Background
Liver sinusoidal endothelial cells (LSECs) are a prime example of organ-specific endothelial cells (ECs) that form the hepatic vascular niche. Their morphology is unique as they are equipped by fenestrae and lack a basement membrane. Among their functions are the angiocrine instruction of surrounding cells and control of organ functions by surface expression or secretion of multiple factors [
1,
2]. Angiocrine secretion of Wnt and Hgf is crucial for the regulation of liver size and liver regeneration [
3‐
5]. Moreover, by the secretion of Bmp2 and Bmp6 LSECs control systemic and hepatic iron homeostasis [
6,
7]. Besides, LSECs are an important immunological component as they contribute to tolerance induction [
8].
Notably, LSECs are also decisively involved in almost all liver diseases such as liver fibrosis, carcinogenesis or hepatic tumor metastasis [
9,
10]. The differentiation of LSECs is critically controlled by the transcription factor
Gata4. Loss of
Gata4 in LSECs drives sinusoidal capillarization leading to anemia in the fetal liver due to disturbed hepatic hematopoiesis [
11]. In the adult liver its loss in LSECs results in a profibrotic angiocrine switch mediating perisinusoidal fibrosis and hepatopathy [
12]. Besides, activation of hepatic endothelial Notch signaling also changes endothelial differentiation towards a continuous EC phenotype [
13]. LSECs represent the first cells circulating tumor cells encounter when disseminating to the liver. E-selectin, ICAM1 or Clec4g mediate tumor cell adhesion and thus facilitate liver metastasis [
14‐
16]. An influence of Notch signaling on disease pathology is highlighted as upregulation in hepatic ECs protects from liver metastasis of colorectal carcinoma and cutaneous melanoma, via downregulation of ICAM1 [
13]. Besides, deposits of fibronectin alongside the hepatic endothelium were shown to bind circulating tumor cells [
17]. Moreover, LSECs were shown to critically control liver metastasis by the secretion of inflammatory cytokines, such as Il-1, Mif or Cxcl12 [
18,
19].
By expression of a variety of surface receptors, such as Stabilin-1 (Stab1), Stabilin-2 (Stab2), LYVE-1 or CD32b, LSECs contribute to the clearance of circulating factors from the blood [
20]. In this regard, the physiologic involvement of LYVE-1/Lyve-1 in blood clearance has not yet been fully defined, although it is known to mediate endocytosis of hyaluronan. LYVE-1/Lyve-1 is expressed on subsets of endothelial cells, mainly the lymphatic endothelium [
21] and sinusoidal vascular ECs such as LSECs [
22]. In comparison to the human system, LYVE-1/Lyve-1 expression appears to be broader on vascular ECs in mice, as it is widely expressed during embryonic development and is also found on subsets of vascular EC in the adult heart, spleen, adrenal gland and lungs [
23,
24]. Besides ECs, subtypes of macrophages, such as some tumor-infiltrating macrophages, also show expression of LYVE-1/Lyve-1 [
25,
26]. However, hepatic Kupffer cells have been described to be Lyve-1 negative [
25]. The glycoprotein LYVE-1/Lyve-1 can bind hyaluronan, an extracellular matrix glycosaminoglycan, and functions as an endocytic receptor [
27] similar to CD44 [
21]. Recently, interaction of Lyve-1 with hyaluronan was found to be involved in lymphatic trafficking of dendritic cells [
28]. Regarding cancer metastasis, in vitro studies imply LYVE-1 to be involved in tumor cell adhesion [
29]. This can be taken advantage of therapeutically, as administration of a monoclonal anti-Lyve-1 antibody strongly decreased tumor formation and lymphatic metastasis of breast cancer in a mouse model [
30]. As of yet, the involvement of LYVE-1/Lyve-1 in tumor cell adhesion to LSECs and hepatic metastasis has not been studied.
Knockout mouse models with varying targeting strategies of the
Lyve-1 locus were generated to decipher the function of Lyve-1. In general, the loss of Lyve-1 does not result in reduced viability or fertility or obvious pathological organ dysfunction. Huang et al. described distended luminal size of lymphatic vessels in the livers of Lyve-1 deficient mice and hypothesized that hyaluronan or PDGF-BB, both ligands of Lyve-1, may act as opening sensors for lymphatic vessels [
31]. However, in the study of Gale et al. lymphatic fluid drainage, hyaluronan plasma levels, organ development, leukocyte distribution and subcutaneous tumor growth were unchanged [
32]. Therefore, they hypothesized that other hyaluronan receptors, such as CD44, may compensate for the loss of Lyve-1 in knockout mice. However, Lyve-1 and CD44 double knockout mice also show no obvious pathological lymphatic phenotype, but a tendency towards increased numbers of leukocytes in the peritoneal cavity after inflammatory stimulation [
33]. Yet, the role of Lyve-1 in hepatic endothelial differentiation and function including metabolic zonation and pathological processes in the liver, such as hepatic metastasis, has not been studied in detail.
Therefore, we here address the role of Lyve-1 in these processes. In this regard, hepatic endothelial differentiation, angiocrine functions and involvement in disease pathogenesis were studied in mice with a constitutive knockout of Lyve-1 (Lyve-1-KO) [
31]. Hepatic metastasis of both CRC and CM were studied by splenic or intravenous injections in Lyve-1-KO. We show that Lyve-1 influences tumor-specific hepatic metastasis of melanoma but not CRC without interfering with angiocrine organ functions of LSECs.
Methods
Animal ethics
The animal ethics Committee of Baden-Wuerttemberg (Regierungspräsidium Karlsruhe) approved all animal experiments. In adherence to the Guide for the Care and Use of Laboratory Animals published by the National Academy of Sciences the animal care was performed.
Animal characterization and experiments
Mice between 6 to 18 weeks of age were used for animal characterization and routine analysis. B6.129S1-
Lyve1tm1Lhua/J (JAX stock #006221) mice with C57BL/6 background were utilized as Lyve-1
−/− group [
31]. Littermates bred in a heterozygous mating mode and purchased wild-type C57BL/6 J mice (Janvier Labs, Le Genest-Saint-Isle, France) served as control group (Ctrl). Heterozygous mice were not used. For DNA extraction and genotyping KAPA HotStart Mouse Genotyping Kit (KK7352, Merck, Darmstadt, Germany) and primers (Metabion international AG, Planegg/Steinkirchen, Germany) were utilized (Additional file
3: Table S1). With a 12 h/12 h day/night cycle all mice were hosted in single ventilated cages (Sealsafe plus DGM™, Tecniplast, Buguggiate, Italy) under specific-pathogen free conditions. Mice were fed ad libitum with a standard rodent diet (V1534-000, Ssniff, Soest, Germany) and always had free access to water.
Cell lines
B16F10
luc2 mouse melanoma cells, WT31 mouse melanoma cells and MC38 colorectal cancer cells were used for in vivo experiments. B16F10
luc2 cells were bought from ATCC (Manassas, VA, USA). WT31 mouse melanoma cells [
34] were kindly provided by O. Sansom (Beatson Institute for Cancer Research, Glasgow, Scotland). The MC38 colorectal cancer cells were a gift from S. Herzig (Helmholtz Zentrum Munich, Germany).
The same passage of the respective cell lines was utilized for in vivo experiments. After thawing, B16F10 luc2 and WT31 cells were held in RPMI Medium (61870044, Gibco™, Thermo Fisher Scientific, Waltham, MA, USA) and MC38 cells were held in DMEM Medium (61965059, Gibco™, Thermo Fisher Scientific, Waltham, MA, USA) respectively with 10% fetal bovine serum (10270106, Gibco™, Thermo Fisher Scientific, Waltham, MA, USA) and 1% of penicillin/streptomycin (15140122, Gibco™, Thermo Fisher Scientific, Waltham, MA, USA) at 37 °C in air with 5% CO2. After thawing of the cells, they were prepared for the in vivo experiments within maximum 1 week. Thereby, the cells were passaged not more than three times. The cells used for the in vivo experiments were mycoplasma-free tested by PCR. Cell authentication was confirmed by cell morphology and pigmentation status.
Liver and lung colonization assay
Liver and lung colonization assays were performed as previously established and described [
13,
35]. Female mice starting with an age of 11 weeks were used in vivo experiments. All experiments were performed with age matched mice.
To study liver colonization, spleen injections or intravenous injections of tumor cells were performed. After spleen injection of 1.5 × 105 MC38 colorectal carcinoma cells, the mice were sacrificed at day 21. When 1.5 × 105 B16F10 luc2 cells were injected into the spleen, the mice were sacrificed after 14 days and BLI measurement of the liver was performed. 1.25 × 106 WT31 melanoma cells were applied by tail vein injection and mice were sacrificed after 19 days. When WT31 melanoma cells were injected into the spleen, 0.2 × 105 cells were used and mice were analyzed at day 21. To analyze the retention of melanoma cells in the liver 3 × 105 B16F10 luc2 cells were injected into the spleen and analysis was performed 90 min after intrasplenic injection by BLI.
BLI
For bioluminescence imaging (BLI) an IVIS® Lumina LT In Vivo Imaging System (Caliper Life Sciences, Perkin Elmer, Waltham, MA, USA) was used. Ten minutes after intraperitoneal application of luciferin (D-luciferin 1-(4,5-dimethoxy-2-nitrophenyl)ethyl ester, 7903, 30 mg/ml, BioVision, Milpitas, CA, USA), mice were sacrificed, livers were removed, put into a Petri dish (664160, Greiner Bio One, Kremsmünster, Austria) and were imaged ex vivo with the IVIS® Lumina LT with an exposure time of 45 s.
Iron quantification in liver
The iron in mouse liver lysates was measured with the Iron Assay KIT (MAK025-1KT, Merck, Darmstadt, Germany) according to the manufacturer’s protocol.
LSEC isolation
LSEC isolation was performed as previously described [
36]. Livers of Ctrl (C57BL/6J) and B6.129S1-
Lyve1tm1Lhua/J with C57BL/6 background [
31] were perfused with a 0.05% collagenase containing amino acid/ saccharide calcium-deprived medium (C2674, Sigma–Aldrich, St. Louis, MO, USA) via the portal vein. Mice were sacrificed and livers were excised. Livers were minced mechanically and digested at 38 °C in a collagenase/Gey’s balanced salt solution (G9779, Sigma-Aldrich, St. Louis, MO, USA). Afterwards livers were filtered through a 250 µm mesh. A gradient was applied by 35% Nycodenz (1002424, Axis-Shield, Alere Technologies, Oslo, Norway) and cells were separated. Utilizing anti-CD146 MicroBeads (ME-9F1, 130-092-007, Miltenyi Biotech, Bergisch Gladbach, Germany) LSECs were isolated by magnetic-activated cell sorting according to the manufacturers’ instructions. Purity of LSECs was checked by their positivity for CD31 and/or Stab2 by flow cytometry with a BD FACSCanto II (BD Biosciences, Franklin Lakes, NJ, USA). For additional analysis, LSECs were pooled from three mice.
Microarray data33
RNA isolation and preparation for microarray analysis were performed as previously described [
36]. Utilizing arrays MoGene-2_0-st from Affymetrix (Santa Clara, CA, USA) the gene expression profiling was executed. cDNA and cRNA were synthetized and hybridized to the arrays in accordance with manufacturer’s recommendation. For annotating the arrays a custom CDF Version 18 (MoGene-2_0-st) with Entrez based gene definitions was utilized. Normalizing raw fluorescence intensity values was performed according to quantile normalization. By using a commercial software package SAS JMP7 Genomics, version4, from SAS (SAS Institute, Cary, NC, USA) differential gene expression was analyzed with the One-Way-ANOVA. Level of significance was determined as a false positive rate of a = 0.05 with FDR correction. The raw as well as normalized data are accessible in the Gene Expression Omnibus database (GEO) (
http://www.ncbi.nlm.nih.gov/). The GEO accession number is GSE199055.
Flow cytometric analysis of liver immune cells
Single cell suspensions from liver tissue were first incubated with Zombie Aqua™ fluorescent dye (Biolegend) to label dead cells. Following Fc-receptor blockade with a CD16/32-producing hybridoma supernatant, cells were incubated with antibodies targeting surface molecules. Cells were fixed and permeabilized using eBioscience™ Intracellular Fixation & Permeabilization reagents (Invitrogen™) according to manufacturer’s instructions. Intracellular staining was performed in permeabilization buffer. All antibodies were purchased from Biolegend and BD Biosciences. Data were acquired by LSRFortessa™ X-20 Flow Cytometer (BD) and analyzed using FlowJo™ Software (BD).
Statistical analysis
Prism 8 (GraphPad Software, La Jolla, CA, USA) was used for performing statistical analysis. Unpaired, two-tailed t-test or Mann–Whitney U-test were applied for statistical testing depending on whether normality was proven by a Shapiro–Wilk test or not. If not stated differently in the figure legends, an unpaired, two-tailed t-test was used. For the analysis of the number of hepatic metastases after spleen injection of WT31 melanoma two outliers were identified by the ROUT method in the Lyve-1-KO group and excluded for the plot and analysis. A P-value < 0.05 was regarded as statistically significant. The graphs represent the mean ± SEM.
For a detailed description of analysis of blood parameters, plasma proteomics, histological methods, antibodies, image acquisition and immunoblot please refer to the Additional file
2.
Discussion
LYVE-1/Lyve-1 is expressed on lymphatic endothelium, subsets of vascular endothelial cells, for example LSECs, and on subgroups of macrophages, such as certain tumor-infiltrating macrophages. Its molecular functions are also diverse, as it has been implied in endocytosis and scavenging, as well as in adhesion and migration of immune and tumor cells.
LSECs are the major endothelial scavenger cell population and responsible for clearance of circulating blood factors. LYVE-1/Lyve-1 was initially assumed to act as a scavenging receptor of hyaluronan (HA) [
21,
27]. However, previous studies in Stab1 and Stab2 deficient mice have clearly identified Stab2 as the major scavenger receptor responsible for the clearance of HA [
20]. In contrast to investigations by Gale et al. with an alternative knockout approach of
Lyve-1 [
32], we found mildly, but consistently elevated plasma levels of hyaluronan in animals with
Lyve-1 deficiency. Notably, hyaluronan plasma levels were not analyzed in the original description of the
Lyve-1 knockout mice designed by Huang et al. and used in this study [
31]. The difference between the two knockout strains first described by Gale et al. and Huang et al. may result from discrepancies of the genetic background, age or environmental factors, such as food or stress levels. Nonetheless, in comparison to Stab2 KO mice HA levels were about 3 to fourfold lower in Lyve-1-KO plasma indicating that Stab2 is the major scavenger receptor responsible for clearance of HA in the blood, while Lyve-1 appears to have a context-dependent smaller role that is mostly compensated by Stab2. One can hypothesize that functions of Lyve-1 might also be compensated by its homologue CD44. Unfortunately, only tissue levels of hyaluronic acid after carrageenan-induced paw edema have been analyzed in Lyve-1/CD44 double knockout mice and did not show any differences compared to Lyve-1-KO and wild-type mice [
33]. Interestingly, Lyve-1/CD44 double knockout mice do not present any overt phenotype. Comprehensive analysis of clearance functions of Lyve-1 by plasma proteomics revealed only CA2 as potential selective protein ligand accumulating in the blood of Lyve-1-KO mice. Factor XIII on the other hand was less abundant in blood plasma of Lyve-1-KO. This indicates that Lyve-1 appears to be involved in the control of blood plasma composition to a smaller extent in comparison to Stab1 and Stab2.
During tumor metastasis, hepatic endothelial cells represent the first barrier interacting with arriving cancer cells via surface receptors like E-selectin, ICAM1 or Clec4g [
14‐
16]. Besides, adhesion of hyaluronan-expressing tumor cells to endothelial cells is mediated by Lyve-1 [
29]. Contrastingly, high levels of plasma HA strongly decreased metastasis of B16F10 melanoma to the lungs by reduction of tumor cell adhesion to the pulmonary capillaries [
38]. Of note, Hirose et al. identified that high levels of HA did not impact melanoma cell invasion, migration or proliferation. However, in our liver colonization model no differences in initial melanoma cell retention were detected among Lyve-1-KO and Ctrl mice. This might be explained by tumor cell-type and context-dependency, as well as by cellular and molecular differences of
in vivo and in vitro models of tumor cell adhesion. Transdifferentiation and capillarization of LSECs can also influence hepatic metastasis e.g. via alterations of tumor cell adhesion [
13]. However, as Lyve-1-KO showed no signs of capillarization or alterations of angiocrine metabolic functions, this cannot explain reduced melanoma metastasis in Lyve-1-KO livers.
Alterations in hepatic metastasis of melanoma but not CRC indicate that the anti-tumor immune response may be altered in Lyve-1-KO. Lyve-1 is involved in hyaluronan-dependent leukocyte trafficking in lymphatic EC [
28,
39] and is expressed on subtypes of macrophages. However, in our Lyve-1-KO mouse model, there were almost no macrophages in the liver detected that co-express Lyve-1. Neither by flow cytometry significant differences in the number of hepatic macrophages were revealed. This is in line with the study of Zheng et al. that did not find an expression of Lyve-1 in macrophages of visceral organs [
25].
Interestingly, loss of Lyve-1 led to iron deposition in F4/80
+ macrophage subpopulations and decreased levels of iron in the plasma. In general, iron metabolism plays a decisive role in macrophages and inflammatory processes [
40]. Macrophages exert a special function during iron metabolism and are of special importance for recycling of iron from senescent erythrocytes and the export into the plasma [
41]. Hereby, the iron exporter ferroportin1 is critically involved as a macrophage-specific loss of ferroportin1 leads to hepatic iron depositions and increases inflammatory responses [
42]. These iron depositions appear like the ones in Lyve-1-KO mice leading us to hypothesize that Lyve-1 may also be involved during hepatic iron turnover. Apart from that, hepatic iron homeostasis is critically controlled by angiocrine secretion of Bmp2 and Bmp6 by LSECs [
6,
7]. However, deficiency of either Bmp2 or Bmp6 leads to a different pattern of iron deposition and a microarray analysis of isolated LSECs revealed no significant differences in the levels of Bmp2 or Bmp6 between Lyve-1-KO and Ctrl mice. Besides, there is accumulating evidence that iron deposition in macrophages promotes a switch towards a pro-inflammatory phenotype during inflammatory conditions, such as non-alcoholic steatohepatitis, anti-cancer immunity or wound healing [
43‐
45]. Regarding liver metastasis, application of ferumoxytolof, an iron supplement, effectively prevented liver colonization of small-cell lung cancer [
46].
Altogether, the significant reduction in the number of B16F10
luc2 and WT31 melanoma liver metastases might be due to an increased premetastatic pro-inflammatory microenvironment in the liver. This is substantiated by microarray analyses of isolated LSECs showing regulation of relevant inflammatory pathways. Although there were no significant changes in the number of CD4
+ and CD8
+ T cells in livers of Lyve-1 KO mice bearing established melanoma metastases, tumor-free livers of Lyve-1 KO mice showed significant increased numbers of CD4
+ and CD8
+ T cells, T
reg cells and eosinophils. These alterations may result from either modulation of leukocyte influx via LSEC or modulation of leukocyte efflux via lymphatic EC. The protective role of T cells for liver metastasis, especially CD8
+ T cells, is highlighted by various studies. Hepatic immunomodulation by α-melittin-nanoparticles increases numbers of CD4
+ and CD8
+ T cells, which protect from hepatic metastasis [
47]. Besides, clinical studies of CRC long-term survivors with liver metastases and a clinical case of metastatic uveal melanoma reveal a prognostic benefit of higher numbers of CD8
+ T cells in the liver [
48,
49]. Our results identify an enhanced premetastatic hepatic immune microenvironment which might promote initial tumor cell rejection and consequently decreased hepatic melanoma colonization, as initial tumor cell retention in mice with
Lyve-1 deficiency was unaltered. The liver specific effect of Lyve-1 on melanoma metastasis could be proven by the intravenous injection route of WT31 melanoma cells, as deficiency of
Lyve-1 led to a significant protection against hepatic but not pulmonary melanoma colonization. Last, the tumor-specific differences in hepatic metastasis in our model might be explained by varying immunogenicity of the CRC cell line MC38 as compared to the two melanoma models.
To treat liver metastasis of CM immune checkpoint inhibition (ICI) or targeted therapies (BRAF/ MEK inhibitors) are state-of-the art options. However, liver metastasis of CM goes along with a poor prognosis for both therapy response to ICI [
47] and targeted therapy [
50]. Therefore, the microenvironment of the hepatic niche needs further attention and new therapeutic targets must be evaluated. Due to the decreased hepatic metastasis in two melanoma models and the increased premetastatic pro-inflammatory state in Lyve-1-KO, Lyve-1 might be an interesting therapeutic target [
30] to further boost therapy response to ICI.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.