*2.1. Shared Genomic Landscape*

In this study, we retrospectively analysed 13 PCa cases that had micro- or macro-metastatic disease at the time of sampling for genomic interrogation. Patient clinical and pathological characteristics are summarized in Table 1. Sixteen tumour samples comprised of nine primary and seven metastatic biopsies and the sites of concurrent or subsequent metastases included: bone (seven cases), lymph nodes (four cases), and brain (one case), while a single case had biochemically relapsed without evidence of macro-metastatic disease on conventional imaging. Figure 1 summarizes the commonalities and differences in the genomic landscape between our primary and metastatic samples, while placing our cases in context with the current knowledge based on large PCa WGS efforts. For the latter, we focused on the study published by Wedge et al. in 2018 for 112 patients (92 primary and 20 metastatic) with the metastases evenly distributed between HSPC and CRPC and biased towards lymph node metastasis (15/20) [8].

Common predisposing germline variants (Table S1) in our samples include the *EHBP1* rs721048 (c.1185+30064G > A) intronic variant in five (38%) and *FGFR4* rs35185519011 (c.1162G > A, p.Gly388Arg) in ten (77%) cases. Reported in 9% of the healthy population [24], the *EHBP1* rs721048 A-allele has been associated with a more aggressive PCa [25]. The functional variant in *FGFR4*, although present in 30% of the healthy population, may predispose PCa patients to an accelerated disease course [26]. Ten patients had one of two SNPs (rs1859962, rs8072254) in non-coding regions of the 17q24.3 locus previously associated with PCa susceptibility [27].

Common somatic alterations include *ETS* fusions (seven cases) and *TP53* alterations (six cases). Tumour mutational burden (TMB) was generally low, ranging from 0.73 to 5.79 mutations/Megabase (mut/Mb) (IQR 1.30–2.09), and did not correlate with disease stage at sampling (Figure 1, Table S2). Percent genome altered (PGA) ranged from 2.2% to 63.9% (IQR 2.79–19.4%) (Table S2).

We observed a prevalence (11/13 cases) of somatic copy number alterations (SCNA) affecting at least one DNA damage response (DDR) pathway gene (Table S4). Losses in *FANCA*, which helps recruit DNA repair proteins to areas of DNA damage [28], were present in five (38%) cases, while one case harboured germline and somatic *BRCA2* alterations. With variation depending on the gene sets tested for and stage of disease, DDR gene alterations occur in approximately 20% of PCas with *BRCA2* alterations reported for 3% of prostatic and 12% of metastatic samples [6,7,29]. Aside from DDR, the most commonly impacted pathways were Phosphoinositide 3-kinase (PI3K, 7/13 (54%) cases), Mitogen-activated protein kinase (MAPK, 10 (77%) cases) and Wnt (9 (69%) cases) (Table 2). PI3K and MAPK are intracellular and extracellular signalling pathways, respectively, that are key to the regulation of the cell cycle and, like certain DDR pathways, are therapeutically targetable (manipulable) with inhibitory drugs [30,31]. The Wnt signalling pathway is a cellular pathway involved in cell growth, embryogenesis and cell cycle progression, the activation of which has been implicated in progression to CRPC and treatment resistance [32]. Previous studies have found that approximately 25% of primary PCas harbour PI3K or MAPK pathway alterations while nearly 50% of metastatic CRPC samples have PI3K alterations [6,7] and 32% MAPK amplifications [30]. In our study, 6 of the 7 samples with somatic alterations impacting PI3K were in the primary tissue, and MAPK alterations were seen in 4/7 (57%) of the metastases and 6/8 (75%) of the primaries.

Overall, SCNAs and SVs, rather than single nucleotide variants (SNVs) and small insertions and deletions (sequences of no more than 50 nucleotides in length, indels), were more commonly acquired in PCa relevant genes (Tables S1–S6). The addition of WGM identified 120 SVs not identified by WGS alone, several of which overlapped with oncogenic and/or tumour suppresser genes (Table 3, Table S6). In particular, large insertions and duplications were typically missed by our short-read WGS approach. However, no recurrent WGM-derived SVs were observed across the cases.

**Figure 1.** Summary of genomic landscape relevant to tumour purity and related to clinical and pathological features for 16 samples from 13 patients, and further compared to the Wedge et al. data. Met-HSPC: metastatic hormone sensitive PCa; met-CRPC: metastatic castration-resistant PCa; TMB: Tumour mutational burden; SNV: single nucleotide variants; Indel: small insertion or deletion; Gains and Losses: somatic copy number alterations (SCNA); SV: Structural variation including large insertions or deletions, inversions, translocations and duplications; PSA: prostate-specific antigen (ug/L); ETS: presence of ETS fusion event; TP53: presence of *TP53* alteration including SNV, SCNA or SV; SPOP: presence of *SPOP* SNV; BCR: biochemical recurrence; ISUP: International Society of Urological Pathologists cancer grade (correlates to Gleason scores). Error bars for Wedge et al. data reflect +/− one standard deviation of the sub-group totals. Sample identifiers in red, with matching red bar plots, are indicative of patients pre-treated with ADT.

Recurrent non-coding events in key PCa-associated genes have been reported [8,11,33–37], including transcription factor (TF) binding sites. Alterations at key non-coding sites within our cases are summarised in Table 4. Common to CRPC, SCNAs or SVs upstream of the androgen receptor (AR) gene were not seen in our cases, which is unsurprising given the hormone sensitive status of most of our patients. All but two of our samples contained non-coding AR binding site mutations (Table 4). Overall, 20% of the somatic SNVs or indels affected at least one TF binding cluster. However, no sample was significantly enriched for mutations within TF binding clusters and no TFs were enriched for mutations. Notably, 0.3% of the 10.5 million TF binding clusters analysed correspond to *JUN*, an average of 1.2% (0.5–1.5%) of somatic SNV in *JUN* binding clusters. JUN is a transcription factor that antagonizes AR signalling [38].

Excluding COSMIC Mutational Signature 1 common to all cancers, we observed a predominance of Mutational Signatures 3 and/or 8 (Figure 2A) that generally reduced in proportion from the clonal to subclonal stages of tumour evolution (Figure 2B). Known to be associated with DDR gene alterations [39–41], Signature 8 was particularly common in the primary 8/9 (89%) versus metastatic 3/7 (43%) samples, with notable loss in both of case 19651's lymph node metastases. In contrast, Signature 5, which is seen in most cancer types, particularly in smokers [39], and Signature 16, most often associated with liver cancer, both increased in the subclonal stage of tumour evolution.

**Figure 2.** COSMIC Mutational Signatures (**A**) Proportion of signatures in each sample, for Signatures with >5% contribution; (**B**) Clonal vs. subclonal signature exposures. mHSPC: metastatic hormone-sensitive prostate cancer; mCRPC: metastatic castration-resistant prostate cancer.

When viewed in detail, each patient had unique features with potential therapeutic implications. This highlights the relevance of genomic information for guiding therapeutic decisions, including data derived from primary tumour tissue. Here, we discuss how the course of treatment for each patient may have been influenced by the availability of their genomic data.



 Diagnosis; Sample biopsy genomic interrogation; Androgen deprivation therapy; prostatectomy; adjuvant sRT=salvage radiotherapy; Rl =Relapse; ECOG=Eastern Cooperative Oncology Group Performance Status; wks: weeks; BCR=Biochemical relapse; mos =months;

 NR = Not

relapsed; NA = Not applicable.



**Table 3.** Structural variants identified by optical mapping as compared to whole genome sequencing. INS: Insertion; DEL: deletion; DUP: duplication; INV: inversion;Intra-Chr: intrachromosomal translocation; Inter-Chr: interchromosomal translocation.



