Abstract

Background & Aims

   Liver paracrine signaling from liver sinusoid endothelial cells to
   hepatocytes in response to mechanical stimuli is crucial in highly
   coordinated liver regeneration. Interstitial flow through the
   fenestrated endothelium inside the space of Disse potentiates the role
   of direct exposure of hepatocytes to fluid flow in the immediate
   regenerative responses after partial hepatectomy, but the underlying
   mechanisms remain unclear.

Methods

   Mouse liver perfusion was used to identify the effects of interstitial
   flow on hepatocyte proliferation ex vivo. Isolated hepatocytes were
   further exposed to varied shear stresses directly in vitro. Knockdown
   and/or inhibition of mechanosensitive proteins were used to unravel the
   signaling pathways responsible for cell proliferation.

Results

   An increased interstitial flow was visualized and hepatocytes'
   regenerative response was demonstrated experimentally by ex vivo
   perfusion of mouse livers. In vitro measurements also showed that fluid
   flow initiated hepatocyte proliferation in a duration- and
   amplitude-dependent manner. Mechanistically, flow enhanced β1 integrin
   expression and nuclear translocation of YAP (yes-associated protein),
   via the Hippo pathway, to stimulate hepatocytes to re-enter the cell
   cycle.

Conclusions

   Hepatocyte proliferation was initiated after direct exposure to
   interstitial flow ex vivo or shear stress in vitro, which provides new
   insights into the contributions of mechanical forces to liver
   regeneration.

Impact and implications

   By using both ex vivo liver perfusion and in vitro flow exposure tests,
   we identified the roles of interstitial flow in the space of Disse in
   stimulating hepatocytes to re-enter the cell cycle. We found an
   increase in shear flow-induced hepatocyte proliferation via β1
   integrin-YAP mechanotransductive pathways. This serves as a useful
   model to potentiate hepatocyte expansion in vitro using mechanical
   forces.

   Keywords: Mechanotransduction, FAK, hepatic sinusoid chip

Graphical abstract

   [47]graphic file with name ga1.jpg
   [48]Open in a new tab

Highlights

     * •
       Ex vivo liver perfusion stimulates hepatocytes to re-enter the cell
       cycle.
     * •
       Flow-induced shear stress induces hepatocyte proliferation
       in vitro.
     * •
       β1 integrin-YAP mechanotransduction is essential for shear-enhanced
       proliferation.

Introduction

   The liver has an outstanding capacity to regenerate after injury and is
   able to restore its original weight within 5 to 7 days after two-thirds
   partial hepatectomy (PHx) in rats.[49]^1 The portal vein flow is
   increased about threefold immediately after two-thirds PHx, followed by
   a series of rapid responses, i.e., the activity of urokinase-type
   plasminogen activator is increased within 1 min, nuclear translocation
   of β-catenin and Notch intracellular domain takes place within 30 min
   to initiate regenerative responses, and hepatocyte growth factor (HGF)
   and epidermal growth factor receptors are activated 1 h later.[50]^2
   The regenerative process is delayed when the portal blood flow is
   decreased by constructing a bypass vessel between the portal vein and
   inferior vena cava after PHx.[51]^3 Specifically, the hemodynamic
   changes in the branched blood vessels after PHx are closely related to
   the subsequent regeneration rate,[52]^4 indicating that there is a
   correlation between liver tissue regeneration and localized blood flow.
   Moreover, blood flow is heterogeneously distributed in the remnant
   branched blood vessels after PHx, resembling the differential
   distribution of hepatocyte proliferation.[53]^5 These mechanical cues
   suggest that blood flow-induced mechanical alterations may act as a key
   trigger for liver regeneration.

   As the first cell type exposed to sinusoidal blood flow, liver
   sinusoidal endothelial cells (LSECs) are subjected to mechanical
   stretch or shear and secrete HGF or downregulate transforming growth
   factor β1 (TGF-β1) expression to promote hepatocyte
   proliferation.[54]^3^,[55]^6 On the other hand, the interstitial flow
   in the space of Disse could also be critical since the sinusoidal
   endothelium is highly permeable, with distributed clusters of fenestrae
   (diameter ranging from 150 to 200 nm) on LSECs and gaps between
   LSECs.[56]^7^,[57]^8 While the effects of interstitial flow on liver
   regeneration after PHx are often neglected because of the technical
   difficulty in studying them, it is plausible that hepatocytes
   underneath the endothelium can also respond to hemodynamic changes in
   the sinusoids by sensing interstitial flow alterations directly. In
   fact, hepatocytes are known to be mechanosensitive in many
   liver-specific functions, which are promoted by low shear stress but
   impaired by high shear stress.[58]^9 Thus, it is reasonably
   hypothesized that the interstitial flow might participate in initiating
   liver regeneration starting from hepatocyte proliferation.

   Mechanotransductive pathways are key to sense and respond to
   mechanically induced cues. For example, yes-associated protein (YAP) is
   known to promote immortalized epithelial cell proliferation in response
   to mechanical stimuli, including low cell density, large cell geometry,
   or stiff substrate.[59]^10 Phosphorylated inactive YAP is retained in
   the cytosol by binding to 14-3-3, catenin or angiomotin, until upstream
   stimuli interrupt this binding and release YAP to translocate into the
   nucleus.[60]^11 One of the key upstream molecules is membranous
   integrin, together with its cytoplasmic partners such as focal adhesion
   kinase (FAK), talin, vinculin and paxillin. Meanwhile, activation of
   the integrin-FAK signaling pathway promotes hepatocellular carcinoma
   (HCC) proliferation but negatively regulates liver-specific functions
   in response to increased matrix stiffness.[61]^12^,[62]^13 On the other
   hand, β1 integrin and YAP signaling are required for liver
   regeneration. YAP knockout in hepatocytes significantly attenuates CAR
   (constitutive androstane receptor)- or SHP2 (tyrosine-protein
   phosphatase nonreceptor type 11)-mediated proliferation of mouse
   hepatocytes after PHx,[63]^14^,[64]^15 while YAP hyperactivation leads
   to liver overgrowth.[65]^16 Hepatocyte-specific knockout or knockdown
   of β1 integrin impairs liver regeneration by inhibiting the signaling
   of related growth factors.[66]^17 These observations suggest that
   mechanosensitive integrin and YAP are essential for hepatocyte
   proliferation, but their roles in shear-initiated hepatocyte
   proliferation are still unknown. Hence, we aimed to determine the role
   of interstitial flow in the initiation of hepatic regeneration as well
   as elucidate the underlying mechanotransductive mechanisms.

Materials and methods

Cell culture

   6- to 8-week-old C57BL/6N male mice (Vital River Laboratories, Beijing,
   China) were used for hepatocyte isolation, as previously
   described.[67]^8 All animal protocols were approved by the
   Institutional Animal and Medicine Ethical Committee at the Institute of
   Mechanics of the Chinese Academy of Sciences. Please refer to the
   supplementary information for more details.

Ex vivo liver perfusion

   Liver perfusion was performed by a modified procedure described
   previously.[68]^6 Briefly, the mouse was anesthetized by pentobarbital
   sodium injection before the portal vein was cannulated with a 24 G
   intravenous catheter connected to a perfusion system. The perfused
   Krebs–Henseleit saline solution was oxygenated, preheated to 37 °C,
   filtered by a bubble trap and finally pumped by a peristaltic pump to
   the liver. 2 h later, the partial liver was excised for gene and
   protein analyses, and the others were fixed for immunofluorescence,
   immunohistochemistry, and H&E staining assays. Fractions of active
   caspase-3^+ and TUNEL^+ hepatocytes were counted to evaluate
   perfusion-induced apoptosis when the intraperitoneal injection of
   D-galactosamine/lipopolysaccharide and the DNase treatment were used as
   respective positive controls. For fluorescent particle diffusion tests,
   0.2 mg/ml of FITC-labeled dextran (10 kDa; Sigma-Aldrich, MO) in
   phosphate-buffered saline (Hyclone, UT) was forced into the portal vein
   at 4 or 8 ml/min and recorded at 496/516 nm (excitation/emission)
   continuously by confocal laser-scanning microscopy (Zeiss LSM880,
   Germany). Dextran accumulation, defined by fluorescence intensity, was
   measured using imageJ software (National Institutes of Health, MD).

Microfluidic device fabrication

   The microfluidic device was constructed using soft lithography as
   described previously,[69]^7 serving as a hepatic sinusoid chip for
   in vitro tests in this work. Briefly, a silicon-wafer SU-8 template
   (Capital Bio Corporation, China) was used as a negative mold to
   generate a polydimethylsiloxane (PDMS; Dow Corning, MI) layer
   containing a channel with dimensions of 100 μm × 1 mm × 10 mm
   (height × width × length). The PDMS layer was bonded firmly to the
   glass coverslip after both PDMS and glass were treated using Plasma
   Sputtering Pump (Yilibotong, China) for 1 min. The integrated device
   was UV-sterilized for 30 min, and the channel was coated with 100 μg/ml
   of collagen I at 37 °C overnight before use. Cell suspension was then
   injected into the channel at a concentration of 1.0 × 10^7 cells/ml and
   dead cells were washed away 4 h later.

Statistical analysis

   The p values were calculated using the two-tailed t-test for any two
   groups if passing the normality test (Shapiro-Will) or using the
   Mann-Whitney U test if not. For multiple group comparisons, we
   implemented the two-way ANOVA test followed by Holm-Sidak test, the
   one-way ANOVA test followed by Holm-Sidak test (if passing the
   normality test), or the Kruskal-Wallis one-way analysis of variance on
   ranks followed by Dunn’s test (if not passing the normality test).
   Statistical significance was set at p <0.05. n values in those figure
   captions denoted the number of independent repetitions. Specifically in
   animal tests, n indicated the number of mice included per condition.

   Other relevant materials and methods are provided in the supplementary
   information and CTAT methods table.

Results

Hepatic flow alteration triggered hepatocyte proliferation ex vivo

   Emerging evidence indicates the potential role of blood flow in liver
   regeneration after PHx. To evaluate if hepatocyte proliferation could
   be regulated by flow-induced mechanical alterations, mouse livers were
   perfused with Krebs–Henseleit buffer in the absence of growth factors
   at a flow rate of 4 or 8 ml/min for 2 h to mimic physiological and
   hepatectomized conditions,[70]^6^,[71]^18 respectively. H&E staining
   indicated that the anatomical structure was kept intact without
   significantly necrotic tissues (upper panel in [72]Fig. 1A).
   Quantitative analysis of active caspase-3^+ cells% and TUNEL^+ cells%
   indicated that few apoptotic hepatocytes were induced by liver
   perfusion ([73]Fig. 1A–C). Taking cyclin D1 as a typical marker to
   identify proliferative cells,[74]^19^,[75]^20 a higher flow rate of
   8 ml/min led to more cyclin D1^+ cells which were colocalized with
   albumin-expressing hepatocytes ([76]Fig. 1D). In addition, expressions
   of those genes associated with the cell cycle, such as Ccna2, Ccnb1,
   Ccnd1, Ccne1, Cdk1 and Mki67 (encoding cyclin A2, cyclin B1, cyclin D1,
   cyclin E1, cyclin-dependent kinase 1 [Cdk1] and Ki-67, respectively)
   were also higher at 8 ml/min than those at 4 ml/min ([77]Fig. 1E). To
   quantify protein level changes, a capillary-based immunoassay was used
   to test the expression of cyclin D1 and its catalytic partner, Cdk4,
   both of which are known to form an essential heterodimeric complex that
   drives cell cycle entry and progression.[78]^21 Protein expression of
   cyclin D1 and Cdk4 supported their relevant mRNA data, also being
   enhanced at 8 ml/min ([79]Fig. 1F–H). Collectively, these results
   suggested that increased perfusion aroused hepatocytes in the quiescent
   G0 state to re-enter the cell cycle in the absence of the potential
   contributions of blood constituents.

Fig. 1.

   [80]Fig. 1
   [81]Open in a new tab

   Exvivo liver perfusion was correlated with hepatocyte proliferation.

   (A–C) Representative images of H&E, active caspase-3, and TUNEL
   staining (A), and relevant quantitative analysis of active caspase-3^+
   cells% (n = 3) (B), and TUNEL^+ cells% (n = 3) (C).
   D-galactosamine/lipopolysaccharide-damaged liver served as positive
   control for H&E and active caspase-3 tests. DNase-treated tissue served
   as a positive control for TUNEL staining. (D) Immunofluorescence
   co-staining of cyclin D1 and albumin (n = 3) in livers perfused ex vivo
   at a flow rate of 4 ml/min or 8 ml/min. (E) Expression of cell cycle
   relevant genes analyzed by RT-PCR at 8 ml/min compared with those at
   4 ml/min (n = 6). (F–H) Protein level expression was determined by
   capillary-based immunoassay (F) with relevant quantitative analysis of
   cyclin D1 (n = 6) (G) and Cdk4 (n = 6) (H). Data were presented as mean
   ± SE. p <0.05∗, 0.001∗∗∗. Statistical analysis was performed by
   two-tailed t test for Ccne1 in E, and by Mann-Whitney U test for other
   markers in E, G, H.

Hepatocytes were exposed directly to interstitial flow in the space of Disse

   Although extensive studies have attributed hepatocyte proliferation to
   those mitogenic factors released by non-parenchymal cells,[82]^2 the
   liver sinusoids are highly fenestrated, which allows blood flow to
   drive interstitial flow in the space of Disse. Based on the previous
   observations that mechanical forces play a role in liver
   regeneration,[83]^3 herein, we hypothesized that PHx-induced
   interstitial flow changes in the space of Disse might contribute to
   initiating liver regeneration, starting from hepatocyte proliferation
   via mechanosensitive pathways. To visualize the interstitial flow,
   blood vessels (red) in the freshly isolated mouse liver were stained
   with tomato lectin and hepatocytes (cyan) were depicted by their
   autofluorescence ([84]Fig. 2A). The lobule structure was clearly
   presented where the enlarged area illustrated the well-distributed
   hepatic plates and sinusoids. Subsequently, 0.2 mg/ml FITC-labeled
   dextran (10 kDa) solution was introduced into the liver via the portal
   vein at 4 ml/min and monitored at different time points ([85]Fig. 2B
   and Movie S1). The labeled dextran first appeared in the liver
   sinusoids and then diffused into the hepatocyte plate region about
   2 min after perfusion. To exclude the deviation resulting from
   anatomical microstructures, the identical position was chosen for
   visualizing dextran diffusion at a high perfusion rate of 8 ml/min
   (Movie S1) after washing dextran away. Serial-shot images revealed that
   the diffusion process was accelerated as indicated by fast dextran
   appearance outside of sinusoids (∼1 min after perfusion) and brightened
   in the hepatocyte plate region at the same time point, demonstrating
   the high permeability of the sinusoidal lumen in contrast to the
   well-established barrier function of large vessels.[86]^22 Indeed, the
   dextran tended to diffuse from microvessels (shown by dotted lines)
   towards hepatocytes (left panel in [87]Fig. 2C). Together with the
   gradually reduced fluorescence intensity from inside to outside the
   liver sinusoid, as revealed by the heatmap using ImageJ software (right
   panel in [88]Fig. 2C), this perfused fluid was inferred to be capable
   of physically contacting hepatocytes through the endothelium. In
   addition, quantitative analysis exhibited significantly higher and
   faster dextran accumulation outside of liver sinusoids after perfusion
   at 8 ml/min compared with 4 ml/min ([89]Fig. 2D). The ratio of dextran
   accumulation from outside to inside the liver sinusoids decreased with
   time and finally reached a plateau ([90]Fig. 2E). Here, the higher
   perfusion rate was associated with a more rapid descending phase,
   consistent with the high permeability interstitial flow at the high
   perfusion rate. By contrast, the steady state remained the same at both
   perfusion rates, indicating that the permeability coefficient
   determined by sinusoidal microstructure was not altered significantly
   during this perfusion process. Meanwhile, the diameter of liver
   sinusoids yielded no significant changes before and after 5 min of
   perfusion at both 4 and 8 ml/min ([91]Fig. 2F and G), implying no
   differential expansion of the sinusoidal lumen at the two perfusion
   rates. Collectively, these results suggested that hepatocytes could be
   exposed to interstitial flow and the flow rate inside the space of
   Disse was increased with increasing perfusion rate. Hence, our cell
   proliferation data ([92]Fig. 1) together with dextran perfusion
   observations ([93]Fig. 2), indicate that hepatocyte proliferation could
   be initiated by perfusion-induced mechanical alterations, such as the
   enhanced interstitial flow in the liver.

Fig. 2.

   [94]Fig. 2
   [95]Open in a new tab

   Interstitial flow sensed by hepatocytes increased with perfusion rate.

   (A) Images of blood vessel staining with tomato lectin (red) and
   hepatocytes visualized by autofluorescence (cyan). (B) Time course of
   FITC-labeled dextran (10 kDa) diffusion was monitored at the identical
   position (rectangle in A) at various flow rates. (C) Enlarged images of
   the identical liver sinusoid from the rectangles denoted in B. Dotted
   lines and arrows in left panel indicated the microvessel profiles and
   diffusion directions, respectively. Right panel presented the heatmap
   of fluorescence intensity measured. (D, E) Quantitative analysis of
   dextran accumulation outside (n = 5) (D) or outside/inside ratio of the
   liver sinusoids (n = 5) (E). (F, G) Diameter of liver sinusoids before
   and after liver perfusion at 4 ml/min (n = 10) (F) or 8 ml/min (n = 10)
   (G). Data were presented as mean ± SE. p <0.05∗. Statistical analysis
   was performed by two-tailed t test in D-G.

   Supplementary video related to this article can be found at
   [96]https://doi.org/10.1016/j.jhepr.2023.100905
   [97]Download video file^ (54.1MB, mp4)

Shear stress initiated hepatocyte proliferation in vitro

   Inspired by the above ex vivo liver perfusion readouts, we next tested
   whether flow-induced shear stress could favor hepatocytes to re-enter
   the cell cycle in vitro. Here hepatocytes were introduced into a
   hepatic sinusoid chip ([98]Fig. 3A) and exposed to flow shear first at
   a given shear stress of 0.05 dyn/cm^2 with systematically varied
   durations (1, 2, 3, 6, 12, and 24 h) ([99]Fig. 3B). Cyclin D1 gene
   expression increased with time to peak at 6 h and then declined,
   demonstrating the vital role of shear duration in controlling the cell
   cycle. To evaluate how shear amplitude affected cell proliferation,
   hepatocytes were exposed to varied shear stresses (0, 0.005, 0.05, 0.5,
   and 5 dyn/cm^2) for a given duration of 6 h ([100]Fig. 3C). Again,
   cyclin D1 gene expression increased with shear stress to peak at
   0.05 dyn/cm^2 and then decreased, implying the key role of shear
   amplitude in modulating the cell cycle. To further validate this
   finding, hepatocyte proliferation was characterized by cyclin D1 and
   EdU staining (an S phase marker) after exposure to various shear
   stresses for 6 h ([101]Fig. 3D). Quantitative analysis demonstrated
   that maximum positive cell ratio was observed at 0.05 dyn/cm^2
   ([102]Fig. 3E and F). Consistently, protein levels of cyclin D1 and
   Cdk4 exhibited similar expression profiles ([103]Fig. 3G–I). Thus,
   these results suggested that flow-induced shear stress was able to
   initiate hepatocyte proliferation in vitro, which was regulated
   cooperatively by shear duration and shear amplitude with an optimized
   parameter set at 0.05 dyn/cm^2 for 6 h.

Fig. 3.

   [104]Fig. 3
   [105]Open in a new tab

   Shear stress alone initiated hepatocyte proliferation in vitro.

   (A) Schematic diagram of an in-house developed hepatic sinusoid chip.
   (B, C) Expression of cyclin D1 gene in mouse primary hepatocytes, after
   exposure to a given shear stress at 0.05 dyn/cm^2 with varied durations
   (n = 3∼6) (B) or to varied shear stresses at given duration of 6 h (n =
   3∼10) (C). (D–F) Typical images of proliferative hepatocytes (D) and
   relevant quantitative analysis of cyclin D1^+% (n = 3) (E) and EdU^+%
   cells (n = 3) (F) at varied shear stresses for 6 h. (G–I) Protein level
   expression was determined by capillary-based immunoassays at varied
   shear stresses for 6 h (G) with relevant quantitative analysis for
   cyclin D1 (n = 3) (H) and Cdk4 (n = 5) (I). Data were presented as mean
   ± SE. p <0.05∗, 0.01∗∗, 0.001∗∗∗. Statistical analysis was performed by
   one-way ANOVA followed by Holm-Sidak test in H, I, and by
   Kruskal-Wallis one-way analysis of variance on ranks followed by Dunn’s
   test in B, C, E, F. Cdk4, cyclin-dependent kinase 4; ss-0.005/0.05/0.5,
   shear stress-0.005/0.05/0.5 dyn/cm^2.

   Physiologically, after PHx, remaining hepatocytes are required not only
   to maintain sufficient proliferation but also to meet hepatic metabolic
   demands. In this work, a hepatocyte-specific phenotype was further
   evaluated after in vitro shear stress exposure. Typical functional
   genes including Alb, Cyp3a11, Cyp1a2, Por, Glul, and Hnf4a were
   increased, or maintained at least, when applying shear stress
   ([106]Fig. S1A). Albumin production in the collected supernatant was
   also elevated by shear stress exposure ([107]Fig. S1B). In concordance
   with this data on gene expression, the capabilities for cellular
   detoxication and metabolism were validated by immunostaining CYP1A2 and
   CYP3A11 ([108]Fig. S1C); the metabolic activity of CYP3A11 was induced
   after the exposure, especially by dexamethasone induction
   ([109]Fig. S1D). In addition, glycogen accumulation and bile canaliculi
   network formation were observed in both 0.05 dyn/cm^2 for 6 h (ss-0.05)
   and static groups ([110]Fig. S1E and F). Hepatocytes were also exposed
   at various concentrations of alcohol to test their resistance
   capability to liver injury. Results indicated that shear stress
   significantly improved cell survival rate under each alcohol
   concentration compared with static control, implying a protective role
   of flow shear in alcohol-induced apoptosis ([111]Fig. S1G). In summary,
   shear stress exposure was able to maintain hepatic functions when
   initiating hepatocyte proliferation.

   To find out how hepatocytes responded to shear stress and initiated
   proliferation, RNA-seq was performed for hepatocytes in the ss-0.05 and
   static control groups. KEGG analysis showed that 40% of items were
   involved in cell growth and death among the top 10 enriched cellular
   processes for differentially expressed genes between the two groups
   ([112]Fig. 4A). Gene ontology classification indicated that there were
   more upregulated genes than downregulated ones in most biological
   processes ([113]Fig. 4B). Specifically, cell-proliferation-relevant
   genes were differentially expressed ([114]Fig. 4C). Further analysis
   implied that typical liver progenitor/cholangiocyte markers of Krt19,
   Epcam, Sox4 and Cd44 were upregulated in the ss-0.05 group
   ([115]Fig. 4D), accompanied by increased expression of the typical
   proliferation markers Ccnd1, Top2a and Aurkb ([116]Fig. 4E). This was
   also confirmed by qPCR analysis of enhanced cell cycle-dominant genes,
   namely Ccna2, Ccne1, Cdk1 and Mki67 ([117]Fig. 4F). Collectively, these
   data suggested that shear stress favored hepatocytes to acquire a
   progenitor phenotype and express cell cycle-regulating markers to
   initiate their proliferation.

Fig. 4.

   [118]Fig. 4
   [119]Open in a new tab

   Hepatocytes tended to present a progenitor phenotype in response to
   shear stress.

   (A) KEGG pathway enrichment analysis for the DEGs between ss-0.05 group
   and static control. Left and right y-axis denoted the top 10 enriched
   items involved in cellular processes and their corresponding
   classifications in the KEGG database, respectively. (B) DEG numbers in
   different biological processes based on GO classification. (C) Heatmap
   visualization of DEGs related to cell proliferation shown in B. (D, E)
   Gene expression profiles of typical liver progenitor/cholangiocyte
   markers (n = 4) (D) and proliferation markers (n = 4) (E) measured by
   RNA-seq. (F) Expression of cell cycle-relevant genes analyzed by RT-PCR
   (n = 3∼4). Data were presented as mean ± SE. p <0.05∗, 0.01∗∗,
   0.001∗∗∗. Statistical analysis was performed by two-tailed t test for
   Ccne1 and by Mann-Whitney U test for other markers in F. ss-0.05, shear
   stress-0.05 dyn/cm^2.

YAP activation was required for shear-initiated hepatocyte proliferation

   To explore the underlying mechanism of shear-initiated proliferation,
   the upregulated genes involved in cell proliferation shown in
   [120]Fig. 4C were further analyzed by KEGG. Top 10 lists showed that
   Hippo was the second most enriched signaling pathway ([121]Fig. 5A).
   Hippo/YAP signature genes were differentially regulated between ss-0.05
   and static groups ([122]Fig. 5B), as expected. Meanwhile, significantly
   higher expression of classic YAP target genes was observed in the
   ss-0.05 group ([123]Fig. 5C). Expression of YAP and its target genes,
   determined by qPCR, supported RNA-seq data ([124]Fig. 5D).
   Consistently, protein level expression of p-YAP/YAP was downregulated,
   confirming YAP activation under shear stress ([125]Fig. 5E).

Fig. 5.

   [126]Fig. 5
   [127]Open in a new tab

   YAP was required for shear-initiated hepatocyte proliferation.

   (A) Top 10 signaling pathways enriched by KEGG based on upregulated
   DEGs in [128]Fig. 3C. (B) Gene set-enrichment analysis of DEGs related
   to Hippo/YAP pathways. (C) Heatmap summarizing the expression of
   classical YAP target genes. (D) Expression of YAP and its target genes
   evaluated by RT-PCR (n = 3∼4). (E) Expression of phosphorylated and
   total YAP determined by capillary-based immunoassays (left).
   Densitometric analysis was performed by p-YAP/YAP (right) (n = 3). (F)
   Gene expression of YAP after siRNA transfection (n = 4). (G–I)
   Representative images of cyclin D1 (left) and relevant quantitative
   analysis of positive cell ratio (right) when hepatocytes were treated
   with YAP siRNA (n = 3) (G), Das (n = 5), VP (n = 5) (H), or LB (n = 4)
   (I) between ss-0.05 group and static control. Data were presented as
   mean ± SE. p <0.05∗, 0.01∗∗, 0.001∗∗∗. Statistical analysis was
   performed by Mann-Whitney U test for Ccn2 in D, by two-tailed t test
   for other markers in D, E, and by two-way ANOVA followed by Holm-Sidak
   test in F–I. Das, dasatinib; LB, leptomycin B; siYap, siRNA targeting
   YAP; ss-0.05, shear stress-0.05 dyn/cm^2; VP, verteporfin.

   To elucidate whether YAP was required for shear-initiated hepatocyte
   proliferation, the hepatocytes were transfected with siRNA targeting
   YAP (siYap). Gene expression of both YAP and its target genes Ccn2,
   Ccn1 and Ankrd1 were significantly reduced by siRNA transfection
   ([129]Fig. 5F and [130]Fig. S2A). Immunostaining (left) and
   quantitative analysis (right) revealed that the cyclin D1^+ ratio
   declined after YAP knockdown ([131]Fig. 5G). Moreover, expression of
   cell cycle-related genes (Ccna2, Ccnd1, Cdk1 and Mki67) was also
   downregulated in the siYap group ([132]Fig. S2B). Consistently, when
   hepatocytes were treated with 20 μM dasatinib or 20 μM verteporfin for
   12 h to inhibit YAP activation,[133]^23^,[134]^24 both the expression
   of YAP target genes and cell cycle-related genes, as well as the cyclin
   D1^+ ratio, were reduced ([135]Fig. S3 and [136]Fig. 5H). By contrast,
   the addition of leptomycin B, a potent inhibitor of the nuclear export
   of proteins,[137]^25 significantly elevated cyclin D1^+ ratio
   ([138]Fig. 5I). These results suggested that YAP activation was
   required for shear-initiated hepatocyte proliferation in vitro.

β1 integrin sensed shear stress and maintained YAP-dependent hepatocyte
proliferation

   We further explored how hepatocytes sensed shear stress on the cell
   membrane and mediated intracellular mechanotransduction, since YAP can
   only impel cell cycle progression by binding DNA in the nucleus. We
   first tested the integrin sensitivity of hepatocytes in response to
   shear stress and data indicated that gene and protein expressions of β1
   integrin were all upregulated ([139]Fig. 6A–C). Subsequently,
   transfecting the cells with siRNA targeting β1 integrin (siItgb1)
   significantly reduced β1 integrin gene expression, cyclin D1^+ ratio,
   and expression of Ccna2, Ccnd1, and Mki67 ([140]Fig. 6D and E, and
   [141]Fig. S4A). On the contrary, treating cells with pyrintegrin, an
   agonist to promote β1 integrin expression, increased cyclin D1^+ ratio
   ([142]Fig. S4B). Furthermore, YAP target genes were downregulated by
   integrin knockdown ([143]Fig. 6F). To determine the interplay between
   YAP and integrin partners, we treated hepatocytes with dasatinib, which
   did not affect β1 integrin expression ([144]Fig. S4C), implying that β1
   integrin served as an upstream regulator of YAP signaling. These
   results indicated that YAP activation and the initiation of hepatocyte
   proliferation in response to shear stress were associated with β1
   integrin signaling.

Fig. 6.

   [145]Fig. 6
   [146]Open in a new tab

   β1 integrin facilitated YAP-dependent hepatocyte proliferation in
   response to shear stress.

   (A-C) Representative images of β1 integrin (A) and its gene (n = 7) (B)
   or protein expression (n = 3) (C) after exposure to shear stress. (D)
   Gene expression of β1 integrin after siRNA transfection (n = 5). (E)
   Representative images of cyclin D1 and relevant quantitative analysis
   of positive cell ratio when hepatocytes were transfected with β1
   integrin siRNA (n = 4). (F) Expressions of YAP target genes after siRNA
   transfection, determined by RT-PCR (n = 4∼5). Data were presented as
   mean ± SE. p <0.05∗, 0.01∗∗, 0.001∗∗∗. Statistical analysis was
   performed by Mann-Whitney U test in B, by two-tailed t test in C, and
   by two-way ANOVA followed by Holm-Sidak test in D-F. siItgb1, siRNA
   targeting β1 integrin; ss-0.05, shear stress-0.05 dyn/cm^2.

FAK was involved in the mechanotransduction of YAP via the Hippo pathway

   Mechanical signals sensed by integrins are transmitted towards the
   intracellular region via its major downstream effectors, FAK and
   integrin-linked kinase,[147]^26 which may serve as the connecting
   player(s) between β1 integrin and YAP. Herein, protein expression of
   p-FAK/FAK was notably induced by shear stress ([148]Fig. 7A). In
   addition, FAK knockdown with siRNA transfection (siFak) reduced Fak
   gene expression, cyclin D1^+ ratio, and expression of Ccna2, Ccnd1, and
   Mki67 ([149]Fig. 7B and C, and [150]Fig. S5). YAP target genes (Ccn2,
   Ccn1 and Ankrd1) were also downregulated by FAK siRNA transfection
   ([151]Fig. 7D). To evaluate whether FAK signaling regulated
   YAP-dependent hepatocyte proliferation via the classical Hippo pathway,
   the core component of Hippo, LATS, was tested along this signaling
   cascade.[152]^27 Results indicated that p-LATS/LATS expression declined
   in response to shear stress and was rescued by FAK siRNA transfection
   ([153]Fig. 7E), indicating its role as a negative regulator of
   shear-initiated hepatocyte proliferation. These results indicated that
   FAK was involved in the cytoplasmic transmission of signals from β1
   integrin to YAP through the Hippo signaling pathway ([154]Fig. 7F).

Fig. 7.

   [155]Fig. 7
   [156]Open in a new tab

   FAK signaling promoted the mechanotransduction from β1 integrin to YAP
   through the Hippo pathway.

   (A) Protein expression of phosphorylated and total FAK. Densitometric
   analysis was performed by p-FAK/FAK (n = 5). (B) Gene expression of FAK
   after siRNA transfection (n = 6). (C) Representative images of cyclin
   D1 and relevant quantitative analysis of positive cell ratio when
   hepatocytes were transfected with FAK siRNA (n = 3). (D) Expression of
   YAP target genes after siRNA transfection, determined by RT-PCR (n =
   3). (E) Protein expression of phosphorylated and total LATS in the
   absence or presence of FAK siRNA treatment (n = 5). Densitometric
   analysis was performed by p-LATS/LATS. (F) Schematic showing the
   regulatory effects of shear stress on hepatocyte proliferation. Data
   were presented as mean ± SE. p <0.05∗, 0.01∗∗, 0.001∗∗∗. Statistical
   analysis was performed by two-tailed t test in A, by two-way ANOVA
   followed by Holm-Sidak test in B-D, and by one-way ANOVA followed by
   Holm-Sidak test in E. siFak, siRNA targeting FAK; ss-0.05, shear
   stress-0.05 dyn/cm^2.

Discussion

   Hepatocyte proliferation is vital in vivo for liver regeneration after
   PHx and critical in vitro for liver repair after injury. A single
   hepatocyte can expand through a maximum number of 34 divisions after
   PHx but loses its proliferative ability in in vitro culture, resulting
   in a shortage of hepatocytes for transplantation, bioartificial liver
   construction, and clinical research.[157]^28 Inspired by the instant
   dramatic hemodynamic alterations during PHx-induced liver regeneration,
   we tested if interstitial flow-driven shear stress in the space of
   Disse could initiate hepatocyte proliferation. Interestingly, primary
   mouse hepatocytes were able to re-enter the cell cycle, which was
   induced not only by ex vivo liver perfusion but also the application of
   in vitro shear stress in the absence of growth factors. Expression of
   genes related to liver progenitors/cholangiocytes were upregulated
   accordingly, consistent with previous findings that regenerative
   hepatocytes after PHx were reprogrammed into an early postnatal-like
   state before proceeding toward the proliferative trajectory.[158]^29
   Thus, this work provided new insights into initiating primary
   hepatocyte proliferation in vitro and could presumably contribute to
   cell expansion in cooperation with other culture methods like liver
   organoid construction.

   Emerging evidence suggests that mechanical forces play an important
   role in the regenerative process.[159]^2 For example, LSECs that line
   hepatic sinusoids first sense blood flow changes, which are followed by
   increased sinusoidal diameter, enlarged cellular fenestrae and widened
   intercellular junctions after PHx.[160]^30 Uniaxial stretched LSECs are
   able to secrete more HGF to promote hepatocyte proliferation.[161]^6 In
   response to shear stress, LSECs release nitric oxide to reinforce
   hepatocytes’ sensitivity to HGF.[162]^31 Hepatic stellate cells (HSCs),
   which reside within the space of Disse, are associated with
   synaptic-type connections to both hepatocytes and LSECs.[163]^32 They
   are also exposed to interstitial flow, like hepatocytes, and subjected
   to mechanical tension produced by the extracellular matrix.[164]^33
   Reduction in interstitial flow in the liver tissue of aged mice yields
   less HGF, whereas both shear stress and mechanical stretch in vitro
   promote HGF release from HSCs.[165]^34 Undoubtedly, these paracrine
   biochemical factors secreted from non-parenchymal cells are able to
   significantly promote liver regeneration. Taking the rapid regenerative
   response after PHx within seconds into consideration, several
   fast-phase changes on hepatocytes themselves may also participate in
   initiating liver regeneration and lead hepatocytes to enter the cell
   cycle, like membrane potential alterations of hepatocytes (within
   minutes).[166]^35 Meanwhile, since hepatocytes are known as
   mechanosensitive cells, another possibility is that hepatocyte
   proliferation is triggered by mechanical loading on hepatocytes
   directly. Combined with our direct observations of interstitial flow in
   the space of Disse, these results proposed that shear stress exposure
   was able to stimulate hepatocytes to enter the cell cycle from a
   quiescent state and thus initiate the liver regeneration process
   together with various biochemical factors.

   Cell cycle entry is driven by the formation of cyclin D and
   cyclin-dependent kinase complexes.[167]^21 In hepatocytes, cyclin D1 is
   markedly synthesized and accumulated in G1 phase, the first phase of
   the cell cycle, and reduced to low levels before entering S
   phase.[168]^36 It can activate Cdk4 to promote the expression and
   activation of other cyclin/cyclin-dependent kinase complexes that
   contribute to cell cycle progression through subsequent phases,[169]^19
   thus serving as a promising candidate for the initiation of cell
   proliferation. In our in vitro model, cyclin D1 was upregulated by
   shear stress exposure and presented time-dependent expression where it
   peaked at 6 h and declined thereafter. Of note, two-thirds PHx in vivo
   led to a peak in cyclin D1 expression at 24-48 h post
   hepatectomy.[170]^37 There are a couple of potential explanations for
   this time-window shifting of cyclin D1 expression. First, PHx not only
   initiates the liver regeneration process but also activates those
   signaling pathways related to its termination in case of excessive
   growth, including the pathway mediated by TGF-β1 (produced by
   non-parenchymal cells) or extracellular matrix (synthesized by
   HSCs).[171]^38 In contrast, the absence of those paracrine mitogenic
   inhibitors like TGF-β1 in our in vitro isolated hepatocyte model may
   result in an earlier peak time point. Second, this earlier time point
   of 6 h is also likely to be the specific peak in response to shear
   stress in vitro rather than the mechano-biochemical coupling stimuli
   in vivo. It is assumed that supplementation of biochemical factors
   originating from PHx would lead to consistent responses in hepatocytes
   in vitro and in vivo, though this warrants further verification.

   Intrahepatic shear stress is hard to measure experimentally due to the
   tiny radius of sinusoids and the varying viscosity within the liver,
   especially for the interstitial flow within the narrow space of
   Disse.[172]^31^,[173]^39 In this in vitro test, hepatocytes were
   exposed to varied shear stress from 0.005 to 5 dyn/cm^2, based on
   summarized parameters from the literature.[174]^40 Interestingly,
   applying shear stress in vitro yielded a biphasic effect on hepatocyte
   proliferation, that is, hepatocyte proliferation was favored at the
   threshold stress of 0.05 dyn/cm^2 while other lower or higher values
   all presented confined benefits, consistent with previous observations
   that an appropriate perfusion rate promoted cell proliferation to
   achieve the best output in the decellularized liver scaffold.[175]^41
   This finding may also account for the insufficient cell division
   following relatively low shear stress after 30% PHx and also for
   small-for-size syndrome following extremely high shear stress after
   extended hepatectomy or liver transplantation using small-sized
   grafts.[176]^30 In fact, hepatocyte proliferation is not homogeneous
   throughout the liver after PHx. Hepatocytes usually start to
   proliferate in the periportal region,[177]^42 while liver homeostasis
   is mainly maintained by midlobular hepatocytes.[178]^43 This
   inhomogeneous proliferation seems positively correlated to sinusoidal
   blood flow distribution, since flow velocity increases gradually from
   the periportal to pericentral region, as measured by leukocyte movement
   in different sinusoidal zones with intravital microscopy.[179]^44
   Interestingly, LSEC permeability is also increased from the periportal
   to pericentral region, as determined by increased porosity of
   fenestrae.[180]^45 Considering that the shear stress is proportional to
   the flow velocity of interstitial flow, it seems meaningful that the
   position of optimal shear stress for hepatocyte proliferation is
   located at midlobular region during liver homeostasis but shifted to
   the periportal region during liver regeneration, after blood flow is
   increased by PHx.

   Hepatocyte proliferation could be accomplished in vitro to a certain
   extent, although little is known about the underlying molecular
   mechanism. For example, an in vitro 3D spheroid model of primary human
   hepatocytes recapitulates the in vivo regenerative phenotype upon
   PHx.[181]^46 Based on analyzing global promoter motif activities,
   Wnt/β-catenin activation and p53 signaling inhibition have been
   identified as critical factors for primary human hepatocyte
   proliferation. In response to mechanical modulation, 3D organoid
   culture triggers two waves of proliferation induced by matrix stiffness
   and cell-cell interactions via the MER1/2-ERK1/2 signaling
   pathway.[182]^20 To the best of our knowledge, primary hepatocyte
   proliferation initiated by shear stress in vitro has not been studied
   yet, although immediate early gene expression is increased in hepatic
   cell lines under laminar or interstitial flow.[183]^39^,[184]^47
   Furthermore, our results indicated that shear stress stimulated
   hepatocytes to re-enter the cell cycle in a YAP-dependent manner,
   regulated by β1 integrin through the Hippo signaling pathway.

   Intracellular mechanotransductive pathways are crucial to understand
   the underlying mechanisms in shear-initiated hepatocyte proliferation.
   As to the sensation of extracellular mechanical signals, our results
   indicated that β1 integrin knockdown or activation significantly
   weakened or strengthened shear-initiated hepatocyte proliferation,
   respectively, indicating the key role of β1 integrin as a shear sensor
   on the cell membrane. These observations were also consistent with the
   findings that shear-activated integrin-FAK signaling regulates HCC
   migration in a time-dependent manner.[185]^48 Subsequent signal
   transmission from the cell membrane to the intracellular region is
   known to be mainly mediated by focal adhesions or cytoskeletons once
   perceived by integrins. Consistently, our data indicated that FAK
   knockdown broke up integrin-YAP signaling transduction and further
   reduced proliferative responses related to YAP shuttle. As to the
   intranuclear responses, activated YAP is able to translocate into the
   nucleus and initiate the transcription of various effector molecules.
   For example, aerobic glycolysis-regulated HCC migration relies on stiff
   matrix-induced YAP nuclear translocation.[186]^49 Accordingly, our work
   indicated that YAP-dependent hepatocyte proliferation was impaired by
   either YAP knockdown or inactivation but enhanced by YAP activation,
   identifying YAP as a key player in shear-initiated hepatocyte
   proliferation. In fact, while the β1 integrin-FAK-YAP pathway has been
   extensively explored in HCC, how it works in primary hepatocytes is
   poorly understood. Moreover, β1 integrin and YAP have previously been
   shown to be essential for liver regeneration,[187]^14^,[188]^17 but
   their roles in regulating shear-initiated hepatocyte proliferation were
   unknown. Intriguingly, we show that β1 integrin could activate
   YAP-dependent hepatocyte proliferation under shear stress, providing
   potential targets for establishing primary hepatocyte expansion
   in vitro.

   In this work, mouse livers were perfused ex vivo and hepatocyte
   proliferation was found to be initiated at a high flow rate. Inspired
   by this observation, primary hepatocytes were exposed to shear stress
   in vitro in a hepatic sinusoid chip. Results indicated that direct
   exertion of shear stress indeed initiated proliferation, in a shear
   duration- and amplitude-dependent manner. The underlying
   mechanotransductive pathway involved the β1 integrin-FAK-Hippo-YAP
   signaling axis, driving hepatocytes to re-enter the cell cycle. These
   findings provide new insights into the mechanisms triggering liver
   regeneration in vivo and potentiating hepatocyte expansion in vitro.

Financial support

   This work was supported by the National Natural Science Foundation of
   China Grants 32130061, 32271366 and 32101056, National Key R&D Program
   of China Grant 2021YFA0719300, China Manned Space Flight Technology
   Project Chinese Space Station Experiment Project YYWT0901EXP0701 and
   the Scientific Instrument Developing Project and the Key Research
   Program of Chinese Academy of Sciences Grants GJJSTU20220002 and
   ZDBS-ZRKJZ-TLC002.

Authors’ contributions

   W Li, N Li, and M Long conceptualized and designed the study; W Li, ZL
   Zhang, Y Wu, J Zhou, XY Shu, and XY Zhang performed experiments; W Li,
   WH Hu, ZL Zhang and H Wu analyzed data; W Li, N Li, Y Du, DY Lü, SQ Lü,
   and M Long wrote the article.

Data availability statement

   The data that support the findings of this study are available from the
   corresponding author upon reasonable request.

Conflict of interest

   The authors declare no competing interests.

   Please refer to the accompanying ICMJE disclosure forms for further
   details.

Footnotes

   Author names in bold designate shared co-first authorship

   Supplementary data to this article can be found online at
   [189]https://doi.org/10.1016/j.jhepr.2023.100905.

Contributor Information

   Ning Li, Email: lining_1@imech.ac.cn.

   Mian Long, Email: mlong@imech.ac.cn.

Supplementary data

   The following are the supplementary data to this article:
   Multimedia component 1
   [190]mmc1.pdf^ (848.1KB, pdf)
   Multimedia component 3
   [191]mmc3.docx^ (210.5KB, docx)
   Multimedia component 4
   [192]mmc4.docx^ (46.3KB, docx)
   Multimedia component 5
   [193]mmc5.pdf^ (627.9KB, pdf)
   Multimedia component 6
   [194]mmc6.pdf^ (4.2MB, pdf)

References