Primary osteosarcoma (OS) of the uterus is a distinctly rare and aggressive disease with fewer than 20 cases reported worldwide. We describe a case of primary uterine OS with rapid development of pulmonary and brain metastasis in a 50-year-old woman. Histopathologic examination of the uterine tumor showed atypical spindle cells producing an osteoid matrix with calcification in keeping with OS. Despite initial response to doxorubicin and ifosfamide, the patient succumbed to brain metastases just 8 months from diagnosis. Whole genome sequencing was performed on tumor and blood samples to analyze genetic alterations in this highly aggressive tumor. A pathogenic somatic missense mutation resulting in substitution of glutamate for lysine at position 653 within the protein kinase domain of the platelet-derived growth factor receptor beta (PDGFRB) was found. The PDGF pathway is involved in cell proliferation and angiogenesis, and it has been implicated in malignancy. Crucially, this pathogenic mutation may be amenable to PDGFR tyrosine kinase inhibition, representing a possible treatment approach in this rare sarcoma.
Although osteosarcoma is the most common primary tumor of the bone, the absolute incidence of osteosarcoma is low at two to four per million.[1,2] Extraskeletal osteosarcoma is a rare malignant mesenchymal neoplasm characterized by the production of malignant osteoid, bone, or chondroid material without attachment to the bone or periosteum.[3,4] Although most extraskeletal osteosarcomas originate from soft tissue of the thigh, retroperitoneum, upper extremities, and trunk,[5,6] primary osteosarcomas from the breast, urinary bladder, heart, lung, and uterus have also been reported.[7–11] Since a case of osteosarcoma of the uterus was reported by Stier and Lyman in 1936, only about 20 cases have been described to date.
Given its rarity, little is known about the pathogenetic mechanisms or the optimal management approach of this aggressive disease. Systemic treatment is usually doxorubicin based, as for a soft tissue sarcoma. Therapeutic outcomes are invariably dismal. We report a case of a primary uterine osteosarcoma with pulmonary and brain metastasis in a 50-year-old woman. Using a whole genome sequencing (WGS) approach, we analyzed tumor and blood samples in an effort to understand underlying mechanisms of tumor development and to identify targetable pathways in this disease.
A 50-year-old Chinese female patient was incidentally found to have a 12-cm uterine mass on pelvic ultrasonography (Fig. 1a) at a health screening examination. The patient was otherwise completely asymptomatic. Apart from a history of fibroids detected on imaging 8 years ago, which was not followed up, the patient had no other medical history and no known family history of malignancy. The patient is married with one child.
The patient consulted a gynecologist, who undertook a laparoscopic total hysterectomy and bilateral salpingo-oophorectomy. The mass was thought to be a large “fibroid” and it was morcellated during the operation as part of a minimally invasive approach. Initial histologic examination was reported as “uterine leiomyosarcoma.” The patient was subsequently referred to our cancer center for consultation and was seen 2 months postoperatively.
Findings from a chest radiograph taken at the initial health screening at an external clinic were normal (Fig. 1b). However, a positron emission tomography/computed tomographic (CT) scan done before consultation at our center (not shown), merely 2 months after the normal chest radiograph findings, showed multiple pulmonary metastases with no residual tumor within the pelvis. At our first consultation, a repeated chest radiograph was performed (Fig. 1c). A chest CT scan, subsequently performed (Fig. 1d), also showed multiple lung metastases.
She initially underwent planning for treatment using single-agent therapy with doxorubicin. Following her first cycle of therapy, imaging revealed no objective response. Ifosfamide was added to her treatment regimen to maximize the chance of tumor response. She subsequently received five cycles of doxorubicin in combination with ifosfamide, achieving radiologic partial response (Figs. 1e, f).
Three months following her last cycle of chemotherapy, she developed sudden right-sided weakness. She was admitted to a different hospital where radiologic imaging revealed multiple intracranial hemorrhages secondary to brain metastases (not shown). She underwent an emergency left frontal craniectomy and left frontoparietal evacuation of hematoma. Histologic examination of her brain lesions confirmed metastatic osteosarcoma. While undergoing rehabilitation and planning for whole brain radiation, merely 1 week from her neurosurgery, her neurological function deteriorated precipitously from further intracranial disease progression. She underwent conservative management and received comfort care in a hospice. She passed on 1 month later, just 8 months from her initial diagnosis and hysterectomy.
We arranged for a review of her hysterectomy specimen, which was reported in the external institution. The hysterectomy specimen itself was not available for further examination and sampling. Nevertheless, the hematoxylin-eosin–stained sections and unstained formalin-fixed paraffin-embedded tissue were retrieved for review and further analysis. They showed multiple leiomyomas, as well as a high-grade sarcoma (Fig. 2). The latter was composed of spindled cells arranged in fascicles and a storiform pattern infiltrating between the smooth muscle bundles, centered in the myometrium. The spindle cells showed cytologic atypia and produced lacelike osteoid matrix with calcification (Fig. 2). There was no epithelial or glandular component. Cells were negative for smooth muscle actin (SMA), caldesmon, and desmin, negating a tumor of smooth muscle differentiation. CD10, which may be seen in endometrial stromal tumors,[8,9] was negative. The estrogen and progesterone receptors (ER and PR), assessed to determine the utility of hormone therapy, were negative. Pancytokeratin MNF-116, performed to exclude a sarcomatoid carcinoma and synovial sarcoma,[10,11] was negative. Tumor protein 63 (p63) was also negative. Staining for programmed cell death protein 1 and programmed death-ligand 1 (PD1 and PDL1) stains was performed with a view to consider immunotherapy for this patient, but these were also negative. Altogether, the diagnosis of uterine osteosarcoma was made as based on the presence of malignant cells producing osteoid, the absence of a carcinomatous component, and exclusion of other types of sarcomas, as well as the possibility of a metastatic osteosarcoma from a bone primary.
Somatic and Germline Alterations
Whole genome sequencing was performed on tumor and blood samples. Pathogenic somatic variants found in tumor tissue are summarized in Table 1. A total of 133 nonsynonymous mutations were identified in tumor tissue, including 119 missense and 13 nonsense single nucleotide variants, as well as 1 frameshift deletion (Supplementary Table 1). A pathogenic missense mutation in platelet-derived growth factor receptor beta (PDGFRB), which substituted glutamate for lysine at position 653 within the protein kinase domain, was found. This was seen at a variant depth of 86 of 108 reads. In silico methods of both Sorting Intolerant from Tolerant (SIFT) and PolyPhen-2 predicted this amino acid substitution to be “deleterious” and “damaging,” respectively. Other potentially relevant somatic nonsynonymous mutations include GDF15 p.P140S and CSMD1 p.D1929E (Fig. 3). POLE p.R1570L, SMARCA4 p.R1341S, and CAMTA1 p.A488E were detected at very low variant allele frequencies and their significance is unknown (Table 1). While POLE mutations are known to lead to a high number of mutations in the context of other cancers, POLE mutations yielding hypermutation are associated with recurrent positional mutations at p.P286R and p.V411L, which is distinct from what we found (p.R1570L). Furthermore, the tumor only had a moderate mutational load of ∼2.4 mutations per megabase. Collectively, these suggest that the detected POLE mutation may not be significant to this case. Notably, we did not detect mutations commonly found in conventional osteosarcomas or leiomyosarcomas, including TP53, RB1, and ATRX. We also found that the tumor had a moderate somatic mutational load of ∼2.4 mutations per megabase and is microsatellite stable.
Germline alterations were also assessed through WGS of the peripheral blood. Only variants of uncertain significance were found. There were no pathogenic mutations found, including in the known osteosarcoma predisposition genes TP53, RB, and CHEK2.
We describe a distinctly rare case of a primary uterine osteosarcoma in a 50-year-old female patient with rapid development of lung and subsequent brain metastases for whom WGS of tumor and blood samples were performed.
Uterine osteosarcomas are rare with published references limited to case reports, all describing highly aggressive disease. While our patient was completely asymptomatic when she underwent a hysterectomy for the large “fibroid” detected on a health screen, these tumors usually occur in postmenopausal women who present with abnormal vaginal bleeding and/or lower abdominal pain. Most patients had distant metastasis at the time of diagnosis, with the lung as the most common site. Distant metastasis to the brain, as seen in our patient, has not yet been reported. Despite extensive treatment, the mean survival time has invariably been less than 1 year. Hence, there is a need to gain more insight to develop better therapeutic approaches for this disease.
The proposed criteria for the diagnosis of uterine osteosarcoma include the presence of malignant cells producing osteoid, the absence of a carcinomatous component, and exclusion of other types of sarcomas, as well as the possibility of a metastatic osteosarcoma from a bone primary. The case described in this study has met all requirements. The tumor displayed morphologic features similar to that of osteosarcoma of the bone. No epithelial, carcinomatous component, or other sarcomatous component was present, excluding the possibility of malignant mixed Müllerian tumor or uterine leiomyosarcoma.
In an attempt to gain a better understanding of this disease, as well as to search for any genetic alterations that could be therapeutically tractable, we performed WGS of the tumor and peripheral blood. To date, there has been one published report for which mutation screening was done with a limited gene panel, but no somatic mutations amenable to targeted therapy were found. We present the first reported case of this rare disease with a whole genome analysis of the tumor as well as germline genetic alterations. Our analysis revealed a pathogenic missense mutation of PDGFRB in the patient's tumor.
The PDGF/PDGFR pathway plays an important role in tumor growth and metastasis in at least three different ways: (1) direct autocrine stimulation of tumor cells; (2) paracrine stimulation of tumor stromal cells and promotion of angiogenesis to overcome hypoxia in the tumor microenvironment; and (3) modulation of tumor interstitial fluid pressure to control the influx and efflux of drugs.
Two approaches to pathway inhibition in clinical use include blocking antibodies against the PDGF receptor to prevent ligand binding and receptor activation, such as Olaratumab, a monoclonal antibody against PDGFRa; and tyrosine kinase inhibitors of the PDGFR, including imatinib, sunitinib, sorafenib, pazopanib, regorafenib, lenvatinib, and axitinib, which also target other receptor tyrosine kinases.
While the PDGF pathway is implicated in up to 30% of cancers,[22,23]PDGFRB mutations are uncommon and the gene is altered in just 2.23% of all cancers, with non–small cell lung carcinoma, colorectal adenocarcinoma, melanoma, breast carcinoma, and malignant glioma having the greatest prevalence of alterations. The utility of PDGFRB mutations as a predictive biomarker has not been established in these cancers and is still under investigation.
At present, the only situation where a genetic alteration of PDGFRB is approved by the US Food and Drug Administration as a predictive biomarker is for imatinib in myelodysplastic syndrome/myeloproliferative disease (MDS/MPD). Rearrangement of PDGFRB has been demonstrated by classical cytogenetic analysis in an exceedingly small fraction of patients with MDS/MPD (∼1%), usually cases with a clinical phenotype of chronic myelomonocytic leukemia. These patients harbor a t(5:12)(q31;p12) translocation involving the PDGFRB and ETV6 genes. Fusion of PDGFRB to the ETV6 gene on chromosome 12 generates a constitutively active tyrosine kinase fusion protein, ETV6-PDGFRB, which is sensitive to imatinib.
In dermatofibrosarcoma protuberans (DFSP), an uncommon locally aggressive cutaneous soft tissue sarcoma, PDGFRB is unmutated but is constitutively activated. Most DFSPs have a characteristic chromosomal translocation, t(17;22), which places the ligand PDGFB under the control of the collagen type I alpha 1 (COL1A1) promoter. This results in constitutive activation of the PDGF receptor beta.[26,27] This is therapeutically tractable, as imatinib may induce clinical responses.
To our knowledge, this is the first report of discovery of a potential therapeutic target in primary uterine osteosarcoma. In an ideal scenario, further functional studies could be performed in vitro to investigate the role of this PDGFRB mutation and whether targeting it in uterine osteosarcoma would affect growth kinetics or invasion, though the lack of a cell line or preclinical model of this rare tumor precludes these experiments. This is a limitation shared by most rare cancers, and perhaps one approach by which patients with rare cancers can quickly access potential treatments is through precision medicine clinical trials where patients are assigned to treatments on the basis of molecular characteristics, such as the NCI-MATCH trial.
In summary, we report a case of primary uterine osteosarcoma with clinicopathologic features similar to those of other reports. A pathogenic missense mutation in PDGFRB was found. This could be a therapeutic target and should be tested in the context of a clinical trial.
Next-generation sequencing of sarcomas may be valuable in the context of rare tumors for which pathogenesis and optimal treatment is unknown.
This may aid in our understanding of these rare tumors and potentially identify novel therapeutic targets, or genetic alterations that may be targetable with existing drugs.
The supplemental material is available with the article online.
Our heartfelt thanks to our patient and her family for allowing the publication of this report, which has helped to further our understanding of this rare disease.
Source of Support: None, Conflict of Interest: None.