Assessment of breast cancer primary tumor material in a 3D in vivo model

Abstract

BACKGROUND:

Breast cancer is the most common malignant tumor in women and highly heterogeneous with a variety of different molecular subtypes. The analysis of the individual tumor biology is necessary to develop a specific and individualized treatment plan for every patient. The chick chorioallantoic membrane (CAM) model, a 3D-in-vivo-tumor-model, could potentially provide a methodology that facilitates the gain of additional information regarding the tumor biology as well as the testing of the tumor’s individual sensitivity to different therapies.

OBJECTIVE:

The objective was to establish the grafting of different breast cancer primaries onto the CAM for tumor profiling and the investigation of different parameters.

METHODS:

Breast cancer primary tissue of different patients was grafted onto the CAM. Subsequently, 3D volume and perfusion measurements were performed during the engraftment period. Histological analyses of the tumors were carried out after the engraftment period.

RESULTS:

The grafting of the breast cancer primaries onto the CAM was successful. The tumors remained partially vital and displayed angiogenic development on the CAM.

CONCLUSIONS:

Breast cancer primary material can be grafted onto the CAM and we observed visible and measurable changes of perfusion over time.

Keywords

Breast cancer primary material breast cancer CAM model

1 Introduction

Breast cancer is not only the most frequently diagnosed cancer but also the cancer type with the highest mortality rate among women [1]. Due to the high level of heterogeneity of the disease, it can be divided into multiple clinical and molecular subtypes. Throughout the last decade, new therapeutic strategies and medical regimens have been developed based on these different subtypes, creating and incorporating an increasingly personalized approach towards treatment protocols [2]. In general, unnecessary therapies should be avoided for the prevention of severe side effects and maintenance of a good quality of life for each patient. To do so, it is crucial to accurately identify those patients that benefit from a specific treatment through analysis of the individual tumor biology.

In addition to the tumor biology itself, the tumor microenvironment plays a major role in tumor pathogenesis. Its complexity is subject to intensified laboratory research. 3D in vitro and in vivo models enable the simulation of the tumor microenvironment as well as the architecture and cellular composition of a tumor and its heterogeneity, making these models essential for modern tumor studies [3]. Although 3D in vitro models bridge the gap between 2D models and 3D animal models, the latter frequently provide definitive tests of specific molecules and processes in translational research [4].

The chick chorioallantoic membrane model (CAM) is a well-known 3D model for the study of various tissue types [5, 6]. This is possible due to the chicken embryo being initially immunodeficient in combination with the high vascular density of the CAM. The model was first described in 1911 for the transplantation of chicken sarcoma tumors [7]. The methodology has been improved over the last century for the optimization of the analysis of different tissue types and now presents a well-established method for laboratory tumor research [8, 9]. Different xenografts, tumors and cell suspensions, derived from various tumors have already been grafted onto the CAM [5]. The CAM 3D-in-vivo-model is an easily accessible method, cost efficient and less problematic from an ethical viewpoint than other xenograft models.

Ausprunk et al. have shown that tumor xenografts on the CAM are revascularized by penetration of proliferating CAM vessels into the tumor tissue, recruited by proangiogenic factors that are released by tumor cells [10]. Angiogenesis is a main factor of tumor pathogenesis and facilitates tumor survival, progression, and metastasis, making it a key element regarding tumor aggressiveness. Tumor angiogenesis on the CAM can be measured with different methods that include semiquantitative, quantitative and immunohistochemistry-based applications [11 –16]. Here, we used LASCA (Laser speckle contrast analysis), a semi-quantitative non-invasive laser-based method that enables the real-time mapping of blood perfusion on the CAM [16].

Tumor proliferation and growth are other vital aspects for the evaluation of tumor biology and aggressiveness. Even though the CAM model only provides a short period of time for observation, tumor growth can easily be measured through gain or loss of tumor weight over time as well as 3D volume changes by utilizing a 3D microscope [17].

Furthermore, the CAM model has been previously used for the transplantation of established breast cancer cell lines [18 –25]. Based on an extensive literature research on PubMed, the grafting of primary breast cancer tissue onto the CAM has not yet been performed.

Here, we investigated whether the CAM 3D-in-vivo-model is a suitable method for primary breast cancer tumor grafting and subsequent analysis of tumor biology. As breast cancer is a very heterogenous disease, it is crucial to identify the individual tumor biology as accurately as possible to enable profound clinical decision making. The CAM 3D-in-vivo-model could offer an additional technique to assess real-time tumor growth and angiogenesis, adding supplementary information to the individual tumor biology.

2 Material and methods

2.1 Primary tumor material

Tissue samples of primary breast cancer tumors were provided by the department of gynecology and obstetrics of St. Marien Hospital Amberg. Written consent was obtained from all patients and all experiments were approved by the ethics committee of the University of Regensburg (Nr. 21-2201-101). We used different molecular breast cancer subtypes. Tissue samples included various primary molecular subtypes of breast cancer defined in large part by the presence of hormone receptors (HR), the absence of the HER2-neu receptor and different expression levels of the proliferation marker Ki67.

2.2 Tumors on the chorioallantoic membrane (CAM) model

The CAM model was performed as previously described (see Fig. 1) [16 , 26]. In brief, fertilized chicken eggs were incubated in an egg incubator (Grumbach, Asslar, Germany) at 37.8°C and 63%humidity for 16 days. On day 4 a window was cut into the eggshell and sealed with tape. The eggs were engrafted with primary tumor material on day 7 to 9. Ahead of tumor placement a small area of the chorioallantoic membrane was gently roughened with a cotton swab. All tumor segments were weighed and measured before and after engraftment on the CAM. The eggs were promptly placed in liquid nitrogen after tumor removal.

Fig. 1

Timeline of procedures performed to assess different aspects of tumor proliferation (days: days of development of the chicken embryo).

2.3 3D imaging and volume measurement

Tumor volumes were measured before and after engraftment on the CAM with a Keyence VHX-7000 digital microscope (Keyence Germany, Neu-Isenburg). Due to the mobile objective of this specific microscope, it was possible to scan the entire surface of the tumor and create several two-dimensional images. The relative height data was then used to reconstruct a 3D image based on the incorporated software’s “Depth from Defocus” algorithm [27].

2.4 Angiogenesis measurement via LASCA

The mapping of blood flow of the CAM surrounding the tumor was performed with LASCA technology on day 8 and day 16 using the PeriCAM perfusion speckle imager (PSI) system HR model as previously described [16]. LASCA is a complex, semi-quantitative method to measure the movement of erythrocytes based on the scattering of laser arrays, speckle patterns and the Doppler effect. The perfusion status of each CAM with tumor was measured. The average perfusion rate was calculated by the PimSoft version 1.5 software.

2.5 Histopathology, cell vitality

Paraffin blocks and histological sections were produced for each tumor before and after engraftment onto the CAM. The histological sections were deparaffinized, hydrated in xylene and graded alcohol series and then stained with H&E according to standard protocol. The H&E-stained sections were then analyzed focusing on the vitality of tumor cells

2.6 Graphs and statistical analyses

Graphs and descriptive statistics were created with Graph Pad Prism 8 software.

3 Results

3.1 Growth of primary breast cancer tumors on the CAM

Different primary breast cancer subtypes were investigated. Patients that received neoadjuvant treatment were excluded from the study. As every tumor had an individual biomarker profile with different characteristics, each tumor was analyzed separately (see Table 1).

Table 1
List of breast cancer primaries. All histological assessments were performed via immunohistochemistry

Tumor Nr. Histological subtype Grading ER^* PR^* Her2^** Ki67

1 Invasive-ductal adenocarcinoma G2 12 0 1+ 20%

2 Invasive-ductal adenocarcinoma G2 12 9 0 12%

3 Invasive mucinous adenocarcinoma G2 12 6 1+ 15%

4 Invasive-ductal adenocarcinoma G2 8 9 1+ 20%

Tumor Nr.	Histological subtype	Grading	ER^*	PR^*	Her2^**	Ki67
1	Invasive-ductal adenocarcinoma	G2	12	0	1+	20%
2	Invasive-ductal adenocarcinoma	G2	12	9	0	12%
3	Invasive mucinous adenocarcinoma	G2	12	6	1+	15%
4	Invasive-ductal adenocarcinoma	G2	8	9	1+	20%

ER: estrogen receptor. PR: progesterone receptor. Her2: Her2 receptor. Ki67: Ki67 index. ^*: IRS score out of 12. ^**: Dako Score

3.2 Histological Sections of tumor engraftments onto the CAM

We observed that tumor tissue remained partially vital after 8 days on the CAM. Vital breast cancer glandular ducts could be found in the histological sections of the tumors that were grafted onto the CAM. Circulating chicken erythrocytes within primary breast cancer tissue were observed. Compared to human adult erythrocytes, erythrocytes of the chick embryo characteristically still carry a nucleus. The tumors showed sections of bradytrophic tissue and areas of calcification because of low perfusion and oxygenation. The tumor surrounding CAM depicted its characteristic three-layered structure (see Fig. 2).

Fig. 2

Growth of breast cancer primary tissue on the CAM. (exemplarily represented by tumor nr. 1, see tab. 1) A: Histological section of primary tumor, H&E stained, before engraftment on CAM, magnification 1,2x. ^*: tumor breast cancer ducts of primary tumor before engraftment on CAM, magnification 40x. B: tumor on CAM on day 16 –complete histological section, H&E stained, magnification 4,5x. 1: calcification. 2: bradytrophic tissue. 3: circulating chicken erythrocytes. 4: vital tumor ducts. 5: exemplary section of the CAM. 1–5: magnification 40x.

3.3 3D imaging and volume measurements of tumors before and after engraftment on the CAM

All tumors (nr. 1–4, see. Table 1) were cut into similar pieces. All pieces were weighed and measured with a 3D microscope before and after engraftment on the CAM (see Fig. 3). Sections of tumor nr. 3 showed a loss of weight and volume over time whereas all other tumors showed only little changes in weight and volume. This observation can be explained by the low proliferation rates indicated by low Ki67 indices of all examined tumors (see Table 1).

Fig. 3

3D volume and weight measurements of all tumors (see tab. 1). A–D: volume measurements of tumor nr. 1–4. Each upper row shows one exemplary tumor section on day 8, each lower row shows the same tumor section on day 16. A: tumor nr. 1, all 3D measurements were performed in ovo, upper row tumor in ovo on day 8, lower row on the left tumor ex ovo on day 16, lower row on the right tumor in ovo on day 16. B: tumor nr. 2, all 3D measurements were performed ex ovo, all pictures ex ovo. C: tumor nr. 3, all 3D measurements were performed ex ovo, all pictures ex ovo. D: tumor nr. 4, all 3D measurements were performed ex ovo, all pictures ex ovo. E: comparison of weight and volume of all tumor sections of tumors nr. 1–4 (t1, t2, t3, t4) on day 8 and on day 16 (d8, d16) (volume of t1 measured in ovo on day 8 and 16, volumes t2–t4 all measured ex ovo). n = 3–15. Mean with SD.

3.4 LASCA perfusion measurements of CAMs with grafted primary tumor

The blood perfusion status of CAMs with engrafted primary tumors was measured in accordance with protocol (see Fig. 1) using the PeriCam PSI system HR model with LASCA technology (see Fig. 4). The changes in perfusion were analyzed by placing all measured eggs in the same position to correctly display and compare the angiogenic development. The perfusion status was calculated by the division of perfusion units through the measured perfusion area. The mean perfusion status of all CAMs of tumor nr. 4 on day 16 was remarkably higher than the one on day 8 with 1,73±0,58 PU/mm² on day 8 vs. 3,06±0,62 PU/mm² on day 16 (see Fig. 4).

Fig. 4

LASCA perfusion measurements of CAM with engrafted tumor (exemplarily represented by tumor nr. 4, see tab. 1) color bar illustrates the perfusion scale. A: CAM with engrafted tumor on day 8. B: CAM with engrafted tumor on day 16. C: picture of examined egg under the visible 650 nm laser during LASCA measurement. D: perfusion image of the CAM with engrafted tumor on day 8. E: Perfusion image of CAM with engrafted tumor on day 16. E: The perfusion status is depicted in perfusion units (PU) divided by the perfusion area in mm² on day 8 (n = 5) and 16 (n = 4). Mean with SD.

4 Discussion

For the first time, we demonstrate that the CAM model can be used for the engraftment of primary breast cancer tissue. Until now the CAM model has only been described for the investigation of breast cancer cell lines but not for the analysis of primary breast cancer material [28 –33].

Major advantages of this in vivo model include easy handling, cost efficiency, the possibility of real-time in situ measurements and little ethical challenges [7]. As shown above, tumors on the CAM can be monitored through 3D volume measurements, either in ovo or ex ovo, and LASCA perfusion imaging. Especially the latter showed notable differences in perfusion on the CAM throughout the tumor engraftment period (Fig. 4). Thus, the CAM model provides a good methodology for the analysis of breast cancer angiogenesis and tumor development. Taking it one step further, this model could be used to test the efficacy of anti-angiogenic agents and chemotherapeutic drugs by applying them onto the surface or by injecting them into one of the exposed blood vessels on the CAM [34 –36]. The CAM model also offers the possibility of examining primary tumor material in its own tumor microenvironment which plays an important role in driving tumor aggressiveness.

Compared to other tumor entities with far higher proliferation rates such as sarcomas, primary breast cancer tumors, especially the molecular subtypes used in this study, don’t show clear changes of tumor growth over the short engraftment period on the CAM [26 , 37]. Breast cancer subtypes with higher proliferation rates and a more aggressive molecular subtype usually require neoadjuvant systemic therapy and were excluded from our study. The little changes of weight and volume over time may be one of the limitations for the usage of this model for primary breast cancer tumors. This requires further investigation. In general, the CAM model has some disadvantages when used in cancer research like inconsistent chick viability and the short observation period [38].

One way of assessing the risk of recurrence in breast cancer is the measurement of cellular proliferation which is mostly done by immunohistochemical antibody staining of the Ki67 antigen, a nuclear marker that is expressed in all cell cycle phases except G0. The level of Ki67 determines the categorization of low risk of recurrence (Ki67 low) and high risk of recurrence (Ki67 high). Clinically, the most widely used cut-off for “Ki67 low” is a Ki67 index lower than 15%[2]. Ki67 as a marker of proliferation is used to assess the tumorigenicity of breast cancer. Immunohistochemical staining of histological tumor sections after engraftment on the CAM should be performed in further studies to assess changes of expression of the estrogen, the progesterone and the Her2 receptor and Ki67 compared to the tumor prior to engraftment. Ki67 could indicate changes in the proliferation rate of the tumor, providing additional information regarding tumor volume and weight changes over the engraftment period. The staining of ER, PR and Her2 might show changes of expression levels over time and will help to improve the identification of vital tumor ducts.

For breast cancer patients, the CAM model could be an applicable method for investigating the individual tumor biology. This model might offer an additional technique to assess real-time tumor development and angiogenesis, generating additional information. Since breast cancer is the most frequent malignant tumor and leading cause for cancer associated death among women, it is crucial to accurately identify those patients that benefit from a specific treatment. The CAM model could also be a suitable model for the testing of different drugs and their effects on the individual tumor. To investigate this, further research is necessary.

5 Conclusion

In this study, we demonstrated the engraftment of primary breast cancer tumor on the CAM for the first time. We established an easily performable and accessible method for further investigation of the individual breast cancer tumor biology. We used different methods to monitor tumor growth and perfusion. The 3D-in-vivo-model could allow real-time testing of anti-cancer drugs and their effects on each individual tumor and might present the next step towards personalized therapies.

Footnotes

Acknowledgments

We thank Lucia Denk for her expert technical assistance.

Conflicts of interest

The authors declare that they have no conflict of interest.

References

Sung

, Ferlay

, Siegel

, Laversanne

, Soerjomataram

, Jemal

, Bray

. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries, CA: a cancer journal for clinicians (2021).

Harbeck

, Penault-Llorca

, Cortes

, Gnant

, Houssami

, Poortmans

, Ruddy

, Tsang

, Cardoso

. Breast cancer. Nat Rev Dis Primers. 2019;5:66.

Boucherit

, Gorvel

, Olive

. 3D Tumor Models and Their Use for the Testing of Immunotherapies. Frontiers in Immunology. 2020;11:603640.

Yamada

, Cukierman

. Modeling tissue morphogenesis and cancer in 3D. Cell. 2007;130:601–10.

Ribatti

. The chick embryo chorioallantoic membrane as a model for tumor biology. Experimental Cell Research. 2014;328:314–24.

Krüger-Genge

, Tondera

, Hauser

, Braune

, Görs

, Roch

, Klopfleisch

, Neffe

, Lendlein

, Pietzsch

, Jung

. Immunocompatibility and non-thrombogenicity of gelatin-based hydrogels. Clinical Hemorheology and Microcirculation. 2021;77:335–50.

Ribatti

. The chick embryo chorioallantoic membrane (CAM). A multifaceted experimental model. Mechanisms of Development. 2016;141:70–7.

, Ishihara

, Chin

, Wu

. Establishment of xenografts of urological cancers on chicken chorioallantoic membrane (CAM) to study metastasis. Precision Clinical Medicine. 2019;2:140–51.

DeBord

, Pathak

, Villaneuva

, Liu

H-C

, Harrington

, Yu

, Lewis

, Sikora

. The chick chorioallantoic membrane (CAM) as a versatile patient-derived xenograft (PDX) platform for precision medicine and preclinical research. American Journal of Cancer Research. 2018;8:1642–60.

10.

Ausprunk

, Folkman

. Vascular injury in transplanted tissues. Fine structural changes in tumor, adult, and embryonic blood vessels, Virchows Archiv. B Cell Pathology. 1976;31–44.

11.

Auerbach

, Lewis

, Shinners

, Kubai

, Akhtar

. Angiogenesis assays: a critical overview. Clinical Chemistry. 2003;49:32–40.

12.

Cimpean

, Seclaman

, Ceauşu

, Gaje

, Feflea

, Anghel

, Raica

, Ribatti

. VEGF-A/HGF induce Prox-1 expression in the chick embryo chorioallantoic membrane lymphatic vasculature. Clinical and Experimental Medicine. 2010;10:169–72.

13.

Nguyen

, Shing

, Folkman

. Quantitation of angiogenesis and antiangiogenesis in the chick embryo chorioallantoic membrane. Microvascular Research. 1994;47:31–40.

14.

Nowak-Sliwinska

, Ballini

J-P

, Wagnières

, van den Bergh

. Processing of fluorescence angiograms for the quantification of vascular effects induced by anti-angiogenic agents in the CAM model. Microvascular Research. 2010;79:21–8.

15.

Strick

, Waycaster

, Montani

, Gay

, Adair

. Morphometric measurements of chorioallantoic membrane vascularity: effects of hypoxia and hyperoxia. The American Journal of Physiology. 1991;260:H1385–9.

16.

Pion

, Asam

, Feder

A-L

, Felthaus

, Heidekrueger

, Prantl

, Haerteis

, Aung

. Laser speckle contrast analysis (LASCA) technology for the semiquantitative measurement of angiogenesis in in-ovo-tumor-model. Microvascular Research. 2021;133:104072.

17.

Troebs

, Asam

, Pion

, Prantl

, Aung

, Haerteis

. 3D monitoring of tumor volume in an in vivo model. Clinical Hemorheology and Microcirculation. 2020;76:123–31.

18.

Kim

, Gautam

, Kim

. -A, Kang KW. Inhibition of tumor growth and angiogenesis of tamoxifen-resistant breast cancer cells by ruxolitinib, a selective JAK2 inhibitor. Oncology Letters. 2019;17:3981–9.

19.

Martowicz

, Spizzo

, Gastl

, Untergasser

. Phenotype-dependent effects of EpCAM expression on growth and invasion of human breast cancer cell lines. BMC Cancer. 2012;12:501.

20.

Mira

, Lacalle

, Gómez-Moutón

, Leonardo

, Mañes

. Quantitative determination of tumor cell intravasation in a real-time polymerase chain reaction-based assay. Clinical & Experimental Metastasis. 2002;19:313–8.

21.

Ngan

, Stoletov

, Smith

, Common

, Muller

, Lewis

, Siegel

. LPP is a Src substrate required for invadopodia formation and efficient breast cancer lung metastasis. Nature Communications. 2017;8:15059.

22.

Nienhuis

, Arjaans

, Timmer-Bosscha

, de Vries , Elisabeth

G E

, Schröder

. Human stromal cells are required for an anti-breast cancer effect of zoledronic acid. Oncotarget. 2015;6:24436–47.

23.

Oliveira

, Granja

, Neves

, Reis

, Baltazar

, Silva

, Martins

. Fucoidan from Fucus vesiculosus inhibits new blood vessel formation and breast tumor growth in vivo. Carbohydrate Polymers. 2019;223:115034.

24.

Pathuri

, Thorpe

, Disch

, Bailey-Downs

, Ihnat

, Gali

. Solid phase synthesis and biological evaluation of probestin as an angiogenesis inhibitor. Bioorganic & Medicinal Chemistry Letters. 2013;23:3561–4.

25.

Pinto

, Ribeiro

, Conde

, Carvalho

, Paredes

. The Chick Chorioallantoic Membrane Model: A New In Vivo Tool to Evaluate Breast Cancer Stem Cell Activity. Tool to Evaluate Breast Cancer Stem Cell Activity. International Journal of Molecular Sciences. 2020;22.

26.

Feder

A-L

, Pion

, Troebs

, Lenze

, Prantl

, Htwe

, Phyo

, Haerteis

, Aung

. Extended analysis of intratumoral heterogeneity of primary osteosarcoma tissue using 3D-in-vivo-tumor-model. Clinical Hemorheology and Microcirculation. 2020;76:133–41.

27.

Keyence: 3D imaging, https://www.keyence.com/ss/products/microscope/glossary/cat4/3d_display/, accessed 29 May 2021.

28.

Achkar

, Kader

, Dib

, Junejo

, Al-Bader

, Hayat

, Bhagwat

, Rousset

, Wang

, Viallet

, Suhre

, Halama

. Metabolic Signatures of Tumor Responses to Doxorubicin Elucidated by Metabolic Profiling in Ovo. Metabolites. 2020;10.

29.

Chen

, Rodriguez

, Barnett

, Hashimoto

, Tang

, Yoneda

. Bone sialoprotein promotes tumor cell migration in both in vitro and in vivo models. Connective Tissue Research. 2003;44(Suppl 1):279–84.

30.

Chen

, Zhou

, Yao

, Dai

, Zhou

, Wang

, Zhang

. -Q. Dipalmitoylphosphatidic acid inhibits breast cancer growth by suppressing angiogenesis via inhibition of the CUX1/FGF1/HGF signalling pathway. Journal of Cellular and Molecular Medicine. 2018;22:4760–70.

31.

Comşa

, Popescu

, Avram

, Ceauşu

, Cîmpean

, Raica

. Bevacizumab Modulation of the Interaction Between the MCF-7 Cell Line and the Chick Embryo Chorioallantoic Membrane, In vivo (Athens, Greece)2017;31:199–203.

32.

Gandhi

, Groß

, Holler

, Coggins

, Patil

, Leupold

, Munschauer

, Schenone

, Hartigan

, Allgayer

, Kim

, Diederichs

. The lncRNA lincNMR regulates nucleotide metabolism via a YBX1 - RRM2 axis in cancer. Nature Communications. 2020;11:3214.

33.

Kang

, Regmi

, Kim

, Banskota

, Gautam

, Kim

. -A. Anti-angiogenic activity of macrolactin A and its succinyl derivative is mediated through inhibition of class I PI3K activity and its signaling. Archives of Pharmacal Research. 2015;38:249–60.

34.

Ausprunk

, Knighton

, Folkman

. Vascularization of normal and neoplastic tissues grafted to the chick chorioallantois. Role of Host and Preexisting Graft Blood Vessels, The American Journal of Pathology. 1975;79:597–618.

35.

Brooks

, Clark

, Cheresh

. Requirement of vascular integrin alpha v beta 3 for angiogenesis. Science (New York, N.Y.). 1994;264:569–71.

36.

Presta

, Rusnati

, Belleri

, Morbidelli

, Ziche

, Ribatti

. Purine analogue 6-methylmercaptopurine riboside inhibits early and late phases of the angiogenesis process. Cancer Research. 1999;59:2417–24.

37.

Walewska

, Dolka

, Małek

, Wojtalewicz

, Wojtkowska

, Żbikowski

, Lechowski

, Zabielska-Koczywąs

. Experimental tumor growth of canine osteosarcoma cell line on chick embryo chorioallantoic membrane (in vivo studies). Acta Veterinaria Scandinavica. 2017;59:30.

38.

Kunz

, Schenker Sähr

A, H

, Lehner

, Fellenberg

. Optimization of the chicken chorioallantoic membrane assay as reliable in vivo model for the analysis of osteosarcoma. . PloS One. 2019;14:e0215312.