/) can help to facilitate curation and sharing of clinical and molecular datasets as well as to develop new research studies using cutting edge methods such as single-cell sequencing, spatial profiling and liquid biopsies. Working together with patient-partnered initiatives in LMS such as the Leiomyosarcoma Project (https://lmsproject.org/) and the Count Me In initiative provides an added opportunity to address patient-led research priorities which will ultimately advance our biological understanding of this rare and aggressive disease.