Patient Characteristics Associated with Growth of Patient-Derived Tumor Implants in Mice (Patient-Derived Xenografts)

Abstract

Background: patient-derived xenografts (PDXs) have defined the field of translational cancer research in recent years, becoming one of the most-used tools in early drug development. The process of establishing cancer models in mice has turned out to be challenging, since little research focuses on evaluating which factors impact engraftment success. We sought to determine the clinical, pathological, or molecular factors which may predict better engraftment rates in PDXs. Methods: between March 2017 and January 2021, tumor samples obtained from patients with primary or metastatic cancer were implanted into athymic nude mice. A full comprehensive evaluation of baseline factors associated with the patients and patients' tumors was performed, with the goal of potentially identifying predictive markers of engraftment. We focused on clinical (patient factors) pathological (patients' tumor samples) and molecular (patients' tumor samples) characteristics, analyzed either by immunohistochemistry (IHC) or next-generation sequencing (NGS), which were associated with the likelihood of final engraftment, as well as with tumor growth rates in xenografts. Results: a total of 585 tumor samples were collected and implanted. Twenty-one failed to engraft, due to lack of malignant cells. Of 564 tumor-positive samples, 187 (33.2%) grew at time of analysis. The study was able to find correlation and predictive value for engraftment for the following: the use of systemic antibiotics by the patient within 2 weeks of sampling (38.1% (72/189) antibiotics- group vs. 30.7% (115/375) no-antibiotics) (p = 0.048), and the administration of systemic steroids to the patients within 2 weeks of sampling (41.5% (34/48) steroids vs. 31.7% (153/329), no-steroids) (p = 0.049). Regarding patient's baseline tests, we found certain markers could help predict final engraftment success: for lactate dehydrogenase (LDH) levels, 34.1% (140/411) of tumors derived from patients with baseline blood LDH levels above the upper limit of normality (ULN) achieved growth, against 30.7% (47/153) with normal LDH (p = 0.047). Histological tumor characteristics, such as grade of differentiation, were also correlated. Grade 1: 25.4% (47/187), grade 2: 34.8% (65/187) and grade 3: 40.1% (75/187) tumors achieved successful growth (p = 0.043), suggesting the higher the grade, the higher the likelihood of success. Similarly, higher ki67 levels were also correlated with better engraftment rates: low (Ki67 < 15%): 8.9% (9/45) achieved growth vs. high (Ki67 ≥ 15%): 31% (35/113) (p: 0.002). Other markers of aggressiveness such as the presence of lymphovascular invasion in tumor sample of origin was also predictive: 42.2% (97/230) with lymphovascular vs. 26.9% (90/334) of samples with no invasion (p = 0.0001). From the molecular standpoint, mismatch-repair-deficient (MMRd) tumors showed better engraftment rates: 62.1% (18/29) achieved growth vs. 40.8% (75/184) of proficient tumors (p = 0.026). A total of 84 PDX were breast models, among which 57.9% (11/19) ER-negative models grew, vs. 15.4% (10/65) of ER-positive models (p = 0.0001), also consonant with ER-negative tumors being more aggressive. BRAFmut cancers are more likely to achieve engraftment during the development of PDX models. Lastly, tumor growth rates during first passages can help establish a cutoff point for the decision-making process during PDX development, since the higher the tumor grades, the higher the likelihood of success. Conclusions: tumors with higher grade and Ki67 protein expression, lymphovascular and/or perineural invasion, with dMMR and are negative for ER expression have a higher probability of achieving growth in the process of PDX development. The use of steroids and/or antibiotics in the patient prior to sampling can also impact the likelihood of success in PDX development. Lastly, establishing a cutoff point for tumor growth rates could guide the decision-making process during PDX development.

Keywords: PDX; cancer; engraftment; mice; models; oncology; patient-derived xenografts; preclinical; prediction; translational; tumor growth.

Figure 1

Process of PDX Development: Samples obtained from patients are carefully selected and submerged in Matrigel. Subcutaneous implants of human tumor fragments are surgically implantes in the lower back of athymic nude mice. The implant is progressively measured during its growth.

Figure 2

Baseline Clinical Characteristics and their association with final tumor growth in PDXs: (A) Analysis showed that 38.1% (72/189) of the samples receiving antibiotics achieved final engraftment vs 30.7% (115/375) in the group that was not treated with antibiotics (p = 0.048). (B): Interestingly, 41.5% (34/48) of samples receiving steroids achieved final engraftment success, against 31.7% (153/329) of the steroids-free samples. These findings were also statistically significant (p = 0.05). (C) Menopausal status was found to be associated with final engraftment success: 34.9% (95/177) of samples obtained from postmenopausal patients achieved final collection, Vs 20,4% (10/39) of the samples derived from premenopausal patients (p = 0.031). (D) High LDH levels were associated with higher engraftment rates. 34.1% (140/411) of samples obtained from patients with baseline LDH levels above the upper limit of normality (ULN) achieved final growth, against 30.7% (47/153) of the samples with normal LDH succeeded (p = 0.047).

Figure 3

Histological characteristics associated with engraftment success. (A) Grade of differentiation reported by pathologist was also assessed. Low grade tumors had an engraftment rate of 25% (47/187), intermediate grade of 34.8% (65/187) and high grade of 40.1% (75/187) suggesting that higher differentiation grades are associated with higher engraftment rates (p = 0.043). (B) Higher ki67 levels were consistent with better engraftment rates: Only 8.9% (9/45) of implants derived from patients with primary tumors with low ki67 levels achieved growth, versus 31% (35/113) of engraftment success achieved in the high ki67 group (Ki67 > 15%) (p = 0.002). (C,D) Interestingly, 42.2% (97/230) with lymphovascular invasion vs 26.9% (90/334) of samples with no invasion achieved final growth (p = 0.0001). Likewise, 41.8% (59/141) of neural invasion-positive samples achieved growth against 30.3% (128/428) (p = 0.008). (E) The expression of hormone receptors (HR) in several tumor types (84 PDX in total) had an impact in total engraftment rate: 57.9% (11/19) of HR negative models grew, vs. 15.4% (10/65) of HR positive models (p = 0.0001).

Figure 4

Tumors coming from samples with MMRd, have a higher engraftment rate: 62.1% (18/29) of MMRd tumors achieved growth vs 40.8% (75/184) of proficient tumors (p = 0.026).

Figure 5

Tumor Growth Rate between models achieving PX1, PX2 and PX3 final success. The models achieving final engraftment success have a higher tumor growth rate during the first and second passage than those that fail: median TGR (mTGR) = 317 (CI 95%: 125–546) vs. mTGR = 14.6 (CI 95% −68.2–274) for those only achieving the first passage, and mTGR 64.7 (CI 95%: −1.70–256) for those only achieving the second passage. These results were statistically significant: p < 0.002 in both cases.

Figure 6

Engraftment success prediction using TGR. Establishing a threshold of 72.8, a sensitivity of 82.4% and a specificity of 5.9% were achieved in predicting engraftment success. This implies that 85.4% of final engraftment successes have a TGR during the first passage superior to 72.8, and that 5.9% of failures present a TGR during PX1 of below 72.8.