Mining the interpretable prognostic features from pathological image of intrahepatic cholangiocarcinoma using multi-modal deep learning.

BACKGROUND: The advances in deep learning-based pathological image analysis have invoked tremendous insights into cancer prognostication. Still, lack of interpretability remains a significant barrier to clinical application. METHODS: We established an integrative prognostic neural network for intrahepatic cholangiocarcinoma (iCCA), towards a comprehensive evaluation of both architectural and fine-grained information from whole-slide images. Then, leveraging on multi-modal data, we conducted extensive interrogative approaches to the models, to extract and visualize the morphological features that most correlated with clinical outcome and underlying molecular alterations. RESULTS: The models were developed and optimized on 373 iCCA patients from our center and demonstrated consistent accuracy and robustness on both internal (n = 213) and external (n = 168) cohorts. The occlusion sensitivity map revealed that the distribution of tertiary lymphoid structures, the geometric traits of the invasive margin, the relative composition of tumor parenchyma and stroma, the extent of necrosis, the presence of the disseminated foci, and the tumor-adjacent micro-vessels were the determining architectural features that impacted on prognosis. Quantifiable morphological vector extracted by CellProfiler demonstrated that tumor nuclei from high-risk patients exhibited significant larger size, more distorted shape, with less prominent nuclear envelope and textural contrast. The multi-omics data (n = 187) further revealed key molecular alterations left morphological imprints that could be attended by the network, including glycolysis, hypoxia, apical junction, mTORC1 signaling, and immune infiltration. CONCLUSIONS: We proposed an interpretable deep-learning framework to gain insights into the biological behavior of iCCA. Most of the significant morphological prognosticators perceived by the network are comprehensible to human minds.

Distribution and density of tertiary lymphoid structures predict clinical outcome in intrahepatic cholangiocarcinoma.

BACKGROUND & AIMS: The prognostic value and clinical relevance of tertiary lymphoid structures (TLSs) in intrahepatic cholangiocarcinoma (iCCA) remain unclear. Thus, we aimed to investigate the prognostic value and functional involvement of TLSs in iCCA. METHODS: We retrospectively included 962 patients from 3 cancer centers across China. The TLSs at different anatomic subregions were quantified and correlated with overall survival (OS) by Cox regression and Kaplan-Meier analyses. Multiplex immunohistochemistry (mIHC) was applied to characterize the composition of TLSs in 39 iCCA samples. RESULTS: A quaternary TLS scoring system was established for the intra-tumor region (T score) and peri-tumor region (P score) respectively. T scores positively correlated with favorable prognosis (p <0.001), whereas a high P score signified worse survival (p <0.001). mIHC demonstrated that both T follicular helper and regulatory T cells were significantly increased in intra-tumoral TLSs compared to peri-tumoral counterparts (p <0.05), and regulatory T cell frequencies within intra-tumoral TLSs were positively associated with P score (p <0.05) rather than T score. Collectively, the combination of T and P scores stratified iCCAs into 4 immune classes with distinct prognoses (p <0.001) that differed in the abundance and distribution pattern of TLSs. Patients displaying an immune-active pattern had the lowest risk, with 5-year OS rates of 68.8%, whereas only 3.4% of patients with an immune-excluded pattern survived at 5 years (p <0.001). The C-index of the immune class was statistically higher than the TNM staging system (0.73 vs. 0.63, p <0.001). These results were validated in an internal and 2 external cohorts. CONCLUSIONS: The spatial distribution and abundance of TLSs significantly correlated with prognosis and provided a useful immune classification for iCCA. T follicular helper and regulatory T cells may play a critical role in determining the functional orientation of spatially different TLSs. LAY SUMMARY: Tertiary lymphoid structures (TLSs) are associated with favorable prognosis in a number of cancers. However, their role in intrahepatic cholangiocarcinoma (iCCA) remains unclear. Herein, we comprehensively evaluated the spatial distribution, abundance, and cellular composition of TLSs in iCCA, and revealed the opposite prognostic impacts of TLSs located within or outside the tumor. This difference could be mediated by the different immune cell subsets present within the spatially distinct TLSs. Based on our analysis, we were able to stratify iCCAs into 4 immune subclasses associated with varying prognoses.

Inverse centrifugal effect induced by collective motion of vortices in rotating thermal convection.

When a fluid system is subject to strong rotation, centrifugal fluid motion is expected, i.e., denser (lighter) fluid moves outward (inward) from (toward) the axis of rotation. Here we demonstrate, both experimentally and numerically, the existence of an unexpected outward motion of warm and lighter vortices in rotating thermal convection. This anomalous vortex motion occurs under rapid rotations when the centrifugal buoyancy is sufficiently strong to induce a symmetry-breaking in the vorticity field, i.e., the vorticity of the cold anticyclones overrides that of the warm cyclones. We show that through hydrodynamic interactions the densely distributed vortices can self-aggregate into coherent clusters and exhibit collective motion in this flow regime. Interestingly, the correlation of the vortex velocity fluctuations within a cluster is scale-free, with the correlation length being proportional ( ≈ 30%) to the cluster length. Such long-range correlation leads to the counterintuitive collective outward motion of warm vortices. Our study brings insights into the vortex dynamics that are widely present in nature.

CircRNA UBAP2 serves as a sponge of miR-1294 to increase tumorigenesis in hepatocellular carcinoma through regulating c-Myc expression.

Circular RNAs (circRNAs) are a class of regulatory RNAs with complex roles in healthy and diseased tissues. However, the oncogenic role of circRNAs in hepatocellular carcinoma (HCC) remains poorly understood, including the mechanisms by which the circular ubiquitin-binding associated protein 2 (circUBAP2) contributes to tumorigenesis. We analyzed the expression of circUBAP2 in 20 paired samples of HCC and healthy tissue as well as in seven HCC cell lines via quantitative real-time polymerase chain reaction. Functional experiments, such as CCK8 viability assays, colony formation assays, wound healing, transwell assays and flow cytometry, were conducted to assess the effects of circUBAP2 in vitro. To further elucidate the mechanisms by which circUBAP2 acts, we conducted dual-luciferase assays, western blots, RNA pull-down assays and rescue experiments. CircUBAP2 was highly upregulated in most HCC tissues and was associated with poor prognosis. HCC patients with high circUBAP2 expression had greater vascular invasion and worse differentiation. Functionally, circUBAP2 overexpression enhanced HCC cell proliferation, migration and invasion and inhibited apoptosis. Furthermore, we found that circUBAP2 upregulated c-Myc expression by sponging miR-1294, thus contributing to hepatocarcinogenesis. Inhibiting circUBAP2 expression in HCC attenuated the oncogenic effects of c-Myc. These findings suggest that circUBAP2 promotes HCC growth and metastasis. CircUBAP2 may have value as an independent prognostic biomarker or as a new target for the treatment of HCC.

Laparoscopic vs. Open Repeat Hepatectomy for Recurrent Liver Tumors: A Propensity Score-Matched Study and Meta-Analysis.

Background: It remains unclear whether the short-term benefits of laparoscopic repeat hepatectomy (LRH) accrue to patients with recurrent liver tumors. The present study aimed to report our own center's experience and perform a meta-analysis to evaluate the safety and feasibility of LRH in comparison with open repeat hepatectomy (ORH) for treating recurrent liver tumors. Patients and Methods: A propensity score-matched study was performed including 426 patients receiving LRH or ORH for recurrent hepatocellular carcinoma between January 2017 and December 2018. Surgical outcomes and perioperative inflammation-based markers, including monocyte-to-lymphocyte ratio, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and systemic immune-inflammation index were collected from medical records and analyzed. Additionally, a systematic literature review was performed to identify relevant studies in PubMed, EMBASE, Web of Science, and Cochrane library databases up to October 1, 2020. Information including patient demographics, pathologic characteristics, and short-term outcomes was extracted and analyzed using random- or fixed-effects models. Results: Of 68 LRHs, 57 were matched with an ORH finally. Our study demonstrated that LRH was significantly associated with less intraoperative blood loss (50 vs. 100 mL; P < 0.001), lower rate of hepatic inflow occlusion (10.52 vs. 33.3%; P = 0.003), and shorter postoperative hospital stay (5 vs. 6 days; P = 0.001) after 1:1 propensity score matching. The operation time, rate of blood transfusion, and postoperative complications were similar between the two groups. Moreover, all four inflammation-based markers were significantly lower in LRH group on postoperative day 1. In the meta-analysis, a total of 12 studies comprising 1,315 patients receiving repeat hepatectomy met the selection criteria. Similar to our own study, the meta-analysis showed shorter hospital stay [standard mean difference (SMD) = -0.51, 95% confidence interval (CI) = -0.79 to -0.22, P < 0.001], less intraoperative blood loss (SMD = -0.79, 95% CI = -1.11 to -0.47, P < 0.001), and lower rate of major postoperative complications [odds ratio (OR) = 0.35, 95% CI = 0.19-0.66, P = 0.001] in the LRH group. There was no difference in the field of overall postoperative complication and operation time between LRH and ORH groups. Conclusion: Compared with ORH, LRH results in relatively better surgical outcomes and faster postoperative recovery. It could be considered a feasible and effective option for the treatment of recurrent liver tumors.

Exploring prognostic indicators in the pathological images of hepatocellular carcinoma based on deep learning.

OBJECTIVE: Tumour pathology contains rich information, including tissue structure and cell morphology, that reflects disease progression and patient survival. However, phenotypic information is subtle and complex, making the discovery of prognostic indicators from pathological images challenging. DESIGN: An interpretable, weakly supervised deep learning framework incorporating prior knowledge was proposed to analyse hepatocellular carcinoma (HCC) and explore new prognostic phenotypes on pathological whole-slide images (WSIs) from the Zhongshan cohort of 1125 HCC patients (2451 WSIs) and TCGA cohort of 320 HCC patients (320 WSIs). A 'tumour risk score (TRS)' was established to evaluate patient outcomes, and then risk activation mapping (RAM) was applied to visualise the pathological phenotypes of TRS. The multi-omics data of The Cancer Genome Atlas(TCGA) HCC were used to assess the potential pathogenesis underlying TRS. RESULTS: Survival analysis revealed that TRS was an independent prognosticator in both the Zhongshan cohort (p<0.0001) and TCGA cohort (p=0.0003). The predictive ability of TRS was superior to and independent of clinical staging systems, and TRS could evenly stratify patients into up to five groups with significantly different prognoses. Notably, sinusoidal capillarisation, prominent nucleoli and karyotheca, the nucleus/cytoplasm ratio and infiltrating inflammatory cells were identified as the main underlying features of TRS. The multi-omics data of TCGA HCC hint at the relevance of TRS to tumour immune infiltration and genetic alterations such as the FAT3 and RYR2 mutations. CONCLUSION: Our deep learning framework is an effective and labour-saving method for decoding pathological images, providing a valuable means for HCC risk stratification and precise patient treatment.

Laparoscopic hepatectomy enhances recovery for small hepatocellular carcinoma with liver cirrhosis by postoperative inflammatory response attenuation: a propensity score matching analysis with a conventional open approach.

BACKGROUND: The concurrent presence of liver cirrhosis and hepatocellular carcinoma (HCC) poses a challenge for laparoscopic surgeons to establish a routine practice. The aim of this study was to gather evidence and produce recommendations on the safe and effective practice of laparoscopic hepatectomy for patients with solitary HCC (≤ 5 cm) and liver cirrhosis. METHODS: Between October 2013 and October 2014, 356 curative hepatectomies were performed for patients pathologically diagnosed with solitary HCC (≤ 5 cm) accompanied by cirrhosis (stage 4 fibrosis). To overcome selection bias, a 1:2 match using propensity score matching analysis was conducted between laparoscopic and open hepatectomy. Perioperative outcomes were compared between the groups, including hospitalization, operation time, blood loss, and surgical complications. Perioperative inflammation-based markers, including systemic immune-inflammation index (SII), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and lymphocyte-to-monocyte ratio (LMR) were collected from medical records and analyzed. RESULTS: There were 43 and 77 patients in the laparoscopic and open groups, respectively. The laparoscopic group had less hepatic inflow occlusion (16.3% vs. 61%; P < 0.001), shorter operation time (155 vs. 170 min; P = 0.004), and shorter postoperative hospital stay (4 vs. 7 days; P < 0.001). Although the difference was not significant (P = 0.154), the rate of postoperative complications tended to be lower in the laparoscopic group (2.3%) compared with the open group (9.1%). The increase in postoperative SII, NLR, and LMR for laparoscopic hepatectomy were significantly lower than for open hepatectomy. NLR < 5.8 on postoperative day 3 was significantly correlated with shorter hospital stay (P < 0.001). CONCLUSIONS: Compared with open hepatectomy, laparoscopic hepatectomy for selected HCC patients, even in the presence of cirrhosis, might result in better perioperative outcomes and postoperative inflammatory response attenuation, and ultimately promote faster recovery. This provides evidence for considering routine laparoscopic hepatectomy through careful selection of patients with HCC.

Vortices as Brownian particles in turbulent flows.

Brownian motion of particles in fluid is the most common form of collective behavior in physical and biological systems. Here, we demonstrate through both experiment and numerical simulation that the movement of vortices in a rotating turbulent convective flow resembles that of inertial Brownian particles, i.e., they initially move ballistically and then diffusively after certain critical time. Moreover, the transition from ballistic to diffusive behaviors is direct, as predicted by Langevin, without first going through the hydrodynamic memory regime. The transitional timescale and the diffusivity of the vortices can be collapsed excellently onto a master curve for all explored parameters. In the spatial domain, however, the vortices exhibit organized structures, as if they are performing tethered random motion. Our results imply that the convective vortices have inertia-induced memory such that their short-term movement can be predicted and their motion can be well described in the framework of Brownian motions.

Identification of FOS as a Candidate Risk Gene for Liver Cancer by Integrated Bioinformatic Analysis.

Liver cancer is a lethal disease that is associated with poor prognosis. In order to identify the functionally important genes associated with liver cancer that may reveal novel therapeutic avenues, we performed integrated analysis to profile miRNA and mRNA expression levels for liver tumors compared to normal samples in The Cancer Genome Atlas (TCGA) database. We identified 405 differentially expressed genes and 233 differentially expressed miRNAs in tumor samples compared with controls. In addition, we also performed the pathway analysis and found that mitogen-activated protein kinases (MAPKs) and G-protein coupled receptor (GPCR) pathway were two of the top significant pathway nodes dysregulated in liver cancer. Furthermore, by examining these signaling networks, we discovered that FOS (Fos proto-oncogene, AP-1 transcription factor subunit), LAMC2 (laminin subunit gamma 2), and CALML3 (calmodulin like 3) were the most significant gene nodes with high degrees involved in liver cancer. The expression and disease prediction accuracy of FOS, LAMC2, CALML3, and their interacting miRNAs were further performed using a HCC cohort. Finally, we investigated the prognostic significance of FOS in another HCC cohort. Patients with higher FOS expression displayed significantly shorter time to recurrence (TTR) and overall survival (OS) compared with patients with lower expression. Collectively, our study demonstrates that FOS is a potential prognostic marker for liver cancer that may reveal a novel therapeutic avenue in this lethal disease.

Serum PON1 as a biomarker for the estimation of microvascular invasion in hepatocellular carcinoma.

BACKGROUND: Microvascular invasion (MVI) is considered as one of the most powerful prognostic factors in hepatocellular carcinoma (HCC). Currently, it could only be diagnosed by post-operative histological examination. This study aimed to assess the diagnostic value of serum paraoxonase 1 (PON1) for MVI. METHODS: In this study, we analyzed data from 754 HCC patients who underwent surgical treatment between December 2010 and November 2011. Serum PON1 was measured by ELISA and receiver operating characteristic (ROC) curve was applied to calculate diagnostic accuracy. RESULTS: MVI was detected in 174 of 505 patients (34.5%) in the test cohort and 84 of 249 patients (33.7%) in the validation cohort. Univariate analyses indicated tumor size, AFP, and PON1 were significantly related with vascular invasion status. ROC curves determined the optimum diagnostic cutoff value for PON1 was 191.12 ng/mL (AUC 0.754, 95% CI: 0.710-0.798, sensitivity 70.67%, specificity 78.11% in the test cohort), which was significantly better than AFP (cutoff value 279.8 ng/mL, AUC 0.666, 95% CI: 0.618-0.714, sensitivity 40.38%, specificity 85.19%, P=0.0063). In the sHCC sub-group, PON1 retained diagnostic value (AUC 0.738, 95% CI: 0.680-0.796, sensitivity 72.82%, specificity 76.57% in the test cohort), while AFP failed to do so (AUC 0.579, 95% CI: 0.511-0.647, sensitivity 26.21%, specificity 86.84%, P=0.0003). These results were further confirmed by the validation cohort. The combination of PON1 and AFP increased the diagnostic accuracy for vascular invasion compared with either test alone (AUC 0.785, 95% CI: 0.744-0.826, sensitivity 75.96%, specificity 77.44%; PON1 plus AFP vs. PON1 alone, P=0.0004; PON1 plus AFP vs. AFP alone, P<0.0001). CONCLUSIONS: Serum PON1 could potentially be used to diagnose MVI and could be used to guide more personalized treatment strategy.