Background: Liver resection (LR) is a potentially curative option for early-stage hepatocellular carcinoma (HCC), and immune checkpoint inhibitors (ICI) are being investigated in the perioperatory setting to improve long-term relapse rates and survival outcomes. Methods: PRIME‐HCC is a phase Ib study testing nivolumab (3 mg/kg, day 1 and day 22) plus ipilimumab (1mg/kg, day 1 only) (N+I) prior to LR in resectable HCC. Primary endpoint was safety, while secondary endpoints included radiological response (per RECIST v1.1) and pathological response on resection specimens (ReS). Translational endpoints encompassed analysis of tissue samples with multiplex immunohistochemistry (MIHC) and gut microbiota profiling with high-throughput sequencing and their correlation with response. Results: As of 7^th of February 2023, 25 patients (pts) were enrolled, all Child-Pugh A, mostly male (n=18, 72%), with cirrhosis (n=13, 52%), and viral hepatitis (44%, 4 HBV and 7 HCV). Any-grade adverse events (AEs) occurred in 88% pts (n=22), of which 24% (n=6) grade (G) 3. No G4-5 AEs were recorded. Any-grade treatment-related (tr)AEs were reported by 72% pts (n=18), including 8% G3 trAEs (n=1 ALT/AST elevation and n=1 dyspnea). After a median follow-up of 14.5 months (95% CI 7.0-22.0), 21 pts were assessed for radiological response and underwent surgery. Median time from treatment start to LR was 2.4 months (interquartile range 2.2-2.7). LR was delayed in 1 pt (4.8%) because of G2 tr-hypothyroidism but occurred uneventfully. Objective response rate was 29% (n=5 partial responses, n=1 complete response [CR]); disease control rate was 95% (n=20), with 1 progressor having a mixed HCC/cholangiocarcinoma. Major pathological response (MPR), defined as ≥70% tumour regression, was observed in 56% (n=9) of the pathologically evaluable ReS (n=16), with 38% pathological CR (n=6). In the MIHC analyses, pretreatment biopsies of MPR pts were significantly enriched in peritumoral CD4+ and CD8+ cells, with no intratumoral difference. When comparing baseline biopsies and ReS, MPR pts had a significant decrease of intratumoral FOXP3+/CD4+ regulatory T cells (Tregs), while peritumoral Tregs increased significantly in non-MPR ReS. In baseline stool samples, α and β diversity did not differ across responders (R) and non-responders (NR), nor it changed after ICI. However, pre-treatment stool samples of NR pts were significantly enriched in Alistipes and Colinsella genera. Conclusions: Neoadjuvant N+I is feasible and tolerable for early-stage HCC, with promising radiological and pathological response rates. Study of tissue and gut microbiota highlights tumor and host-related changes. Immune infiltrate and gut microbiota composition lend themselves as predictors of anti-tumour efficacy of N+I combination. Clinical trial information: NCT03682276.