**3. Results**

We enrolled three consecutive patients with a diagnosis of ICC confirmed at pathology. Table 1 summarises the patients' characteristics. The mean tumour size was 92 mm (range 60–120). At PET-CT, the mean SUVmax was 11.2 (8.9–14.7), the mean SUVmin was 5.3 (5.1–5.5), and the mean difference between the two was 5.8 (3.5–9.6). The synchronisation of PET-CT with IOUS and its navigation was successful in all patients.


**Table 1.** Clinical characteristics of the patients.

Table 2 summarises the pathology data. The IH of ICC was evident in different analyses. One patient had a lower tumour grading in the SUVmin area than in the SUVmax one (G1 vs. G2). One patient had a phenotypic IH, i.e., variable CK19 positivity in areas with a different uptake. One patient had a molecular IH: FGFR2 translocation was evident in the high-uptake area, while it was not in the low-uptake one. PET-CT uptake was also associated with the proliferative index in two patients (70% in the SUVmax area vs. 10% in the SUVmin area of one patient; 70 vs. 20%, respectively, in one). Finally, IH on PET-CT corresponded to heterogeneous immune infiltration: SUVmax areas had a higher CD8+ infiltrate in all patients (a mean of 15 vs. 8%), and a higher CD4+ (30 vs. 10%), CD68+ (25 vs. 10%), and CD163+ (30 vs. 12%) infiltrate in two patients. Metabolic indexes, PD1, PD-L1, and p53 expression were similar between areas.


**Table 2.** Summary of the pathology results.

**CD163**, marker of M2 macrophages; **CD3** (cluster of differentiation 3), marker of T-lymphocytes; **CD4**, marker of helper/inducer T-lymphocyte; **CD68**, pan-macrophage or M1 marker; **CD8**, marker of suppressor/cytotoxic T-lymphocyte; **CK19**, cytokeratin 19; **CK7**, cytokeratin 7; **CS**, citrate synthase; **FGFR2**, fibroblast growth factor receptor 2; **G6PD**, glucose-6-phosphate dehydrogenase; **Ki67**, proliferation index; **LOH**, loss of heterozygosity; **Neg**, negative; **p53**, tumour suppressor protein; **PD-1**, programmed cell death protein 1; **PD-L1**, programmed death ligand 1; **SUV**, (standardised uptake value) semiquantitative parameter of FDG uptake; **WT**, wild type.

The pathology data of IOUS-guided biopsies and macrobiopsies after resection obtained from the same area were concordant.

#### **4. Discussion**

ICC is an aggressive malignancy with a poor prognosis. Standard chemotherapy has scarce disease control [12], but targeted therapies and immunotherapy could change this scenario. Some of the commonest ICC mutations concern the p53 pathway, Ras/Raf/MEK/ERK pathway, metabolic pathway (IDH1/IDH2), FGFR2, and 1p36 [10,11,13]. To date, targeted therapies for FGFR2 rearrangements and IDH1 mutations have been approved, and some other drugs have had tissue-agnostic approval [14,15]. However, the effectiveness of systemic therapies is limited by profound tumour genetic heterogeneity [10,11]. Walter et al. depicted varying expression patterns of MSH6 (mismatch repair protein) in peripheral and central areas of ICC [16]. Goyal et al. reported intra-tumoural clonal heterogeneity, in terms of acquired resistance to FGFR inhibition, in patients with FGFR2-fusion-positive tumours [17]. The possibility of predicting IH with non-invasive imaging is of major interest but has not been demonstrated yet.

ICC malignant cells have increased their expression of glucose transporters and a high activity of hexokinase, which leads to augmented glucose metabolism [4]. It corresponds to an increased FDG accumulation at PET-CT, especially in moderately and poorly differentiated ICCs [5]. High glucose metabolism in ICC is expected to be associated with increased tumour aggressiveness. Indeed, Seo et al. reported a high SUV as an independent predictor of postoperative recurrence [6].

We investigated the association between the heterogeneous uptake of ICC at PET-CT and IH. Among liver tumours, ICC is the most adequate for this analysis: it is usually diagnosed at an advanced stage (large masses); FDG PET-CT uptake is often non-homogeneous; and resectable patients do not receive preoperative chemotherapy, which could compromise PET-CT findings. Navigation technology provided a fundamental contribution. It

is commonly used to guide the percutaneous interstitial treatment of tumours not visible on ultrasound [7]. In liver surgery, navigation technology merges preoperative and intraoperative imaging to identify the anatomy and the correct plane [8]. We used the fusion of preoperative PET-CT with IOUS to have an accurate identification of tumour areas with a different uptake at PET-CT. The intraoperative analysis allowed us to maximise the precision of the biopsy and unequivocally ascertain the capability of PET-driven biopsies to detect IH.

In the present series, PET-CT effectively caught IH. FDG uptake was associated with proliferative index (Ki67) and tumour grading: the areas with the highest SUV were the most aggressive parts of ICC. We also observed interesting results concerning genetic mutations and immune infiltration. In one patient, PET-CT identified a heterogeneous mutational status of FGFR2 (a wild-type in the SUVmin area and translocated in the SUVmax one). The remaining two patients had a wild-type status of FGFR2 in all biopsies. Considering immune infiltration, tumour areas with a higher FDG uptake had higher levels of T-lymphocytes (CD3+ and CD4+/CD8+) and macrophages (CD68+/CD163+) compared to areas with a lower uptake. Those data are clinically relevant: FGFR2 mutations are the target of approved drugs [14,18,19], and the immune infiltrate is a major determinant of prognosis in ICC patients [20,21]. A further result deserves consideration. In general, the key enzymes of anaerobic glycolysis and mitochondrial respiration (G6PD and CS, respectively) did not show a clear association with the SUV. Even if FDG PET-CT detects an augmented glucose metabolism [22], the different uptake did not always correspond to a heterogeneous metabolic pattern. Due to the study design (three patients), we can formulate some hypotheses about the mechanisms underlying the heterogeneous FDG uptake, also considering that they can vary among patients. In one patient, the expression of citrate synthase, which is used in oxygen-dependent ATP production, dropped in the high-SUV area. This finding is consistent with the concept of increased glucose consumption in the hypoxic areas of the tumours, which are forced to switch to the energy-inefficient anaerobic glycolysis and thus require many times more substrate for the same ATP output [23]. In the remaining two patients, the high-uptake areas were probably related to a major increase in the proliferative index (of both patients) and an FGFR2 translocation (in one). Investigations on the link between the latter gene and glucose metabolism are thus far limited, but the FGF/FGFR pathway involves anti-apoptosis signalling, proliferation, and angiogenesis [24].

The present study is in line with modern oncological research. Advanced imaging and analyses achieved excellent results for ICC, being able to provide a non-invasive prediction of tumour pathology data and prognosis [25,26]. Focusing on PET-CT, Yugawa et al. demonstrated that FDG uptake is associated with immune infiltration [27]. Fiz et al. reported that the radiomic analysis of the ICC and peritumoral tissue accurately predicts tumour grading, microvascular invasion, and survival [28]. Our preliminary data are coherent with such literature but represent a major step forward, thanks to IH mapping.

The proposed approach is clinically relevant for at least two reasons. First, the fusion of two different imaging modalities—morphological (ultrasound) and functional (PET)—provided a non-invasive depiction of ICC heterogeneity and detection of the most significant tumour portions. Even if our data are preliminary, the concordance of these results among multiple samplings from the same area strengthens the reliability and the reproducibility of the present technique. PET-driven biopsies could become a new standard in ICC patients: to catch the most relevant and aggressive areas of the tumour, have a more precise prediction of prognosis, and schedule a more effective patient-tailored treatment. Theoretically, the same approach could be applied to other liver tumours (primary or metastatic) and tracers. Liver metastases from colorectal cancers have both an intense FDG uptake with heterogeneous areas and a proven intralesional heterogeneity that correlates with prognosis [29–31]. Metastatic neuroendocrine tumours have a known inter-lesional heterogeneous DOTA peptides uptake, which can bear relevance for treatment strategies [32]. The tracers of the PSMA molecules can visualise the heterogeneity of prostate

cancer metastases and, more recently, primary hepatic malignancies [33,34]. The uptake can depict variations in vascularity across the tumoural volume [34].

Second, the present technique was the first one that reliably associates the different FDG-uptake areas with tumour heterogeneity at a phenotypic, molecular, and genetic level and with immune infiltration. Our experience only provides a preliminary exploration of the concept but could be the basis for a better understanding of IH, a precision-medicine approach, and the identification of new biomarkers and therapeutic targets. By analysing a larger population, we could identify the SUV values and patterns which are able to non-invasively predict tumour characteristics.

### **5. Conclusions**

The present study demonstrates that the fusion of morphological and functional imaging modalities may allow an in vivo and reliable evaluation of tumour heterogeneity. Discrepant intra-tumoural phenotypic, molecular, and genetic patterns were identified, as well as heterogeneous immune infiltrations. The proposed approach could increase the efficacy of percutaneous biopsies and could be the basis for a better understanding of IH.

**Author Contributions:** Conceptualisation, L.V., E.L., F.F., L.D.T., A.A., L.R., L.S., A.C. and G.T.; methodology, L.V., E.L., F.F., A.D. and L.S.; software, E.L. and L.S.; formal analysis, L.D.T. and A.D.; investigation, L.V., E.L., L.D.T., A.D., L.S., A.C. and G.T.; data curation, A.A., F.F. and L.R.; writing—original draft preparation, L.V., E.L., F.F. and A.D.; writing—review and editing, L.D.T., A.A., L.R., L.S., A.C. and G.T.; supervision, L.V., A.A., L.R., L.S., A.C. and G.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board (or Ethics Committee) of Humanitas Clinical & Research Hospital (protocol code 146/20 on 20 February 2020).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available from the corresponding author on reasonable request.

**Conflicts of Interest:** The authors declare no conflict of interest pertinent to the present manuscript. Considering the conflicts of interest in general, we state that: (1) L.V. received speaker's honoraria from Johnson & Johnson. (2) A.C. received speaker's honoraria from Advanced Accelerator Applications, General Electric Healthcare, Sirtex Medical Europe and AmGen Europe; received travel grants form General Electric Healthcare and Sirtex Medical Europe; he is a member of Blue Earth Diagnostics' and Advanced Accelerator Applications' advisory boards; received scientific support, in terms of a three-year Ph.D. fellowship, from the Sanofi Genzyme. (3) L.R. reports receiving consulting fees from Amgen, ArQule, AstraZeneca, Basilea, Bayer, BMS, Celgene, Eisai, Exelixis, Genenta, Hengrui, Incyte, Ipsen, IQVIA, Lilly, MSD, Nerviano Medical Sciences, Roche, Sanofi, Servier, Taiho Oncology, Zymeworks; lecture fees from AbbVie, Amgen, Bayer, Eisai, Gilead, Incyte, Ipsen, Lilly, Merck Serono, Roche, Sanofi; travel expenses from AstraZeneca; and institutional research funding from Agios, ARMO BioSciences, AstraZeneca, BeiGene, Eisai, Exelixis, Fibrogen, Incyte, Ipsen, Lilly, MSD, Nerviano Medical Sciences, Roche, Zymeworks. All other authors have no relevant disclosures.

#### **References**

