Clonal Lineage Tracing with Somatic Delivery of Recordable Barcodes Reveals Migration Histories of Metastatic Prostate Cancer

Abstract

The patterns by which primary tumors spread to metastatic sites remain poorly understood. Here, we define patterns of metastatic seeding in prostate cancer using a novel injection-based mouse model-EvoCaP (Evolution in Cancer of the Prostate), featuring aggressive metastatic cancer to bone, liver, lungs, and lymph nodes. To define migration histories between primary and metastatic sites, we used our EvoTraceR pipeline to track distinct tumor clones containing recordable barcodes. We detected widespread intratumoral heterogeneity from the primary tumor in metastatic seeding, with few clonal populations instigating most migration. Metastasis-to-metastasis seeding was uncommon, as most cells remained confined within the tissue. Migration patterns in our model were congruent with human prostate cancer seeding topologies. Our findings support the view of metastatic prostate cancer as a systemic disease driven by waves of aggressive clones expanding their niche, infrequently overcoming constraints that otherwise keep them confined in the primary or metastatic site. Significance: Defining the kinetics of prostate cancer metastasis is critical for developing novel therapeutic strategies. This study uses CRISPR/Cas9-based barcoding technology to accurately define tumor clonal patterns and routes of migration in a novel somatically engineered mouse model (EvoCaP) that recapitulates human prostate cancer using an in-house developed analytical pipeline (EvoTraceR).

Figure 1.. The EvoCaP platform for defining cancer evolution.

A, The BC10 platform is compatible with loxP technology. Lentivirus is injected into prostate of the EvoCaP mice. EvoTraceR is an R package for specific analysis of the barcode. B, BC10 and MG or NMG are driven by an hU6 promoter. Amplicon DNA can be extracted, amplified and sequenced for analysis of edits in the barcode (Amp-Seq). C, Plasmid components used in BC10 plasmid technology. D, Diagram of experimental procedure. Primary MEFs extracted from mice harboring the Pten^loxP/loxP;Trp53^loxP/loxP genotype are infected with lentivirus at day 0. At days 7, 14, and 28, DNA is collected for amplicon analysis. Additional readouts include eGFP fluorescence, FLuc bioluminescence, and protein expression by Western blot. E, KO of Pten/p53 and activation of Cas9-eGFP with Cre expression in virus-infected cells at day 7. NVC = no virus control (negative). F, The number of eGFP⁺ cells increases over time and is consistent between viruses. Individual groups were analyzed at each time point using parametric unpaired t-tests with two-tailed p-values ± SD (95% confidence interval). (Created with BioRender.com).

Figure 2.. EvoCaP mice develop metastatic PCa to relevant secondary sites.

A, Timeline for EvoCaP injections and analysis. Primary tumors and metastases may be tracked longitudinally during the course of live imaging based on BL⁺ signal. B, Kaplan-Meier survival curve displaying rate of death in mice injected with BC10-containing lentivirus that developed metastatic prostate cancer over the 60 week period of monitoring. C, Longitudinal BL⁺ analysis of BC10-injected mice shows advancing tumor progression and metastasis over 60 weeks. The top three panels show BL⁺ progression in EvoCaP-LP cohort. Bottom four panels show BL⁺ progression in EvoCaP-HP cohort. Black line on border indicates that mouse did not survive through 60 weeks. D, Box plot displaying quantitation of BL⁺ radiance in all mice that survived up to a 60 week period. After week 28, animals that did not develop BL⁺ signal were no longer followed and penetrance was determined. Only mice with primary and metastatic eGFP⁺ signal are included in the analysis. MMUS1466 was sacrificed between week 56-60 and final BL⁺ data was included for week 60. Percentages listed under week 28 indicate only those mice with metastases that exhibit BL⁺ signal. Data are displayed in log scale. E, Bioluminescence imaging (BLI) enables post-mortem visualization of primary tumor and metastases and fluorescence enables robust isolation of eGFP⁺ cells using microscopy for downstream analysis. Scaling for BL⁺ signal is identical to that used in 2C. F, Frequency of eGFP⁺ signal post-mortem in analyzed tissues from all mice exhibiting metastasis. FL* in some lymph nodes indicates that fluorescence was not measured but the lack of BL⁺ signal or enlargement makes presence of tumor cells unlikely. The numbers in each green box represent the number of different sites that has at least one metastasis. In MMUS1874, two metastases were isolated in the same bone but counted as one site (FMR). Local invasion into seminal vesicle is also depicted. Fluorescence shows metastasis distribution in different organs (frequency of metastases: bones > liver > lungs > lymph nodes). (Created with BioRender.com)

Figure 3.. Scheme for BC10 design.

A, CFD scores were developed based on assessment of types of mismatches at every position within the guide sequence. We assigned CFD scores based on four mismatched nucleotides and positioned the 10x TS on the barcode in order of decreasing predicted activity (maximum cut TS01: 1.0; minimum cut TS10: 0.1). Cas9 creates heritable marks (insertions and deletions) in target sites with likelihood predicted by CFD score. ASVs are identified based on deletions (rectangles) and insertions (diamonds) in the BC10 and transformed to the Boolean matrix (color: mark or white: absent mark in BC10). B, EvoTraceR comparison to CRISPResso2 and ampliCan. C, ASVs are grouped based on common marks with nucleotide resolution (truncal mutations) and are considered as related clonal populations (CPs). ASVs could be found in different metastatic sites. D, BC10 analysis in the EvoCaP enables descriptions of CPs and routes of metastatic spread. (Created with BioRender.com)

Figure 4.. Barcode analysis reveals high shared editing patterns between primary and metastatic sites.

A, Fluorescence images in prostate (PRL), median and right liver lobes (LVM, LVR), and left rib (RBL) in MMUS1495. B, BC10 edits are characterized by deletions and insertions caused by Cas9, mainly at 1.0/0.9 and 0.5/0.4 target sites as predicted by CFD score. Frequency (y-axis) of deletions (blue) and insertions (red) across the entire amplicon (260 bp, x-axis) in different organs. C, The majority of marks in the BC10 are short (<30 bp deletions and <6 bp insertions) affecting from 1-4x different sites (x-axis: 1-52, light blue and light red bars). By comparison, larger indels are rarely observed (x-axis: >130, darker blue bars). X-axis signifies the number of base pairs within the BC10 regions that a specific edit spans. Edits greater than 26 bp constitute a region larger than one TS. D-E, Eco-statistical measures of heterogeneity; D, Shannon’s index measures alpha-diversity, which quantifies intraclonal heterogeneity. E, Beta-diversity is calculated using Bray-Curtis dissimilarity scores, and is representative on inter-organ heterogeneity. Organs with scores closest to 1 are the least similar to one another, while a score of zero would indicate that the tissues are exactly homogeneous in terms of their edits. F-H, Phylograms are shown for the most information-rich CPs with the highest number of counts. F, CP01, G, CP02, and H, CP03, We depict genealogies of selected CPs composed of their intrinsic ASVs. Phylogenetic analysis (phylogram) performed using Cassiopeia suite with the greedy algorithm. ASVs are expressed as lengths of BC10 deletion (blue) or insertion (red). Individual CPs are shown with descendent subclones (branches and leaves).

Figure 5.. Migration histories of mouse with advanced prostate cancer.

A-F, Reconstructed migration histories of CP01, CP02, and CP03 from MMUS1495. A-C, MACHINA-generated metastatic trajectories for A, CP01, B, CP02, and C, CP03. Circles represent ASV frequency within a particular organ. Intensity of edge color denotes degree of expansion of an ASV within the defined organ. Diamonds represent ancestral states inferred by MACHINA. D-F, Individual transition matrices for D, Transition matrix for all CPs in MMUS1495. E-G, Individual transition matrices for E, CP01, F, CP02, and G, CP03, showing trajectories of metastatic spread between organs and extent of expansion within an organ. Numbers range from 0 (absent in site) to 1 (confinement within site). Transition events occurring at rates below 1% fall below the threshold and not shown. Seeding trajectories are displayed above each matrix.

Figure 6.. Tissue seeding topologies of clonal populations.

A, Possible seeding topologies delineating the routes of tumor cell expansion and spread within and between different organs. B, Example of expected migration patterns based on possible seeding topologies. C-E, Actual seeding topologies exhibited by mice in EvoCaP-LP (C) and EvoCaP-HP (D) cohorts compared to seeding topologies of clones analyzed from a human PCa dataset (E). Pie charts indicate percentages of each possible type of seeding topology based on total number of CPs exhibiting each topology. Total CPs undergoing a specific type of topology are indicated by percentages. Topologies were considered for CPs in the primary tumor that were confined to the primary site, disseminated to a secondary site (either seeded individually or in parallel with more than one site), or re-seeded from a secondary site. Topologies of CPs from metastatic sites include those that remained confined to the same site, seeded another secondary organ, or traveled bidirectionally between different metastatic sites. F, Frequency of seeding topologies observed across human patients from Gundem et al., 2015 (10) (n=4), HP mice (n=5), and LP mice (n=5). Percentages for all CPs for each mouse and human patient are shown. No significant differences between groups were detected for any seeding topology (unpaired two-sample Wilcoxon test, Bonferroni-adjusted p>0.05).