Deletion of Cd44 Inhibits Metastasis Formation of Liver Cancer in Nf2-Mutant Mice

Primary liver cancer is the third leading cause of cancer-related death worldwide. An increasing body of evidence suggests that the Hippo tumor suppressor pathway plays a critical role in restricting cell proliferation and determining cell fate during physiological and pathological processes in the liver. Merlin (Moesin-Ezrin-Radixin-like protein) encoded by the NF2 (neurofibromatosis type 2) gene is an upstream regulator of the Hippo signaling pathway. Targeting of Merlin to the plasma membrane seems to be crucial for its major tumor-suppressive functions; this is facilitated by interactions with membrane-associated proteins, including CD44 (cluster of differentiation 44). Mutations within the CD44-binding domain of Merlin have been reported in many human cancers. This study evaluated the relative contribution of CD44- and Merlin-dependent processes to the development and progression of liver tumors. To this end, mice with a liver-specific deletion of the Nf2 gene were crossed with Cd44-knockout mice and subjected to extensive histological, biochemical and molecular analyses. In addition, cells were isolated from mutant livers and analyzed by in vitro assays. Deletion of Nf2 in the liver led to substantial liver enlargement and generation of hepatocellular carcinomas (HCCs), intrahepatic cholangiocarcinomas (iCCAs), as well as mixed hepatocellular cholangiocarcinomas. Whilst deletion of Cd44 had no influence on liver size or primary liver tumor development, it significantly inhibited metastasis formation in Nf2-mutant mice. CD44 upregulates expression of integrin β2 and promotes transendothelial migration of liver cancer cells, which may facilitate metastatic spreading. Overall, our results suggest that CD44 may be a promising target for intervening with metastatic spreading of liver cancer.


Introduction
As the third leading cause of cancer-related mortality (830,000 deaths worldwide in 2020), primary liver cancer represents a major public health challenge [1]. The most common histologic type of primary liver cancer, hepatocellular carcinoma (HCC), constitutes approximately 75-85% of primary liver cancer worldwide [1]. Intrahepatic cholangiocarcinoma (iCCA) is the second most common, comprising 10-15% of primary liver cancer cases [1]. Mixed hepatocellular cholangiocarcinoma, presenting histologic features of

Experiments Involving Animals
In total, 122 mice aged between 2 and 48 weeks were analyzed to investigate the influence of NF2 and CD44 on different stages of liver tumor development and progression. All animals used in this study were housed under constant temperature and humidity conditions, followed a 12-h light/dark cycle, and had ad libitum access to food and water. All experiments involving mice were approved by the local authorities (AZ: 23 177-07/G 16-1-028). Alb-Cre (B6.Cg-Tg(Alb-cre)21Mgn/J) mice were from The Jackson Laboratory (Sacramento, CA, USA). Generation and genotyping of Nf2 flox/flox transgenic mice (with targeted exon 2) was described by Giovannini et al. [41]. Generation of Cd44 flox/flox mice (with targeted exon 3) was described by Dhar et al. (2018) [42]. Generation and genotyping of Cd44 −/− mice was described by Ma et al. (2020) [43]. All mice were inbred onto C57BL/6 genetic background. Nf2 flox/flox ;Alb-Cre; and Cd44-deficient mice were crossed to obtain the desired phenotypes. Nf2 flox/flox ;Alb-Cre; mice were monitored weekly for occurrence of tumors or decline of health, then euthanized after two years at the latest and analyzed.

Histochemistry and Immunohistochemistry
Tissue samples were fixed in 4% paraformaldehyde (PFA) and embedded in paraffin. Next, 4 µm thick sections were deparaffinized, re-hydrated, and stained by hematoxylin and eosin (H&E) or were used for antigen detection by immunohistochemistry or immunofluorescence microscopy. Conventional H&E and Gomori silver staining were performed, following standard procedures, at the Institute of Pathology, Mainz. Detection of antigens was conducted on tissue sections after antigen-retrieval in the target retrieval solution, citrate pH 6 (DakoCytomation GmbH, Hamburg, Germany). Tissue sections were permeabilized in 0.1% Triton X-100 (Carl Roth GmbH, Karlsruhe, Germany) in phosphate-buffered saline (PBS, Life Technologies GmbH, Darmstadt, Germany) at room temperature for 7 min. Afterwards, tissue sections were washed with PBS. Endogenous peroxidases were quenched using 3% H 2 O 2 (Carl Roth GmbH, Karlsruhe, Germany) in methanol at room temperature for 10 min (this step was performed only for subsequent chemiluminescent detection). Tissue sections were again washed with PBS. Unspecific binding was blocked by pre-incubation with blocking solution consisting of 2% normal horse serum (Vector Laboratories, Newark, CA, USA), 1% bovine serum albumin (BSA, Sigma-Aldrich, Taufkirchen, Germany), and 0.01% Triton X-100 in PBS. Primary antibodies were diluted in blocking solution and applied on sections at 4 • C overnight. Primary antibodies used for immunohistochemistry are listed in Table 1. After vigorous washings with 0.01% Triton X-100 and 0.01% Tween20 in PBS, sections were incubated either with the fluorophore-conjugated secondary antibody diluted in blocking solution for 1 h, or with ZytoChem HRP One-Step Polymer anti-Mouse/Rabbit/Rat solution (Zytomed Systems GmbH, Berlin, Germany) at room temperature for 30 min. Afterwards, sections were again washed with 0.01% Triton X-100 and 0.01% Tween20 in PBS. For fluorescent detection, the sections were embedded in ProLong Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific GmbH, Darmstadt, Germany). Alternatively, chemiluminescent detection was performed using Liquid DAB+Substrate Chromogen System (DakoCytomation GmbH, Hamburg, Germany), following manufacturer's instructions. Afterwards, the nuclei were counterstained with hematoxylin (Merck KGaA, Darmstadt, Germany) for 20 sec, and the samples were then dehydrated and embedded in xylene-based mounting medium (Kindler GmbH & Co., Freiburg, Germany). Paraffin-embedded lung, liver, and tumor sections were evaluated by an expert pathologist (B.K.S).

TUNEL Assay
Click-iT ® Plus TUNEL Assay (catalog number C10618, Life Technologies GmbH, Darmstadt, Germany) was applied for in situ apoptosis detection, per manufacturer's protocol. To induce DNA strand breaks for positive control, fixed and permeabilized Alb-Cre liver sections were incubated with 1 unit of Dnase I (Qiagen, Hilden, Germany) diluted in 1xDNase I Reaction Buffer (20 mM Tris-HCl, pH 8.4, 2 mM MgCl2, 50 mM KCl), at room temperature for 30 min.

Clonogenic Assay under Cell Attachment Conditions
Liver cells were seeded in triplicates at a density of 100 and 1,000 cells per well in a 6-well plate, in DMEM supplemented with 10% FBS. Medium was changed every 3 days. After 1 week, the colonies were fixed in 4% PFA for 5 min and stained with 0.05% crystal violet in PBS for 30 min. Photographs of colonies were taken on a ChemiDoc MP Imaging System (Bio-Rad Laboratories GmbH, Feldkirchen, Germany), and the number of colonies was counted using Cell Counter software v0.2.1 (by Nghia Ho).

Colony Growth under Nonadherent Conditions/Anoikis Assay
First, 96-well plates were made anchorage-resistant by coating them with a 1% agar layer. Liver cells were seeded in triplicates at a density of 0.2 × 10 5 cells per well on an anchorage-resistant 96-well plate in DMEM supplemented either with 10% FBS or without FBS. After 7 days of culture, the colonies were photographed using a Leica DFC290 microscope, and the colony surface area was measured using ImageJ software (National Institutes of Health (NIH), Bethesda, MD, USA).

Scratch Migration Assay
Liver cell lines were grown on 6-well plates in DMEM, supplemented with 10% FBS. Scratches were generated in confluent cell monolayers with a 200 µL pipette tip, and cell debris was removed by washing with PBS. Afterwards, the cells were allowed to migrate into the generated gap in serum-free medium for 72 h. To inhibit proliferation, cytosine b-D-arabinofuranoside (AraC, Sigma-Aldrich, Taufkirchen, Germany) was added at a concentration of 10 µM prior to migration assay. The cells were photographed using a Leica DFC290 microscope (Leica Microsystems CMS GmbH, Mannheim, Germany), and the cell-free gap was measured using ImageJ software.

Tumor Cell Transendothelial Migration Assay
The ability of tumor cells to migrate through a confluent monolayer of endothelial cells towards a chemoattractant was tested using commercially available QCM tumor cell transendothelial migration assay (catalog number ECM558, Merck KGaA, Darmstadt, Germany), per manufacturer's protocol. Briefly, tumor cells were allowed to migrate at 37 • C for 23 h through a confluent monolayer of tumor necrosis factor-α (TNFα, Thermo Fisher Scientific GmbH, Darmstadt, Germany) -stimulated C57BL/6 mouse primary lung microvascular endothelial cells, from medium without serum towards medium supplemented with 10% FBS.

Western Blot
Cells and liver tissues were harvested in RIPA (50 mM Tris HCl pH 8.0; 150 mM NaCl; 1% NP40; 0.1% SDS; 1 mM EDTA) and M-PER (Thermo Fisher Scientific GmbH, Darmstadt, Germany) lysis buffers, respectively, containing a halt protease and phosphatase inhibitor cocktail. Protein concentration was determined using the bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific GmbH, Darmstadt, Germany). Lysates were mixed with SDS-PAGE sample buffer (final concentration: 50 mM Tris HCl, pH 6.8; 2% SDS; 10% glycerol; 0.1% bromphenol; 100mM dithiotreitol) and denatured by boiling at 97 • C for 7 min. Proteins were separated on 8% acrylamide gels using SDS-PAGE and transferred to nitrocellulose membranes. Blocking was performed at room temperature for 1 h, using 5% low-fat milk powder in Tris-buffered saline (TBS) containing 0.1% Tween20 solution (TBS-T). Membranes were incubated with primary antibody at 4 • C overnight, followed by HRPlabeled secondary antibody at room temperature for 1 h. Primary and secondary antibodies used for immunoblot analysis are listed in Tables 1 and 2. A total of 5% BSA in TBS-T was used as diluent for CD44 antibody, clone E7K2Y. All other primary and secondary antibodies were diluted in 5% low-fat milk powder in TBS-T. Blots were developed using Clarity Western ECL Substrate (Bio-Rad Laboratories GmbH, Feldkirchen, Germany) and visualized with Chemi-Doc MP Imaging System (Bio-Rad Laboratories GmbH, Feldkirchen, Germany). The intensity of immunoblot bands was quantified with ImageJ software.

Immunocytochemistry
Liver cells grown on coverslips were fixed with 4% PFA in PBS for 20 min, washed in PBS, and permeabilized with 0.1% Triton X-100 in PBS for 7 min. Non-specific epitopes were blocked with blocking solution (1% BSA in PBS) for 45 min. Cells were probed in primary antibody solution at room temperature for 1 h, washed extensively with PBS, and incubated with fluorophore-conjugated secondary antibody solution at room temperature for 1 h. After additional washing with PBS, the cells were mounted on cover slides with a ProLong™ Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific GmbH, Darmstadt, Germany). Primary and secondary antibodies used for immunofluorescence (IF) detection are listed in Table 1. A total of 1% BSA in PBS was used as diluent for all primary and secondary antibodies. For negative controls, primary antibody was omitted.

Antibodies
Tables 1 and 2 list antibodies used in the study.

RNA Sequencing and Gene Expression Analysis
Liver tissues were derived from Cd44 +/+ ;Nf2 ∆2/+ and Cd44 -/-;Nf2 ∆2/+ mice described previously [43]. Three samples of each group were analyzed. Total RNA was isolated using RNeasy mini kit (Qiagen), per manufacturer's instructions. Sequencing of RNA samples was achieved using Illumina's next-generation sequencing methodology [44]. The quality check and quantification of total RNA was completed using the Agilent Bioanalyzer 2100 in combination with the RNA 6000 nano kit (Agilent Technologies). Library preparation required Illumina's TruSeq stranded mRNA library preparation kit, per manufacturer's description. Agilent Bioanalyzer 2100 in combination with the DNA 7500 kit enabled quantification and quality check of libraries. Said libraries were sequenced in two lanes on a HiSeq2500 running in 51 cycle/single-end/high-output mode. Sequence information was extracted in FastQ format using Illumina's bcl2fastq v.1.8.4 and resulted in an average of around 24 million reads per sample. Reads were aligned to the mouse reference genome sequence (ENSEMBL Mus musculus GRCm38.84) using HISAT2 (hisat2-2.0.2-beta), with read counts determined by read summarization with feature Counts (subread-2.0.0). R programming language and related packages were engaged to further analyze RNA sequencing data. Differentially regulated genes were determined using the rlog function of the DESeq2 package. All genes with Benjamin Hochberg adjusted p-value (padj) < 0.05 were considered significant.

Real-Time Quantitative Reverse Transcription PCR (Real-Time qRT-PCR)
Total RNA was isolated from mouse tissues or cell lines using PeqGOLD total RNA kit (VWR International GmbH, Darmstadt, Germany), per manufacturer's instructions. cDNA was prepared from 1 µg of total RNA using iScript cDNA synthesis kit (Bio-Rad Laboratories GmbH, Feldkirchen, Germany) and used for qRT-PCR reaction at a final dilution of 1:75. We used iQ TM SYBR green supermix (Bio-Rad Laboratories GmbH, Feldkirchen, Germany) on CFX Connect™ real-time PCR detection system (Bio-Rad Laboratories GmbH, Feldkirchen, Germany) for PCR amplification. Oligonucleotides were used at a final concentration of 157 nM. Tbp housekeeping gene was used as the reference for testing efficiency of Icam1 knockdown, with Hprt1 and Actin housekeeping genes used as references for all other real-time qRT-PCRs. Total Cd44 was detected using primers that recognized constant regions of all isoforms (Cd44all Fwd1 and Cd44all Rev1). CD44 isoforms containing variant exon sequences were amplified using forward oligonucleotides specific for stable exon 5 (stable Ex5 Fwd 1) and reverse oligonucleotides specific for appropriate variant exon. Oligonucleotides used for real-time qRT-PCR are listed in Table 3. Relative expression values were calculated using REST-MCS software [45]. siRNAs targeting mouse Icam1 (ID:s67995 and s67994), mouse Cd44 (ID:s63660 and s63661), and negative control siRNA No.1 were acquired from Thermo Fisher Scientific GmbH (Darmstadt, Germany). All siRNA transfections were performed in 6-well plates using the liposomal transfection reagent Lipofectamine 2000 (Thermo Fisher Scientific GmbH, Darmstadt, Germany), per manufacturer's instructions.

Statistical Analysis
The Chi-Square test of independence determined whether there is a significant association between the presence or absence of metastases in Cd44-positive and Cd44-negative livers; and presence or absence of variant Cd44 isoforms in liver tumors and control livers. For all quantitative analyses, comparisons between groups were made with unpaired Student's t-test unless otherwise stated. Two-tailed p-value ≤ 0.05 was considered as significant (ns: p > 0.05, *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001, ****: p ≤ 0.0001).

Deletion of Cd44 Does Not Influence Progressive Liver Enlargement in Nf2 flox/flox ;Alb-Cre Mice
We used Nf2-mutant mice to model liver tumor development and progression. Deficiency of Nf2 was confined to the liver by crossing conditional Nf2-mutant (Nf2 flox/flox ) mice with albumin-Cre (Alb-Cre) transgenic mice. To investigate the role of CD44 in liver tumor development and progression, Nf2 flox/flox ;Alb-Cre mice were crossed with Cd44-mutant mice, either Cd44 flox/flox , to obtain deletion of Cd44 exclusively in the liver, or Cd44 −/− , resulting in deletion of Cd44 in the entire body. A total of 122 mice aged between 2 and 48 weeks were analyzed to investigate the influence of NF2 and CD44 on different stages of liver tumor development and progression. Alb-Cre mice were the control group. Table 4 lists the number of mice analyzed at different ages. Table 4. Number of analyzed mice. "wk-old" stands for week-old.

Genotype
Age The presence of the desired genetic modification was confirmed by PCR-based genotyping ( Figure S1). As expected, recombinant deleted alleles of Cd44 and Nf2 were only detected in livers of conditional knockout mice ( Figure S1). In contrast, Cd44 −/− mice exhibited allele deletion in all analyzed tissues ( Figure S1). Presence of a faint floxed band of Cd44 and Nf2 gene in the liver of conditional knockout mice may indicate non-complete recombination of floxed alleles or infiltration by non-hepatic cells, which is consistent with previous reports on Alb-Cre-mediated excision [6,46,47]. The Nf2 flox/flox ;Alb-Cre mice were born at expected Mendelian rations and appeared healthy. The most prominent difference from control animals was progressive liver enlargement observed in all analyzed Nf2 flox/flox ;Alb-Cre mice, independent of the Cd44 gene status (Figure 1a,b). It should be mentioned that deletion of the Cd44 gene alone has no influence on normal liver phenotype [43,48,49]. The liver weight/body weight ratio of Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre animals increased from 6.94% at 20 weeks to 16.06% at 48 weeks, whereas the percent ratio of liver weight/body weight of control animals remained relatively constant (5.52% at 20 weeks and 5.46% at 48 weeks) during the analysis period. In addition, livers isolated from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre, Cd44 −/− ;Nf2 flox/flox ;Alb-Cre and Cd44 flox/flox ;Nf2 flox/flox ;Alb-Cre mice appeared paler and more solid than control livers and harbored tumors at later ages (32-and 48-weeks) (Figure 1b).  Table 4. "wk-old" stands for week-old. We tested liver biomarkers in the serum to analyze the function of enlarged livers (Figure 1c,d), but found no significant difference in serum levels of bilirubin ( Figure 1c) or alanine aminotransferase (ALT, Figure 1d) between control and Nf2 flox/flox ;Alb-Cre mice, which indicates normal function of biliary system and no significant hepatocyte injury, respectively. Thus, deletion of the Nf2 gene in the liver and concomitant liver enlargement is not associated with impairment of liver function. Consistently, during this analysis (maximally 48 weeks) we observed no increased mortality in Nf2 flox/flox ;Alb-Cre mice compared to control group. Nor did we observe any significant differences in the appearance or size of the liver between Cd44-positive and Cd44-negative Nf2 flox/flox ;Alb-Cre mice. Thus, loss of the Cd44 gene does not influence the liver overgrowth observed in Nf2-mutant mice.

Ductular Proliferation Observed in Nf2 flox/flox ;Alb-Cre Mice Is Not Influenced by Loss of Cd44
To investigate the Nf2-dependent histopathological changes, livers isolated from Moderate ductular reaction in the portal tract areas was already found in 2-and 6-weekold Nf2-mutant mice and progressed in the 20-to 32-week-old cohort, ultimately leading to a massive tumor-like ductular process radiating from and bridging portal areas in livers of 48-week-old Nf2-mutant mice (Figures 2 and S2). As revealed by Gomori silver staining, livers of 48-week-old mice exhibited progressive fibrosis, visible as bridges of collagen that connected individual portal tracts and entrapped islands of hepatocytes, with pseudolobule formation and distortion of liver architecture, indicative of cirrhosis ( Figure S3). Comparable pathological changes were observed in livers isolated from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre, Cd44 −/− ;Nf2 flox/flox ;Alb-Cre, and Cd44 flox/flox ;Nf2 flox/flox ;Alb-Cre mice (Figures 2, S2 and S3), suggesting that the status of the Cd44 gene has no influence on ductular proliferation or liver fibrosis.
In younger mice, ductular reaction and fibrosis were most pronounced in subcapsular areas (Figures 2 and S2, asterisks). These regions showed increased cell turnover when compared to central regions containing hepatocytes ( Figure 3). Both the rate of cell death (Figure 3a,c, measured by TUNEL assay) and cell proliferation (Figure 3b,d, indicated by positive Ki67 staining) were increased three-to four-fold in subcapsular areas when compared to central areas.
There were no significant differences in cell proliferation or apoptosis between Cd44positive and Cd44-negative Nf2-mutant livers ( Figure 3).

CD44 Exhibits Similar Expression Pattern to Ductular/Liver Progenitor Cell Markers, but Is Not Required for Ductular Process in Nf2 flox/flox ;Alb-Cre Mice
To further characterize ductular proliferations observed in Nf2-mutant mice, ductular/liver progenitor cell markers cytokeratin 19 (CK19) and transcription factor SOX9 were analyzed. In addition, hepatocyte nuclear factor 4 alpha (HNF4α) was detected to distinguish hepatocytes, and Ki67 staining was performed to identify proliferating cells. At later ages, the ductular process was distributed throughout the portal and periportal areas of Nf2-mutant livers. Ki67-staining highlighted an increased rate of proliferation in ductular reactions from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre and Cd44 −/− ;Nf2 flox/flox ;Alb-Cre livers (Figures 4a and S4a). Patches of proliferating ductular reactions radiated from portal tracts, whereas hepatocytes were largely Ki67-negative (Figures 4a and S4a). HNF4α showed positive nuclear staining in hepatocytes but was absent in ductular reactions (Figures 4b and S4b). Instead, tumor-like ductular proliferations were positive for CK19 and SOX9 staining (Figures 4c,d and S4c,d). Figures 4c,d and S4c,d demonstrate the comparably marked CK19 and SOX9 staining of cholangiocytes of ductular reaction/oval cell proliferation highlighting fibrotic septae between portal tracts and the formation of pseudolobuli in Nf2-mutant livers; in the Alb-Cre control livers, only the originary bile ducts expressed CK19 and SOX9 antigens and no ductular proliferation was noted. Despite some regional differences in intensity and thickness of CK19 and SOX9 -positive cell layer, we did not observe any significant differences between Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre and Cd44 −/− ;Nf2 flox/flox ;Alb-Cre livers (Figures 4c,d and S4c,d).
compared to control group. Nor did we observe any significant differences in th pearance or size of the liver between Cd44-positive and Cd44-negative Nf2 flox/flox ;Al mice. Thus, loss of the Cd44 gene does not influence the liver overgrowth observ Nf2-mutant mice. Moderate ductular reaction in the portal tract areas was already found in 2-6-week-old Nf2-mutant mice and progressed in the 20-to 32-week-old cohort, ultim leading to a massive tumor-like ductular process radiating from and bridging port   Furthermore, we investigated localization of CD44 in mutant livers. Immunohistochemical analysis (Figures 4e and S4e) and immunofluorescent staining ( Figure S5) revealed that CD44 is expressed in bile ducts and ductular proliferations, but not in hepatocytes ( Figure 4e). As expected, Cd44 −/− -livers were negative for CD44 staining (Figures 4e, S4e and S5). Together, our results suggest that despite a similar expression pattern to ductular/liver progenitor cell markers, Cd44 is not required for ductular proliferation in Nf2 flox/flox ;Alb-Cre mice.

YAP Signaling Is Elevated in Nf2-Deficient Livers
NF2/Merlin acts as an upstream regulator of the Hippo signaling tumor suppressor pathway that inhibits oncogenic YAP activity (Figure 5b). Accordingly, Nf2-deficient livers exhibited enhanced staining of YAP and increased transcriptional activity of YAP target genes when compared to control mice ( Figure 5). YAP was predominantly localized in the ductular reactions of Nf2 flox/flox ;Alb-Cre mice, whereas hepatocytes were largely negative for YAP staining (Figure 5a). Expression of YAP target genes Ccn1 and Ccn2 increased more than tenfold in livers isolated from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre mice compared to those isolated from Alb-Cre mice (Figure 5c). Transcriptional activity of YAP was even higher in Cd44-negative Nf2-mutant livers compared to Cd44-positive livers. However, these effects were only significant for the Ccn2 gene. Increased activity of YAP oncogene is compatible with its role in mediating tumorigenic progression in Nf2-mutant livers. The effects of Cd44 deletion on expression of YAP target genes are difficult to interpret given the high variations in Ccn1 and Ccn2 mRNA levels between single Cd44-individuals.  [43]. CD44 provides a link between Merlin and the plasma membrane. Merlin directly binds and recruits the LATS1/2 kinase to the plasma membrane, prompting its phosphorylation by the MST1/2 kinase bound to adapter protein SAV1 [50]. In turn, LATS1/2, in a complex with small regulator protein MOB1A/1B, phosphorylates and inactivates YAP/TAZ, blocking their role as TEAD/MEAD transcription factor co-activators [50].   [43]. CD44 provides a link between Merlin and the plasma membrane. Merlin directly binds and recruits the LATS1/2 kinase to the plasma membrane, prompting its phosphorylation by the MST1/2 kinase bound to adapter protein SAV1 [50]. In turn, LATS1/2, in a complex with small regulator protein MOB1A/1B, phosphorylates and inactivates YAP/TAZ, blocking their role as TEAD/MEAD transcription factor co-activators [50].

Deletion of Cd44 Does Not Influence Primary Liver Tumor Generation but Virtually Aborts Metastasis Formation in Nf2 flox/flox ;Alb-Cre Mice
Furthermore, we investigated the influence of CD44 on liver tumor development and progression in Nf2-mutant mice. We observed that nearly all Nf2 flox/flox ;Alb-Cre mice older than 32 weeks, whether Cd44-positive or Cd44-negative, developed frank liver tumors, and primarily HCCs. However, transitional cellular and histological features that were intermediate between HCCs and iCCAs were also frequently observed. The former exhibited a striking variety of histopathological types, including macrotrabecular and pseudoglandular morphologies ( Figure S6). However, we did not observe any significant differences in the spectrum of tumors detected in Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre, Cd44 −/− ;Nf2 flox/flox ;Alb-Cre, and Cd44 flox/flox ;Nf2 flox/flox ;Alb-Cre mice.
The level of total Cd44 mRNA was significantly increased in liver tumors isolated from 48-week-old Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre mice compared to age-matched control livers isolated from Alb-Cre mice ( Figure 6a). As expected, Cd44 −/− tumors did not express relevant Cd44 mRNA levels ( Figure 6a). Moreover, nearly all livers isolated from 48-week-old Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre mice were positive for expression of exon v3-, v6-, and v7-bearing CD44 variant isoforms that have previously been implicated in cancer (Figure 6b). Livers isolated from Alb-Cre mice were applied as controls, as it was not possible to identify normal liver areas in 48-week-old Nf2 flox/flox ;Alb-Cre animals given the extensive tumorlike ductular process. The increased expression of total Cd44 mRNA and presence of Cd44 variant isoforms, including exon v3, v6, and v7 sequences in liver tumors from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre mice, is compatible with the function of CD44 and its isoforms in the formation of primary liver tumors and/or metastases. However, the Cd44 gene status had no influence on either the number of primary liver tumors detected or on the level of the hepatocellular carcinoma biomarker AFP (α-fetoprotein) in the serum of Nf2-mutant mice (Figure 6c,d). In sum, our results suggest that CD44 does not influence primary liver tumor generation in Nf2-mutant mice.
Next, we were interested in the role of CD44 in liver tumor progression. Liver metastases were detected in the lungs of four of the nine analyzed 48-week-old Cd44 +/+ ;Nf2 flox/flox ; Alb-Cre mice (Figure 6e). These included two HCCs, one iCCA, and one mixed HCC-iCCA metastasis. In contrast, no metastases were found in nine of the analyzed Cd44 −/− ;Nf2 flox/flox ; Alb-Cre or five of the analyzed Cd44 flox/flox ;Nf2 flox/flox ;Alb-Cre mice. No extrahepatic lesions other than lung metastases were detected. Together, our results indicate that CD44 is not required either for ductular proliferation or for generation of HCCs or the other primary liver tumors observed in Nf2 flox/flox ;Alb-Cre mice. Instead, we identified a critical role of CD44 in the generation of liver tumor metastases in the lung.

Deletion of the Cd44 Gene Does Not Influence Colony-Forming Ability or Migration into Cell-Free Gap of Nf2-Negative Liver Cells
Liver tumor cells were isolated from Nf2-mutant mice to further investigate the role of CD44 in liver tumor development and progression. Subsequently, the cells isolated from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre, and Cd44 −/− ;Nf2 flox/flox ;Alb-Cre livers will be identified as Cd44 +/+ and Cd44 −/− , respectively. As expected, all cell lines generated from Nf2 flox/flox ;Alb-Cre livers were negative for NF2/Merlin expression (Figure 7a). Because Alb-Cre-dependent recombination of Nf2 flox alleles is restricted to liver cells, these results indicate a lack of contamination from other cell types. Through immunoblot analysis, we also confirmed that only Cd44 +/+ cells, and not Cd44 −/− cells, were positive for CD44 (Figure 7a). The isolated cells also did not express hepatocyte marker HNF4α, but they were positive for ductular/liver progenitor cell markers CK19 and SOX9 ( Figure S7). Although there were cell line-dependent variations in CK19 and SOX9 expression levels, we observed no significant differences in the expression and subcellular localization of CK19 and SOX9 between Cd44-positive and Cd44-negative cell lines ( Figure S7b-d). Similarly, levels of phosphorylated LATS and YAP were highly variable between analyzed cell lines; however, these differences were independent of the Cd44 gene status, suggesting that, in the absence of Merlin, CD44 has no influence on Hippo signaling ( Figure S8).  Three independent Cd44 +/+ and two-three independent Cd44 −/− cell lines were tested. NIH3T3 cells were applied as a positive control for detection of Merlin. The cells were seeded at 50% confluency in DMEM medium supplemented with 10% FBS and left overnight to attach, then subjected to immunoblot. Merlin was detected using anti-Merlin antibody, clone EPR2573(2). CD44 was detected using anti-CD44 antibody, clone E7K2Y. GAPDH was detected to control equal loading of samples. After confirming that the isolated cells exhibit expected expression pattern of Nf2, Cd44, and characteristic liver markers, further functional assays were conducted. At first, the self-renewing capability of Cd44-positive and Cd44-negative cells derived from Nf2 flox/flox ;Alb-Cre livers was compared in a clonogenic assay. Deletion of the Cd44 gene did not influence the capability of Nf2-negative liver cells to form colonies under adherent (Figure 7b,c) or nonadherent conditions (Figure 7d,e). Liver cells grown on attachmentresistant plates formed one single spherical colony per well of a 96-well plate (Figure 7d,e). The size of generated spherical colonies grown in medium enriched with 10% FBS was larger compared to those generated in serum-deprived medium (Figure 7d,e). However, we did not observe any significant differences between Cd44-positive and Cd44-negative colonies (Figure 7d,e). Colony-forming assays are indicative of tumor-initiating potential of investigated cells [51]. Thus, our in vitro results suggest that Cd44 +/+ and Cd44 −/− cells do not differ in their tumor-initiating capability. This observation is compatible with our in vivo results demonstrating no difference in tumor development or growth between Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre and Cd44 −/− ;Nf2 flox/flox ;Alb-Cre mice (Figure 6c,d).
Attachment-independent growth is also a prerequisite for metastatic spread of tumor cells. A large percentage of tumor cells that enter the bloodstream or the lymph during metastatic dissemination die due to loss of cell-cell and cell-matrix contacts, undergoing a specific type of apoptosis known as "anoikis" [52]. Our results suggest that CD44 does not influence the ability of liver cells to evade anoikis (Figure 7d,e). Migration is another prerequisite for metastasis formation. To test the influence of CD44 on cell migration in vitro, a conventional scratch migration assay was conducted. CD44 had no influence on the ability of isolated liver cells on migration into an introduced cell-free gap (Figure 7f,g). Thus, the inhibitory effect of the Cd44 gene deletion on metastasis formation in vivo (Figure 6e,f) cannot be explained by the influence on anoikis or simple migration.

CD44 Contributes to Transendothelial Migration of Nf2-Negative Liver Cells by Upregulating Levels of Integrin Subunit Beta 2
An important hint about the potential mechanism underlying CD44-dependent metastasis formation was provided by our previous study [43]. We performed RNA-sequencing analysis on Cd44-positive and Cd44-negative livers isolated from Nf2-heterozygous mice (Figure 8a,b). We identified that several genes involved in transendothelial migration were among the top-ranked genes differentially regulated between Cd44-positive and Cd44negative livers (Figure 8c). The full list of differentially regulated genes can be found in Table S1. Adhesion of tumor cells to endothelial cells and extravasation are important rate-limiting steps in the metastatic cascade. Integrin αLβ2 (LFA-1, lymphocyte functionassociated antigen 1) and integrin α4β1 (VLA-4, very late antigen-4) are two major integrins implicated in transendothelial migration of leukocytes and tumor cells [53,54]. Interestingly, expression of genes encoding integrin subunit alpha L (ItgaL) and integrin subunit beta 2 (Itgb2), as well as the substrate of integrin α4β1, VCAM-1 (vascular cell adhesion molecule-1), were significantly downregulated in Cd44-negative livers compared to Cd44-positive counterparts (Figure 8c). These results strongly indicate a role of CD44 in regulating the major effectors of transendothelial migration of tumor cells.
Real-time qRT-PCR facilitated confirmation that the livers isolated from Cd44 +/+ ; Nf2 flox/flox ;Alb-Cre animals express significantly higher levels of Itgb2 and Vcam1, compared to Cd44 −/− ; Nf2 flox/flox ;Alb-Cre livers, whereas the levels of mRNAs of ItgaL, Itga4 and Itgb1 were not significantly different (Figure 8d). These results are consistent with the RNA sequencing data, although there is a lack of differences in ItgaL expression levels between Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre and Cd44 −/− ; Nf2 flox/flox ;Alb-Cre animals. The observed discrepancy could depend upon the presence of frank liver tumors in Nf2 flox/flox ;Alb-Cre mice, but not in the analyzed Nf2-heterozygous mice. For instance, it is possible that both Cd44-positive and Cd44-negative tumors generate high levels of ItgaL, and in this case, the level of Itgb2 would be the limiting factor in the generation of functional integrin αLβ2 (LFA-1).  Liver cells isolated from Cd44 −/− ;Nf2 flox/flox ;Alb-Cre animals showed significantly reduced efficiency of transendothelial migration in vitro compared with those isolated from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre mice (Figure 8e,g). Moreover, CD44-specific antibody significantly blocked transendothelial migration of Cd44-positive liver cells (Figure 8f,g). These results indicate that CD44 may promote metastatic dissemination of liver tumor cells by facilitating transendothelial migration through upregulation of Itgb2 and Vcam1 expression.

ICAM-1 May Overtake CD44 Function as a Coreceptor of Receptor Tyrosine Kinase MET in Cd44-Negative Liver Cells
Considering that Merlin inhibits putative tumor-promoting functions of CD44 (as discussed in the introduction), we expected that deletion of Nf2 would lead to more a severe phenotype in Cd44-positive mice when compared to Cd44-negative equivalents. This proved not to be the case; the Cd44 gene status had no influence on the development of primary liver tumors in Nf2 flox/flox ;Alb-Cre mice, suggesting that deletion of Nf2 leads to tumor-promoting mechanisms independent of Cd44. Alternatively, we considered that compensatory molecules may overtake CD44 function in Cd44-knockout mice. It has been demonstrated that ICAM-1 overtakes CD44 function as a coreceptor of receptor tyrosine kinase MET during liver regeneration [55].
To test the possibility that ICAM-1 may compensate for CD44 loss during tumorigenesis in Nf2 flox/flox ;Alb-Cre mice, we first investigated Icam1 expression in the livers of Nf2 flox/flox ;Alb-Cre mice. Interestingly, Cd44 −/− livers showed significantly increased levels of Icam1 mRNA compared to Cd44 +/+ livers, suggesting a compensatory mechanism (Figure 9a). Next, we tested the effect of Cd44 and Icam1 downregulation on HGF/MET signaling in liver cell lines isolated from Nf2-mutant mice. Cd44 and Icam1 expression was downregulated using small interfering RNA (siRNA). As readout for HGF/MET signaling, phosphorylated ERK was detected (see also diagram in Figure 9b). The levels of phosphorylated MET and ERK were strongly elevated in liver cells treated with 20 ng/mL of HGF (Figure 9c,d). In Cd44 +/+ liver cells, the activation of MET was strictly dependent on CD44 as the levels of phosphorylated ERK were inhibited by Cd44-knockdown ( Figure 9c). As expected, siRNA targeting Cd44 had no influence on MET-ERK signaling in Cd44-negative liver cells (Figure 9c). Interestingly, Icam1 knockdown inhibited ERK phosphorylation only in Cd44-negative liver cells, whereas it had no effect on MET/ERK signaling in Cd44positive liver cells (Figure 9d). These results suggest that CD44 is sufficient for MET activation in the absence of ICAM-1. Nevertheless, ICAM-1 can overtake CD44 function as a coreceptor of MET in Cd44-knockout cells. This could explain why Cd44 knockout in the liver of Nf2 flox/flox ;Alb-Cre mice led to milder effects on tumor development and growth than expected based on previous reports [32][33][34][35][36].

Deletion of Nf2 in the Liver Leads to CD44-Independent Liver Enlargement
This study analyzed the influence of CD44 on the phenotype of mice with a Nf2 deletion in the liver. To study the relative contribution of CD44-and Merlin-dependent signaling pathways on liver tumor development and progression, Cd44-deficient mice were crossed with Nf2 flox/flox ;Alb-Cre mice. Our results demonstrate that deletion of the Cd44 gene had no effect on liver size or primary tumor development in Nf2-mutant mice. Instead, CD44 was required for metastatic spreading of Nf2-deficient liver tumors. CD44 facilitates transendothelial migration of liver cancer cells by promoting expression of integrin subunit beta 2 (Itgb2). Moreover, our results indicate that ICAM-1 may substitute for CD44 function as a coreceptor of receptor tyrosine kinase MET, and thus may compensate for CD44 loss during other tumor-relevant processes in Nf2-deficient livers.
Nf2 flox/flox ;Alb-Cre mice showed progressive liver enlargement and developed HCCs, iCCAs, and tumors of mixed histology, often within the same liver ( Figure 1). These results are consistent with previous reports demonstrating that deletion of the Nf2 gene in the liver [5,6]-or genetic inactivation of other members of the Hippo signaling pathway Mst1/2 [7][8][9], Lats1/2 [10,11], Mob1a/b [12], or Sav1 [9,13]-leads to liver overgrowth and promotes tumorigenesis. Nevertheless, the phenotype of our Nf2 flox/flox ;Alb-Cre mice was milder overall than previously reported [5,6]. In contrast to Benhamouche et al. [6], we did not observe any significant effects of Nf2-deletion on liver function or animal survival. The penetrance of Nf2-deficient liver phenotype could possibly be influenced by the genetic background of applied mice; C57BL/6 in our study, FVB/N in the study of Benhamouche et al. [6], and not specified in the study of Zhang et al. [5]. Association between genetic background and penetrance of a phenotype of genetically engineered mice is very well documented [56,57]. Other factors, including the strategy to impair Merlin function, were exactly the same in our study and in those of Benhamouche et al. [6] and Zhang et al. [5]; thus, the recombination efficiency of Nf2 flox alleles should also be comparable.

Origin of Active Cells during Ductular Reaction Observed in Nf2 flox/flox ;Alb-Cre Mice
The liver overgrowth observed in Nf2 flox/flox ;Alb-Cre mice was due to ductular proliferation of bile ducts in the portal tracts. We observed the appearance of small overproliferating cells that were positive for ductular/liver progenitor cell markers CK19 and SOX9 (Figure 4c,d and Figure S4c,d); however, the origin of these cells is unclear. The origin of active cells during ductular reaction could be cholangiocytes, hepatocytes, or hepatic progenitor cells [58]. The albumin promoter becomes active in embryonic hepatoblasts that give rise to both hepatocytes and cholangiocytes, and thus, recombination of floxed alleles in mice carrying the Alb-Cre transgene is expected to affect all major types of liver cells [6,46,47]. Previous studies reached contradictory conclusions about the origin of proliferating cells appearing in the livers of Nf2 flox/flox ;Alb-Cre mice. Based on the positive anti-pan-cytokeratin, A6, and anti-CD34 staining and morphological features characteristic for oval cells, Benhamouche et al. [6] concluded that overproliferating cells appearing in Nf2-deficient livers originated from putative liver progenitor cells. To support this hypothesis, they generated Nf2-deficient embryonic liver progenitor cells (hepatoblasts) and showed that, upon injection into the livers of immunocompromised mice, multiple foci of proliferating undifferentiated cells were generated. These cells progressed within 3-4 weeks to form larger neoplasia that exhibited both cholangiocytic and hepatocytic features [6]. In contrast, the injected control Nf2-positive hepatoblasts were scattered uniformly throughout the liver and did not form neoplasms [6]. Benhamouche et al. [6] concluded that disruption of the Nf2 gene is sufficient to affect a primary expansion of both embryonic and adult liver progenitor cells in vivo and to reproducibly yield both major forms of primary liver cancer. By employing a combination of lineage tracing and clonal analysis, Yimlamai et al. [59] reached a different conclusion. They discovered that deletion of Nf2 or overexpression of active YAP form (S127A), specifically in adult hepatocytes, using AAV-Cre, results in their dedifferentiation, driving liver overgrowth and appearance of ductal cells bearing characteristics of hepatic progenitors [59]. Overexpression of active YAP form in the biliary/progenitor compartment, on the other hand, led to hyperplasia, but did not result in an oval-cell-like appearance [59]. While Benhamouche et al. [6] interpreted their findings as liver progenitor-cell expansion and transformation driven by loss of Hippo signaling, those of Yimlamai et al. [59] suggest that hepatocytes might be the source of active cells during the ductular reaction found in Nf2 flox/flox ;Alb-Cre mice. Considering the findings of Yimlamai et al. [59] and the fact that hepatocytes continue to express Alb-Cre postnatally, it is plausible to assume that the overproliferating cells observed in Nf2-deficient livers originate from dedifferentiated hepatocytes [6,46,47]. Yet the question remains regarding how Nf2 deficiency induces dedifferentiation of some hepatocytes, leaving remaining hepatocytes unchanged. We observed that, despite Nf2 deficiency, the livers of Nf2 flox/flox ;Alb-Cre mice also contained differentiated hepatocytes that did not exhibit gross morphological alterations or aberrant Ki67 expression in vivo. A slight increase in apoptosis was observed in subcapsular areas, consistent with atrophy of liver parenchyma in these regions (Figure 3a,c). Moreover, Benhamouche et al. [6] observed that Nf2 deletion in adult mouse liver leads to very mild periporal hyperplasia, compared to dramatic and widespread ductular proliferation in Alb-Cre-mediated Nf2 deletion [6]. Therefore, deletion of Nf2 alone is not sufficient to reprogram liver cells. They showed that additional proliferative stimuli in the form of partial hepatectomy were required to induce emergence of overproliferating progenitor-like cells and tumor formation in mice with postnatal Nf2 deletion [6]. Importantly, liver regeneration occurred normally in the absence of Merlin, indicating that, even after receiving a proliferative stimulus and undergoing cell division, Nf2 −/− hepatocytes can appropriately re-enter a quiescent state [6]. Thus, the exact interplay between signaling components that determine hepatocyte dedifferentiation and acquisition of mature phenotype remains to be determined. To mediate its major tumor suppressor functions, Merlin must be bound to the plasma membrane. This association occurs indirectly through interactions with plasma membrane proteins, including CD44, as the best characterized example [27,60]. Accordingly, numerous NF2 mutations in cancer patients are predicted to perturb the interaction of Merlin with CD44 [26] and anticipated to interfere with the tumor suppressor activity of Merlin [23]. We hypothesized that disruption of the Merlin-CD44 complex, through loss of Merlin, may unleash putative tumor-or metastasis-promoting functions of CD44. Accordingly, Cd44 expression was elevated in liver tumors of Nf2 flox/flox ;Alb-Cre mice when compared to control livers (Figure 6a), which is compatible with CD44 function in liver tumorigenesis. However, the status of Cd44 gene had no effect on primary liver tumor development in Nf2 flox/flox ;Alb-Cre mice (Figure 6c,d). Likewise, the ductular proliferation (Figures 2-4) and liver size (Figure 1a) were unaffected by Cd44 loss. In contrast to our study, previous reports demonstrated that deletion of the Cd44 gene inhibits initiation of colon carcinoma in Apc(Min/+) mice [61] and oncogenic progression in the testis of RHAMM-deficient mice [62]. Thus, our results suggest that deletion of Nf2 may specifically release tumorpromoting mechanisms that are independent of CD44; or alternatively, CD44 function is compensated by other molecules in Cd44-deficient Nf2-mutant mice.
As reported elsewhere, these alternative signaling pathways which mediate the phenotype of Nf2 flox/flox ;Alb-Cre mice may include epidermal growth factor receptor (EGFR), RAC1 and YAP signaling. Benhamouche et al. [6] demonstrated that the phenotype of Nf2mutant livers was effectively suppressed by treating the Nf2-deficient mice with erlotinib, an inhibitor of EGFR kinases [6], whereas Zhang et. al. [5] demonstrated that Nf2-mutant phenotype was suppressed by heterozygous inactivation of Yap. The suppression of Nf2mutant phenotype by loss of Yap was specific, since Yap deficiency did not suppress liver overgrowth and tumorigenesis induced by the expression of an oncogenic KRAS mutant (G12D) [5]. Moreover, deletion of Rac1 was shown to block tumor initiation but paradoxically exacerbate hepatomegaly induced by Nf2 loss [63]. Liver-specific knockout of angiomotin (Amot), which functions upstream of both RAC1 and YAP [64], also effectively suppressed the Nf2mutant phenotype [64]. Despite previous indications that CD44 promotes EGFR [65][66][67] and RAC1 [68,69] signaling and attenuates activation of the Hippo pathway [70][71][72][73][74] in vitro, Cd44 knockout had no effect on the phenotype of Nf2-mutant livers. As such, our results suggest that CD44 does not regulate the above-mentioned pathways in Nf2-deficient livers. Consistently, knockout of Cd44 did not prevent YAP activation in Nf2-deficient livers and cells ( Figure 5 and Figure S8). Paradoxically, transcriptional activity of YAP was even higher in livers isolated from Cd44 −/− ;Nf2 flox/flox ;Alb-Cre compared to those isolated from Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre animals ( Figure 5). However, this effect is difficult to interpret because of high variations in mRNA levels of YAP target genes between single Cd44 −/− individuals (Figure 5c). The unexpectedly mild effect of Cd44 deletion on the phenotype of Nf2-deficient livers led us to hypothesize that compensatory molecules may overtake CD44 function during liver tumor development in Cd44 −/− ;Nf2 flox/flox ;Alb-Cre mice. Existence of compensatory molecules was previously proposed to explain the striking discrepancy between in vitro studies, or interfering with CD44 function later during development, and the phenotype of Cd44 −/− mice [48,49]. Cd44 knockout mice exhibit no overt phenotype during development and only have mild abnormalities in the adult [48,49]. In contrast, interfering with CD44 function later in development, using antisense oligonucleotides or specific CD44 antibodies, led to marked deficits in organ development, skin homeostasis, neuronal axon guidance, numerous immune functions, and haematopoiesis ( [75,76]; as reviewed in [37]). These data strongly suggest that CD44 functions can be substituted during early embryogenesis (in knockout mice), whereas at later stages, this is not the case.
Interestingly, a previous study demonstrated that ICAM-1 can overtake the function of CD44 as a coreceptor of the receptor tyrosine kinase MET during liver regeneration in Cd44null mice [55]. In a number of cellular systems [38,39], or in mice with haploinsufficiency of MET or HGF [77], coactivation of MET was shown to be strictly dependent on the expression of CD44, and specifically, on its isoforms which bear sequences encoded by variant exon v6. Indeed, all tested Cd44 +/+ livers were positive for expression of CD44v6bearing isoforms (Figure 6b), which is compatible with the role of CD44 as MET coreceptor. Interestingly, livers isolated from Cd44 −/− ;Nf2 flox/flox ;Alb-Cre mice showed elevated levels of mRNA encoding Icam1 (Figure 9a), suggesting potential compensatory effect. Most strikingly, siRNA-mediated downregulation of Cd44 inhibited MET activation in Cd44 +/+ liver cells, whereas Icam1-specific knockdown interfered with MET-ERK signaling only in Cd44 −/− liver cells (Figure 9c,d). Our results suggest that, in mice with early obliteration of CD44, receptor tyrosine kinase MET may recruit alternative coreceptor molecules, including ICAM-1 [55]. Not only is MET signaling relevant during liver regeneration; it also plays an important role in liver tumorigenesis [21,58]. Thus, due to expression of compensatory molecules, potential effects of CD44 on liver tumor development or growth may be overseen in Cd44-knockout mice. It is possible that interfering with CD44 function later in life will enable identification of additional tumor-relevant CD44 properties.

CD44 Facilitates Metastatic Dissemination of Nf2-Deficient Liver Tumors
Our results demonstrate that CD44 is essential for metastatic spreading of Nf2-deficient liver tumors (Figure 6e,f). Namely, formation of lung metastasis was observed in 44% of Cd44 +/+ ;Nf2 flox/flox ;Alb-Cre animals, but in none of the analyzed Cd44-deficient Nf2-mutant mice (Figure 6e,f). Of note, one of the earliest reports on Cd44-knockout mice demonstrated deficits in migration/homing of myeloid-progenitor cells and lymphocytes [48,49]. Our previous data and other studies suggest that tumor and immune cells might use similar mechanisms to pass from the bloodstream into surrounding tissues and generate metastatic colonies [43,54,78,79]. It has been demonstrated that CD44 regulates expression and activation of integrins VLA-4 and LFA-1 on the surface of leukocytes and tumor cells [43,53,54,78,80,81]. Activated LFA-1 and VLA-4 integrins bind to their respective ligands, ICAM-1 and VCAM-1, on endothelial cells, enabling firm adhesion and passage of migratory cells through endothelium [53,81]. Interestingly, the genes encoding integrin LFA-1 (Itgal and Itgb2), as well as VCAM-1, were among the top-ranking genes differentially regulated between Cd44-positive and Cd44-negative livers of Nf2-heterozygous mice (Figure 8c). Real-time qRT-PCR results confirmed that expression of Itgb2 and Vcam1 is positively regulated by CD44 in Nf2-deficient livers (Figure 8d). Moreover, migration of liver cells through a monolayer of endothelial cells was effectively inhibited by Cd44knockout, or by blocking antibodies specific for CD44 (Figure 8e-g). These outcomes suggest that CD44 has an important role in transendothelial migration of liver cells by regulating expression of integrins and their ligands.

Conclusions
Overall, our results suggest that CD44 facilitates metastatic dissemination of liver tumor cells by upregulating expression of integrin subunit beta 2 and VCAM-1, which are required for transendothelial migration of liver tumor cells. The pro-metastatic function documented in our study could apparently not be substituted by other molecules in Cd44knockout mice. CD44 and integrins might represent a suitable target for interfering with metastatic spreading of liver cancer. Additional CD44 tumor-promoting functions might be revealed when interfering with CD44 function later in the development.