Evolutionary paths to ARSI resistance in LNCaP/AR xenograft model

(A) The LNCaP/AR barcode cell line was generated by lentiviral infection with ClonTracer library and grafted to hormonally intact mice (control) or physically pre-castrated mice treated with enzalutamide (Enz) (pre-engraftment ARSI). (B) Mean tumor volumes ± SEM (N=8 in each group). (C-D) Pie charts (C) and bar graphs (D) showing barcode distribution in the control and ARSI-resistant (ARSIR) tumors. In bar graphs, each bar represents each barcode with y-axis showing relative ratio, and the x-axes are identical across all the graphs in a decreasing order of abundance of 500 most enriched barcodes in pregraft (cell population used for xenograft, top graph). Barcode distribution of the full tumor set can be found in Supplementary Fig. 1. (E) Heatmap depicting Pearson correlation analysis (r) of barcodes between the tumors and hierarchical clustering on the values. (F) The LNCaP/AR barcode cell line was grafted to hormonally intact mice and when the tumors reached ∼200mm3, half of the mice were physically castrated and treated with Enz (post-engraftment ARSI) to test if the pre-existing ARSI-resistant clones will be selected in ARSIR tumors. (G) Mean tumor volumes ± SEM after Enz treatment (N=6 in control and N=7 in Enz treated group). (H) The relative ratio of two pre-existing ARSI-resistant clones enriched in control and ARSIR tumors. (I-J) Pie charts (I) and bar graphs (J) showing barcode distribution in control and ARSIR tumors. Barcode distribution of the full tumor set can be found in Supplementary Fig. 3. (K) Heatmap depicting Pearson correlation analysis (r) of barcodes between the tumors and hierarchical clustering on the values.

Evolutionary paths to ARSI resistance in CWR22Pc xenograft model

(A) The CWR22Pc cells containing cancer associated fibroblast (CAF) were infected with lentivirus containing ClonTracer library and grafted to hormonally intact mice (control) or physically pre-castrated mice treated with enzalutamide (Enz) (pre-engraftment ARSI). (B) Mean tumor volumes ± SEM (N=10 in control and N=8 in Enz treated group). Note that the Enz treatment is stopped at 21 weeks after grafting. (C-D) Pie charts (C) and bar graphs (D) showing barcode distribution in control and ARSI-resistant (ARSIR) tumors. In bar graphs, the barcodes in x-axes are identical across all the graphs in a decreasing order of abundance of 500 most enriched barcodes in pregraft. Barcode distribution of the full tumor set can be found in Supplementary Fig. 7. (E) Pearson correlation analysis (r) of barcodes between the tumors shows that the control and ARSIR tumors are clustered separately. (F) The CWR22Pc barcode cell line was grafted into hormonally intact mice and when the tumors reached ∼200mm3, half of the mice were physically castrated and treated with Enz (post-engraftment ARSI). (G) Mean tumor volumes ± SEM after Enz treatment (N=5 in control and N=6 in Enz treated group). (H-I) Pie charts (H) and bar graphs (I) showing barcode distribution in control and ARSIR tumors. Barcode distribution of the full tumor set can be found in Supplementary Fig. 6. (J) Pearson correlation analysis (r) of barcodes between the tumors shows that the control and ARSIR tumors are clustered together.

Evolutionary paths to ARSI resistance in LNCaP/AR model converge on GR upregulation.

(A) The clonal isolation was performed by sorting single cell of LNCaP/AR barcode cell line into 96-well plate and sanger sequencing of barcode in each clonal line. One clone enriched across control tumors (Pre-Intact), one clone enriched across ARSIR tumors in pre-engraftment ARSI setting (Pre-ARSIR), and four clones not enriched in the tumors (cloneA through D) were isolated. (B) The clones were cultured in regular media with vehicle or hormone-deprived media with 10μM Enz for 8 days. The cell viability values of Enz treated cells were normalized to the values of vehicle treated cells ± SD of two biological replicates. P-values were determined by one-way ANOVA compared to the value of Enz treated Pre-Intact clone. (C) GR is upregulated in Pre-ARSIR1 clone and 3 days of 10μM Enz treatment further increased GR level in Pre-ARSIR1 clone. (D) qRT-PCR shows that GR mRNA levels (mean ± SD) are upregulated in ARSIR tumors derived from pre-engraftment (Fig. 1B) and post-engraftment (Fig. 1G) ARSI studies. (E) Knockdown of AR or GR re-sensitizes Pre-ARSIR1 clone to 10μM Enz. The cell viability values were normalized to the value of vehicle treated shRenilla (shRL) ± SD of two biological replicates. P-values were determined by two-way ANOVA compared to shRL. (F) The GFP-labeled Pre-ARSIR1 clone was infected with shRL, shAR or shGR and mixed with Azurite-labeled Pre-Intact clone and clonesA-D infected with shRL. The mixed population was then cultured with 10μM Enz and the relative ratio of cells with blue, green and no color were analyzed using flow cytometry at day 0, 8 and 15. (G) The LNCaP/AR barcode cell line was grafted into hormonally intact mice and when the tumors reached ∼400mm3, the mice were physically castrated and treated with Enz. The tumors were collected at 0, 1 and 4 weeks after treatment, and at 14 weeks when became resistant to the treatment, and stained with RFP (red, grafted LNCaP/AR cells), GR (green) and Ki-67 (grey). (H) The numbers of RFP+/GR+ cells were normalized to the numbers of total RFP+ cells in each tissue section. Mean ± SD, N=4 in each group.

Evolutionary paths to ARSI resistance in CWR22Pc model converge on HER3.

(A) One clone enriched across the ARSIR tumors in pre-engraftment ARSI setting (Pre-ARSIP), three clones not enriched in the tumors (cloneA through C) and CAFs were isolated from CWR22Pc barcode cell line. (B) The clones were cultured in control or hormone-deprived media for 6 days. The cell viability values in control media were normalized to the value of Pre-ARSIP ± SD and the values in hormone-deprived media were normalized to the values in control media ± SD of two biological replicates. NS = not significant, *p<0.05, two-way ANOVA compared to Pre-ARSIP. (C) The clones were cultured in regular or hormone-deprived media with increasing concentration of conditioned media (CM) derived from CAF for 6 days. The cell viability values were normalized to the value of no CM ± SD of two biological replicates. ****p<0.0001, two-way ANOVA compared to cloneA. (D) The clones were cultured in serum-free media for 2hrs and cultured with each indicated concentration of conditioned media (CM) for 10min. (E) The clones were cultured in hormone-deprived media with 0.01μM neratinib and increasing concentration of conditioned media (CM) derived from CAF for 6 days. The cell viability values were normalized to the value of no CM ± SD of two biological replicates. (F) NRG1 and total HER2, and phospho-HER3, AKT and ERK levels are increased in ARSIR tumors from pre-engraftment ARSI study (Fig. 2B). (G) NRG1 and total HER2, and phospho-HER2, HER3 and ERK levels are increased in some of the ARSIR tumors from post-engraftment ARSI study (Fig. 2G).

ARSI resistance in established tumors requires inter-clonal cooperation.

(A-B) Mix of 6 LNCaP/AR clones (Pre-Intact, Pre-ARSIR1 and clonesA-D) were grafted to control, pre-engraftment ARSI treated (A), or post-engraftment ARSI treated (B) mice. Barcode distribution in pregraft, control and ARSIR tumors are shown in pie charts. (C) GR protein levels were accessed in control and ARSIR tumors in B. (D) Mix of 6 LNCaP/AR clones or Pre-Intact clone alone were grafted to post-engraftment ARSI setting. (E) Mean tumor volumes ± SEM after ARSI treatment (N=7 in each group). P-value was determined by two-tailed t-test at each time point. (F) Mix of indicated clones were grafted to post-engraftment ARSI setting. Tumor volumes were measured 12 weeks after ARSI treatment (N=9 in Pre-Intact and Pre-Intact + CloneA, N=7 in Pre-Intact + Pre-ARSIR1, and N=5 in Pre-Intact + CloneA + Pre-ARSIR1 group). Mean tumor volumes ± SEM. P-value was determined by two-tailed t-test. For all panels, NS = not significant, *p<0.05.

Enrichment of pre-existing resistance clones in pre-engraftment ARSI setting.

(A) The LNCaP/AR barcode cell line was generated by lentiviral infection with ClonTracer library to enable 11,600 cells labeled with unique barcode sequence (details can be found in Materials and methods). The infected cells then went through 13 doublings and used for xenograft experiments. (B) Significantly lower number of barcodes are enriched in ARSI-resistant (ARSIR) tumors from Fig. 1B. Mean ± SD, N=7 in each group. ***p<0.001, two-tailed t-test. (C-D) Pie charts (C) and bar graphs (D) showing barcode distribution in control and ARSIR tumors. In bar graphs, each bar represents each barcode with y-axis showing relative ratio, and the x-axes are identical across all the graphs in a decreasing order of abundance of 500 most enriched barcodes in pregraft (cell population used for xenograft, top graph). (E) Distribution of the top 50 barcodes in pregraft shows that the barcodes most enriched in ARSIR tumors (blue bars) are some of the most enriched barcodes in pregraft.

In vitro effects of AR activation and inhibition on growth of pre-existing ARSI-resistant clones.

(A) LNCaP/AR barcode line was culture with either DMSO vehicle, 10μM Enz or 10nM DHT for 7 days, and then gDNA was extracted for barcode sequencing. (B-C) Distribution of the top 500 (B) or the top 50 (C) barcodes enriched in LNCaP/AR barcode cell line before the treatment (day 0, top graph) and in triplicates of each treatment group. The relative abundance of pre-existing ARSI-resistant clones (blue bars in Supplementary Fig. 1E and here) are increased by vehicle or Enz, and decreased by DHT treatment.

ARSI resistance in established LNCaP/AR tumors is adaptive despite the presence of pre-existing resistant clones.

(A) Significantly lower number of barcodes are enriched in ARSIR tumors from Fig 1G. N=5 in control and N=6 in ARSIR group. ***p<0.001, two-tailed t-test. (B-C) Pie charts (B) and bar graphs (C) showing barcode distribution in control and ARSIR tumors. In bar graphs, the barcodes in x-axes are identical across all the graphs in a decreasing order of abundance of 500 most enriched barcodes in pregraft. (D) Top, Venn diagrams showing the overlaps between barcodes enriched in control and ARSIR tumors in pre-engraftment ARSI (Fig. 1B) and post-engraftment ARSI (Fig. 1G) studies. The % overlap is based on the number of barcodes enriched in control tumors. Bottom, scatter plots showing the average % of each barcode in tumors and bar graphs showing cumulative % of barcodes enriched only in control (blue), ARSIR (red) or both groups (yellow).

Enrichment of pre-existing persister clones in pre-engraftment ARSI setting in CWR22Pc.

(A) The CWR22Pc cells containing mouse-derived cancer associated fibroblast (CAF) were infected with ClonTracer library to generate cell line labelled with 7,823 unique barcode, and then went through 14 doublings to have enough number for xenograft experiments. Since the CAFs are also labelled with barcode, we FACS sorted the human EpCAM-positive/mouse MHC-class1-negative population from the cells used for injection at each xenograft experiment and referred to it as ‘pregraft’. (B-D) Pie charts (B) and bar graphs (C-D) showing barcode distribution in control or ARSI-resistant (ARSIR) tumors from Fig. 2B. In bar graphs, the barcodes in x-axes are identical across all the graphs in a decreasing order of abundance of 500 (C) or 50 (D) most enriched barcodes in pregraft.

Replicate experiment showing enrichment of pre-existing resistance clones in setting of pre-engraftment ARSI therapy.

(A) Mean tumor volumes ± SEM of the repeated pre-engraftment ARSI experiment (N=6 in control and N=11 in Enz treated group). Note that the Enz treatment is stopped at 20 weeks after grafting. (B-C) Pie charts (B) and bar graphs (C) showing barcode distribution in control and ARSIR tumors. In bar graphs, the barcodes in x-axes are identical across all the graphs in a decreasing order of abundance of 500 most enriched barcodes in pregraft. (D) Heatmap depicting Pearson correlation analysis (r) of barcodes between the tumors shows that the control and ARSIR tumors are clustered separately.

ARSI resistance in established CWR22Pc tumors is adaptive despite the presence of pre-existing resistant clones.

(A-C) Pie charts (A) and bar graphs (B-C) showing barcode distribution in control and ARSIR tumors from Fig. 2G. In bar graphs, the barcodes in x-axes are identical across all the graphs in a decreasing order of abundance of 500 (B) or 50 (C) most enriched barcodes in pregraft. (D) Top, Venn diagrams showing the overlaps between barcodes enriched in control and ARSIR tumors in pre-engraftment and post-engraftment ARSI studies in Fig. 2. The % overlap is based on the number of barcodes enriched in control tumors. Bottom, scatter plots showing the average % of each barcode in tumors and bar graphs showing cumulative % of barcodes enriched only in control (blue), ARSIR (red) or both groups (yellow).

Replicate experiment confirming that ARSI resistance in established CWR22Pc tumors is adaptive despite the presence of pre-existing resistant clones.

(A) Mean tumor volumes ± SEM of the repeated post-engraftment ARSI study (N=8 in control and N=6 in Enz treated group). (B-C) Pie charts (B) and bar graphs (C) showing barcode distribution in control and ARSIR tumors. In bar graphs, the barcodes in x-axes are identical across all the graphs in a decreasing order of abundance of 500 most enriched barcodes in pregraft. (D) Pearson correlation analysis (r) of barcodes between the tumors shows that the control and ARSIR tumors are clustered together. (E) Top, Venn diagrams showing the overlaps between barcodes enriched in control and ARSIR tumors in repeated pre-engraftment and post-engraftment ARSI studies (Supplementary Figs. 5 and 7). The % overlap is based on the number of barcodes enriched in control tumors. Bottom, scatter plots showing the average % of each barcode in tumors and bar graphs showing cumulative % of barcodes enriched only in control (blue), ARSIR (red) or both groups (yellow).

Pre-existing ARSI-resistant LNCaP/AR clones have higher baseline AR activity and GR expression.

(A-B) Pre-ARSIR1 and Pre-ARSIR2 clones have higher IC50 values of Enz compared to other clones. IC50 values ± SD of three biological replicates. P-values were determined by one-way ANOVA. (C-D) Pre-ARSIR1 and Pre-ARSIR2 clones have higher AR reporter activities compared to the other clones. The values of luciferase AR reporter activities are normalized to Pre-Intact clone ± SD of three biological replicates. P-values were determined by one-way ANOVA. (E) GR is upregulated in both Pre-ARSIR1 and Pre-ARSIR2 clones and Enz treatment further increase GR expression. (F) The knockdown level of AR and GR from the cells in Fig. 3E. (G) Knockdown of AR or GR re-sensitize Pre-ARSIR2 clone to 10μM Enz. The cell viability values were normalized to the value of vehicle treated shRenilla (shRL) ± SD of two biological replicates. P-values were determined by two-way ANOVA compared to shRL. (H) The knockdown level of AR and GR from the cells in G. For all panels, NS = not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Mean tumor volumes ± SEM (N=8 in each group) at 5 weeks after grafting the 6 LNCaP/AR clones into hormonally intact mice. **p<0.01, ****p<0.0001, one-way ANOVA.