Development of clinically relevant in vivo metastasis models using human bone discs and breast cancer patient-derived xenografts

Abstract

Background: Late-stage breast cancer preferentially metastasises to bone; despite advances in targeted therapies, this condition remains incurable. The lack of clinically relevant models for studying breast cancer metastasis to a human bone microenvironment has stunted the development of effective treatments for this condition. To address this problem, we have developed humanised mouse models in which breast cancer patient-derived xenografts (PDXs) metastasise to human bone implants with low variability and high frequency.

Methods: To model the human bone environment, bone discs from femoral heads of patients undergoing hip replacement surgery were implanted subcutaneously into NOD/SCID mice. For metastasis studies, 7 patient-derived xenograft tumours (PDX: BB3RC32, ER+ PR+ HER2-; BB2RC08, ER+ PR+ ER2-; BB6RC37, ER- PR- HER2- and BB6RC39, ER+ PR+ HER2+), MDA-MB-231-luc2, T47D-luc2 or MCF7-Luc2 cells were injected into the 4th mammary ducts and metastases monitored by luciferase imaging and confirmed on histological sections. Bone integrity, viability and vascularisation were assessed by uCT, calcein uptake and histomorphometry. Expression profiling of genes/proteins during different stages of metastasis were assessed by whole genome Affymetrix array, real-time PCR and immunohistochemistry. Importance of IL-1 was confirmed following anakinra treatment.

Results: Implantation of femoral bone provided a metabolically active, human-specific site for tumour cells to metastasise to. After 4 weeks, bone implants were re-vascularised and demonstrated active bone remodelling (as evidenced by the presence of osteoclasts, osteoblasts and calcein uptake). Restricting bone implants to the use of subchondral bone and introduction of cancer cells via intraductal injection maximised metastasis to human bone implants. MDA-MB-231 cells specifically metastasised to human bone (70% metastases) whereas T47D, MCF7, BB3RC32, BB2RC08, and BB6RC37 cells metastasised to both human bone and mouse bones. Importantly, human bone was the preferred metastatic site especially from ER+ PDX (100% metastasis human bone compared with 20-75% to mouse bone), whereas ER-ve PDX developed metastases in 20% of human and 20% of mouse bone. Breast cancer cells underwent a series of molecular changes as they progressed from primary tumours to bone metastasis including altered expression of IL-1B, IL-1R1, S100A4, CTSK, SPP1 and RANK. Inhibiting IL-1B signalling significantly reduced bone metastasis.

Conclusions: Our reliable and clinically relevant humanised mouse models provide significant advancements in modelling of breast cancer bone metastasis.

Keywords: Bone metastasis; Breast cancer; ER+; ER−; PDX.

Fig. 1

Viability of human bone xenograft following implant into NOD SCID mice. Two 0.5-cm³ pieces of human femoral head were implanted subcutaneously into 8-week-old female NOD SCID mice (n = 5/group). Animals were culled and bone implants removed 0, 7, 14, 21 and 28 days following implantation for analysis of viability (a). Percentage of viable osteocytes was assessed following Giemsa staining (b). Uptake of fluorescently labelled calcein by viable cells 0 and 28 days following implantation of bone was detected by multiphoton microscopy (c). Structure of bone was assessed by uCT and bone integrity shown as % bone volume compared with total tissue volume (BV/TV%) (d). Effects on osteoclast and osteoblast numbers and origin are shown in (e). (i) X60 photomicrographs of osteoclasts and osteoblasts lining human bone 28 days after implantation. (ii) Numbers of osteoclasts and osteoblasts per mm of bone surface, and (iii) origin of these cells as identified by expression of mouse and human CTSK (osteoclasts) and RANKL (osteoblasts/osteocytes) normalised to GAPDH. All graphs show mean ± SEM from 5 mice per time point. Data represents mean ± SEM with statistical significance determined by one-way ANOVA. *p < 0.01, **p < 0.001, ***p < 0.001 compared with 0 h control

Fig. 2

Vascularisation of human bone xenografts following implantation into NOD/SCID mice. Representative 3D reconstructions of bone architecture (a), tomato-conjugated lectin bound to carbohydrates on the luminal surfaces of blood vessels (b) and their merged location within the bone (c). Determination of species origin of endothelial cells is shown in panels d–k, representative immunofluorescent images of bone sections on day 0 (d–g) and day 28 (h–k). Single-channel DAPI (d and h), FITC-conjugated mouse CD31 (e and i), TRITC-conjugated human CD31 (f and j) and merged dual stained images (g and k). Total number of blood vessels present in 10 fields per bone was unchanged over the time course. Presence of single stained human blood vessels decreased significantly from day 14 (l (i)), but a concomitant increase in dual stained vessels was detected (l (ii)). Gene expression of human and mouse VEGF confirmed the immunofluorescent time curve (l (iii)). Data represents mean ± SEM with statistical significance determined by one-way ANOVA. ***p < 0.0001

Fig. 3

Evidence for functional activity of vasculature in bone implants. The spleen of NOD/SCID mice implanted with human bone was dissociated for the detection of human B cells disseminated (n = 5/group). Single-cell suspensions were labelled with antibodies specific for human IgG-PE and CD19-FITC and mouse CD45. FACS scatter plots were gated for all viable cells (R1), then mouse CD45-negative cells (R2). R5 from each spleen sample was plotted as a histogram for human CD19+ IgG+ cells for each time point. Data represents mean ± SEM, statistical significance determined by one-way ANOVA, *p < 0.01, **p < 0.001

Fig. 4

Effects of bone viability on breast cancer metastasis. Eight-week-old female NOD/SCID mice were implanted with femoral bone (mixed, spongy and subchondral) (a), subchondral only (b) or subchondral bone that had been devitalized prior to implantation (c). Four weeks after implantation of bone, all mice received injections of MDA-MB-231 breast cancer cells into mammary ducts 4 and 9; primary tumour growth and metastasis to human bone implants can be seen in a and b. Histogram (d) represents mean ± SEM of percentage of mice that developed metastases in human bone implants from 3 independent experiments each containing 10 mice/group, statistical significance determined by one-way ANOVA. **p < 0.001; ND, not detected

Fig. 5

Effects of intraductal injection of breast cancer PDX on metastasis to human bone implants. Intraductal injection of a sub-set of breast cancer PDX into NOD/SCID mice pre-implanted with human subchondral bone results in metastasis to human/mouse bone. Representative photomicrographs (× 60) of primary tumour sections and human bone sections are shown in (a and b respectively) and uCT images of mouse tibiae are shown in (c). Human bone metastases in mice from which representative H&E images are shown are outlined in orange squares. Metastases in mouse bone from which uCT images are shown are highlighted in green squares, and lesions in mouse tibiae are outlined in yellow

Fig. 6

Gene and protein expression in primary tumours and matched bone metastases from ER+ and ER− PDX. a Fold change in gene expression from RNA isolated from human bone metastatic deposits compared with the corresponding primary PDXs. Data shown are mean ± SEM for 2 independent repeats of 3 replicates. **p < 0.001, ***p < 0.0001. b Photomicrographs of primary tumours and metastases in human bone implants following immunohistochemical detection of S100A4, Fibronectin, Ras IL-1B, IL1R1 and γ Catenin. All data shown are from ER+ PR+ HER2− BB2RC08 and ER− PR− HER2− BB2RC37 PDXs

Fig. 7

IL-1B signalling drives breast cancer metastasis to bone. a Gene expression of IL-1B in human bone implants isolated from NOD SCID mice 4 weeks after intraductal injection of MDA-Td Tomato cells. Data shown are from human bone implants bones isolated from mice in which MDA-Td Tomato cells did not metastasise and from human bone implants with MDA-Td Tomato metastases compared with bone from the same patient isolated from a mouse not injected with breast cancer cells. Data shown are mean ± SEM for n = 5–10 mice per group. Statistical significance determined by one-way ANOVA, *p < 0.01, **p < 0.001, ***p < 0.0001. b Effects of inhibiting IL-1 signalling on metastasis to human bone implants. MDA Td Tomato cells were othrotopically injected into NOD SCID ϒ mice 4 weeks after implantation of human bone. Mice were randomised to receive 1 mg/kg/day anakinra or placebo. Percentage of animals with metastasis in human bone was quantified 8 weeks post tumour cell injection. Data shown are from 7 to 10 mice per group