Continuous sensing of IFNα by hepatic endothelial cells shapes a vascular antimetastatic barrier

Hepatic metastases are a poor prognostic factor of colorectal carcinoma (CRC) and new strategies to reduce the risk of liver CRC colonization are highly needed. Herein, we used mouse models of hepatic metastatization to demonstrate that the continuous infusion of therapeutic doses of interferon-alpha (IFNα) controls CRC invasion by acting on hepatic endothelial cells (HECs). Mechanistically, IFNα promoted the development of a vascular antimetastatic niche characterized by liver sinusoidal endothelial cells (LSECs) defenestration extracellular matrix and glycocalyx deposition, thus strengthening the liver vascular barrier impairing CRC trans-sinusoidal migration, without requiring a direct action on tumor cells, hepatic stellate cells, hepatocytes, or liver dendritic cells (DCs), Kupffer cells (KCs) and liver capsular macrophages (LCMs). Moreover, IFNα endowed LSECs with efficient cross-priming potential that, along with the early intravascular tumor burden reduction, supported the generation of antitumor CD8+ T cells and ultimately led to the establishment of a protective long-term memory T cell response. These findings provide a rationale for the use of continuous IFNα therapy in perioperative settings to reduce CRC metastatic spreading to the liver.


Continuous IFNα administration prevents spontaneous hepatic colonization of 164 orthotopically implanted CT26 LM3 cells 165
To confirm the above-mentioned results in a different metastatic setting, we developed an 166 orthotopic CRC model of liver metastases by implanting invasive CRC cells into the mouse 167 cecal wall. As previously reported (Zhang et al, 2013), invasive CRC cells were generated 168 by serial intracecal injections of the parental CT26 cells into CB6 mice ( Fig S3A). The 169 percentage of metastatic livers in intracecally implanted mice significantly increased as 170 CT26 cells were passaged, with an almost 100% of animals bearing multiple liver 171 metastases after 3 rounds of in vivo selection (Fig S3B-D). Three-time passaged cells 172 (termed CT26 LM3 ) were then orthotopically implanted in the cecal wall of CB6 mice and 7 173 days later the animals were treated with either NaCl or IFNα ( Fig 3A). 174 Consistent with our previous results (Fig 2B), serum IFNα levels peaked at day 2 after MOP 175 implantation (Fig 3B), without causing myelotoxicity (Fig 3C), and MRI analyses performed 176 14 days later revealed that continuous IFNα therapy did not alter the growth of primary 177 intracecal tumors (Fig 3D,E), while IFNα treatment significantly reduced both number and 178 size of hepatic lesions (Fig 3D,F) with 60% of mice spared from metastatic lesions (Fig 3H). 179 The primary intracecal tumors ( Fig S4A) and liver metastases ( Fig 3G) detected after 180 orthotopic implantation of CT26 LM3 cells were also characterized by immunohistochemistry 181

(IHC). This analysis showed that primary intracecal tumors and liver metastatic lesions of 182
NaCl-treated control mice were highly proliferative (as denoted by Ki67 positivity), exhibited 183 marked signs of angiogenesis (as denoted by CD34 staining) and, accordingly with previous 184 reports ( Catarinella et al., 2016;Tauriello et al, 2018), were devoted of F4/80 + resident 185 macrophages and CD3 + T cells (Fig 3G,H). Similar results were also observed in IFNα-186 treated primary intracecal tumors (Fig S4A,B). The absence of liver metastases in the 187 majority of IFNα-treated mice is reflected by a reduced Ki67 or CD34 staining and an 188 apparently normal distribution of F4/80 + macrophages and CD3 + T cells (Fig 3G,H). The few 189 VeCad Ifnar1_KO mice treated with either NaCl or IFNα displayed very similar numbers and 242 sizes of hepatic lesions (Fig 4C,D) or survival rates (Fig 4F), indicating that the antimetastatic 243 properties of IFNα requires Ifnar1 signaling on HECs. VeCad Ifnar1_KO mice exhibited 244 increased tumor burden ( Fig 4D) and mortality rates ( Fig S6D) when compared to NaCl-245 treated Ifnar1 fl/fl mice, suggesting that hepatic endothelial Ifnar1 signaling exerts significant 246 anti-tumor activity even in the context of physiologic endogenous intrahepatic levels of type 247 I interferons. Furthermore, histological analysis of hepatic CRC lesions from NaCl-and 248 IFNα-treated VeCad Ifnar1_KO mice euthanized at day 21 after MC38 intramesenteric injection 249 indicated that these tumors resembled NaCl-treated Ifnar1 fl/fl lesions, showing signs of 250 angiogenesis (as denoted by CD34 positivity) and similar content of F4/80 + macrophages 251 and CD3 + T cells within the intrahepatic CRC foci (Fig S7A,B). is smaller than their own (12 ± 0.1 µm), similar to what we previously reported (Catarinella 263 et al., 2016). As the process of trans-sinusoidal migration -a critical limiting step in the 264 metastatic cascade -is known to occur within 24 hours of CRC challenge (Chambers et al., 265 2002;Valastyan & Weinberg, 2011;Wolf et al, 2012), the intrahepatic number and 266 localization of MC38 GFP cells were then studied at this time point. Confocal IF quantification 267 revealed that, compared to NaCl-treated Ifnar1 fl/fl animals, MC38 GFP cells were about ~2-268 fold less abundant in IFNα-treated Ifnar1 fl/fl mice and ~3-fold more abundant in VeCad Ifnar1_KO 269 mice treated with NaCl or IFNα (Fig 5A top). Moreover, confocal 3D reconstructions of liver 270 sinusoids from IFNα-treated Ifnar1 fl/fl mice unveiled that by 24 hours most MC38 GFP cells 271 localize intravascularly (i.e. they did not invade the liver parenchyma), while in NaCl-treated 272 Ifnar1 fl/fl controls and in NaCl-or IFNα-treated VeCad Ifnar1_KO mice only few MC38 GFP cells 273 remain within the liver vasculature (i.e. they invaded the liver parenchyma) (Fig 5A bottom,274 5B and Video S1-4). These results indicate that HECs, including LSECs, negatively control 275 trans-sinusoidal CRC migration upon IFNα sensing.  (Pandey et al, 2020;Wohlfeil et al, 2019). By contrast, IFNα treatment 293 of VeCad Ifnar1_KO mice showed no effect (Fig S9A,B). 294 Next, we evaluated the status of the vascular glycocalyx (GCX), a fibrous network of 295 glycoproteins and proteoglycans that lines the LSECs and projects intraluminally (Reitsma 296 et al, 2007). Notably, enhanced GCX deposits can act as a repulsive barrier that prevents 297 tumor cell interactions with endothelial cells, adhesion molecules or chemokines have been 298 previously identified as negative correlates of transendothelial migration (Glinskii et al, 2005;299 Mitchell & King, 2014;Offeddu et al, 2021;Wilkinson et al, 2020). Continuous IFNα 300 treatment modified this network as well, increasing its thickness (Fig 5E,F top) and the 301 expression of one of its major components, the heparan sulfate (HS) (Reitsma et al., 2007) 302 (Fig 5E,F bottom). Of note, VeCad Ifnar1_KO mice displayed reduced GCX thickness 303 independently of NaCl-or IFNα-treatment (Fig 5E,F). Additionally, we evaluated the vascular 304 and perivascular status of cell adhesion molecules such as selectins and integrins, which 305 have been positively associated with the transendothelial migration of tumor cells (Glinskii 306 et al., 2005;Wilkinson et al., 2020). The expression of ICAM1, E-selectin (CD62E) (Fig  307   S8G,H) and the integrins ITGB2 (CD18) or ITGA4 (CD49d) (Fig S8C) was up-regulated in 308 IFNα-treated Ifnar1 fl/fl controls, while significantly reduced or attenuated in IFNα-treated 309 VeCad Ifnar1_KO mice. The notion that a more modest upregulation of some these markers 310 was still evident in the latter mice may reflect the capacity of liver cells other than HECs to 311 respond to IFNα. Altogether, these results indicate that numerous phenotypic modifications 312 of the liver microvasculature previously associated with the deficient extravasation of both 313 normal and transformed cells of different origin (Guidotti et al, 2015;Valastyan & Weinberg, 314 2011) also occur because of continuous IFNα sensing by HECs, including LSECs. Notably, 315 these microvascular modifications were reverted after the discontinuation of IFNα therapy 316 with no impact on long-term liver functionality/viability (Fig S9C-I). 317 318

HECs acquire an antimetastatic transcriptional profile upon continuous IFNα sensing 319
To confirm the above-mentioned data and to shed new light on the transcriptional changes 320 that HECs adopt to limit CRC trans-sinusoidal migration, we performed RNA-seq analyses 321 on CD31 + endothelial cells isolated from the liver of Ifnar1 fl/fl or VeCad Ifnar1_KO mice 7 days 322 after NaCl or IFNα treatment (Fig S10A). Using SEM to assess the % of CD31 + cells bearing 323 the typical sinusoidal fenestrae, we determined that our preparations contain ~ 96% of bona-324 fide LSECs (Fig S10B), consistently to previous reports (Liu et al, 2011;Su et al, 2021). 325 When compared to HECs isolated from NaCl-treated Ifnar1 fl/fl mice, HECs derived from 326 IFNα-treated animals of the same lineage showed 381 transcripts that were differentially 327 expressed ( Fig 6A). As expected, many of these up-regulated transcripts belonged to the 328 ISG family, including Irf7, Irf9, Mx1, Mx2, Isg15, Stat1 and Oasl1 (Fig 6A,B). Pre-ranked 329 gene set enrichment analyses (GSEA) of IFNα-treated LSECs also revealed a significant 330 enrichment of transcripts involved in interferon signaling or in the induction of varying 331 cytokines and chemokines ( Fig 6C). Several transcripts related to the ECM/GCX 332 organization or the cell-cell/cell-matrix adhesion pathways were upregulated as well (Fig  333   6B,D). Of note, the expression of Itga4 and Itgb2 -previously shown to be increased by IFNα 334 treatment at the protein level ( Fig S8C) -was also enhanced at the transcriptional level (Fig  335   6B). A similar association did not hold true for Icam1 and Sele, suggesting that the increased 336 protein expression we observed earlier ( Fig S8G) occurred independently of transcriptional 337 activity ( Fig 6B). Notably, GSEA also identified gene sets involved in the IFNα-dependent 338 activation of innate and adaptive immune responses or in TCR-dependent signaling 339 pathways ( Fig S10C). 340 Keeping the HECs transcriptional profile of NaCl-treated Ifnar1 fl/fl mice as a point of 341 reference, a total of 566 genes were differentially expressed (DEGs) in HECs isolated from 342 NaCl-treated VeCad Ifnar1_KO mice, of which 373 (mostly ISGs and genes involved in the 343 immune response or in the antigen processing) were downregulated (Fig 6A-C). These latter 344 results indirectly suggest that -when compared to HECs capable of sensing low levels of 345 endogenous type I IFNs, as those present in NaCl-treated Ifnar1 fl/fl mice -LSECs devoted 346 of Ifnar1 may be less prepared to stimulate innate and adaptive immunity ( Fig 6B). The 347 downregulation of transcripts involved in cell-cell adhesion molecules and matrix remodeling 348 (GO) analysis of HECs confirmed that Ifnar1-proficient, but not Ifnar1-deficient, HECs 357 upregulate transcriptional pathways involved in the production of immunostimulatory 358 cytokines and chemokines, the capacity to process and present antigens or the regulation 359 of immune responses ( Fig 6D). Altogether, the data support the hypothesis that, upon IFNα 360 sensing, HECs and particularly LSECs not only acquire a transcriptional profile that can 361 reinforce their barrier function, but they may also enhance HECs/LSECs immunostimulatory 362 functions contributing to antitumor activity. 363 364 Continuous IFNα sensing improves immunostimulatory properties of HECs to 365 provide long-term tumor protection 366 First, HECs/LSECs isolated from the liver of Ifnar1 fl/fl or VeCad Ifnar1_KO mice 7 days after 367 continuous NaCl or IFNα treatment were assessed for the relative surface protein 368 expression of MHC-I, CD86 (a costimulatory molecule (Katz et al, 2004)) or the interleukin 369 6 receptor alpha (IL-6RA, a molecule that LSECs use to properly cross-prime antigens to 370 naïve CD8 + T cells (Bottcher et al, 2014)). Following IFNα treatment, Ifnar1-bearing LSECs 371 significantly increased MHC-I, CD86 and IL-6RA expression (Fig S11A), while no induction 372 was detected in Ifnar1-negative LSECs (Fig S11A). We then analyzed the ability of IFNα-373 treated LSECs or splenic DCs (sDCs) from Ifnar1 fl/fl and VeCad Ifnar1_KO mice to stimulate the 374 cross-priming of naïve CD8 + T cells in vitro. To this end, viable CD31 + HECs and CD11c + 375 sDCs were isolated and purified (Fig S12A,B). sDCs were cultured to acquire mature CD8α + 376 (~25%) or plasmacytoid (45%-50%) phenotypes endowed with cross-priming capacity (Fig  377   S12C,D) (Fu et al, 2020). HECs and sDCs from Ifnar1 fl/fl or VeCad Ifnar1_KO mice previously 378 pulsed with the SIINFEKL peptide or soluble ovalbumin (sOVA) in the presence or absence 379 of either IFNα or NaCl were then co-cultured with naïve OT-I CD8 + T cells and their relative 380 cross-priming capacity was defined by the percentage of these latter cells to express both 381 CD44 and IFNγ ( Fig S11B). IFNα stimulation of Ifnar1-bearing HECs (HECs from Ifnar1 fl/fl 382 mice or sDCs from Ifnar1 fl/fl and VeCad Ifnar1_KO mice) pulsed with SIINFEKL or sOVA promptly 383 increased their cross-priming capacities, while the same IFNα treatment failed to do so in 384 Ifnar1-negative cells (HECs from VeCad Ifnar1_KO mice) (Fig 7A,B and Fig S11B,C). Once 385 exposed to IFNα and pulsed with sOVA, HECs and sDCs from Ifnar1 fl/fl mice cross-primed 386 naïve OT-I CD8 + T cells to a similar extent ( Fig 7B and Fig S11C), highlighting once more 387 the immunostimulating potential of IFNα treatment on HECs, including LSECs. We also 388 evaluated the splenic composition of central memory T cell populations (Tcm, 389 (Fig 7C and Fig S11D) as a proxy of potential systemic memory 390 responses against tumor antigens (Sallusto et al, 2004;Stone et al, 2009;Yu et al, 2019). 391 Splenic naïve T cells (Tn, CD8 + CD44 -CD62L + ) were also evaluated. Looking at Ifnar1 fl/fl or 392 VeCad Ifnar1_KO mice continuously treated with NaCl or IFNα and euthanized by day 21 after 393 challenge, we found that only IFNα-treated Ifnar1 fl/fl mice showed an increased proportion of 394 Tcm and decreased percentage of Tn when compared to NaCl-treated Ifnar1 fl/fl controls (Fig  395  7C), suggesting that IFNα-responsive LSECs may promote antitumor immune memory in 396 secondary lymphoid organs. 397 To assess whether IFNα-stimulated HECs and LSECs promoted memory responses 398 endowed with antitumor potential, Ifnar1 fl/fl -cured mice (defined as animals that 7 days after 399 IFNα treatment initiation were intramesenterically challenged with MC38 cells and survived 400 as disease-free animals until day 50) or naïve Ifnar1 fl/fl control mice were subcutaneously 401 rechallenged with MC38 cells (Fig 7D). Notably, while the latter animals developed 402 subcutaneous tumors that increased in size over time, none of the Ifnar1 fl/fl -cured mice 403 showed detectable lesions at any time point studied ( Fig 7E). These results indicate that 404 continuous IFNα treatment promotes protection against secondary tumor challenge even 405 after IFNα therapy discontinuation. The results also suggest that this effect may be 406 dependent on the capacity of IFN-sensitive HECs and LSECs to foster antitumor immunity, 407 especially tumor-specific effector CD8 + T cell responses that are well-known to control tumor 408 growth in vivo in different experimental settings (Dobrzanski et al, 2000;Katlinski et al., 409 2017;Klebanoff et al, 2005;Yu et al., 2019). In this study, we used different mouse models of CRC liver metastasis to show that the 413 continuous perioperative administration of relatively low IFNα doses provides significant 414 antitumor potential in vivo without provoking overt toxicity. Moreover, under the 415 pharmacological conditions we defined (route, dosage, treatment duration, and chemical 416 nature of the recombinant protein), we did not observe counter-regulatory mechanisms 417 affecting IFNα efficacy (Katlinski et al., 2017), or significant systemic side effects, as our 418 strategy avoids the short tissue-oscillatory IFNα bursts that are often achieved after high 419 and pulsed administrations, often associated with efficacy-limiting toxicities (Weber et al., 420 2015). These results are consistent with previous preclinical work indicating that the 421 intrahepatic delivery of IFNα through a gene/cell therapy approach curbs CRC liver 422 metastases by acting primarily on unidentified non-hematopoietic stromal cell populations 423 (Catarinella et al., 2016). 424 Given the pleotropic nature of IFNα, we demonstrated that the antimetastatic activity of IFNα 425 is neither based on the direct inhibition of primary intracecal tumor growth, favoring the 426 hypothesis that IFNα therapy does not modify the number of cells that spread from primary 427 tumors and seed into the liver -nor on the direct inhibition of metastatic cell growth within 428 the liver. These data is consistent with the high IFNα concentrations required to activate the 429 "tunable" direct antiproliferative functions of this cytokine, likely exceeding the levels 430 achieved in our system (Catarinella et al., 2016;Schreiber, 2017). In addition, IFNα therapy 431 does not require indirect stimulation of hepatocytes, HSCs, DCs, KCs or LCMs to exert its 432 antimetastatic functions. Rather, the results pinpointed HECs/LSECs as key local and early 433 sensors of IFNα that ultimately limit CRC cell invasion into the liver. 434

435
Mechanistically, we showed that IFNα-stimulated LSECs inhibit the trans-sinusoidal 436 migration of circulating CRC cells normally occurring within 24 hours of their initial 437 intrahepatic landing. This effect is associated with phenotypic changes that IFNα-stimulated 438 LSECs acquire or induce in the liver microenvironment. Among these changes, we observed 439 a reduction in the overall LSEC porosity (i.e., sinusoidal fenestrae were reduced in number 440 and size), an enhancement in the subendothelial deposition of basal membrane components 441 (including collagen IV and laminin) and an upregulation of Lyve-1, a marker of hepatic 442 capillarization (Pandey et al., 2020;Wohlfeil et al., 2019). Along these lines, it is noteworthy 443 that in the "healthy" liver, functioning as a common site for CRC metastases, LSECs contain 444 numerous fenestrae of up to 200 nm in diameter and normally lack the typical basal 445 membrane that characterizes the microvasculature of most other tissues and organs 446 (Jacobs et al, 2010). It is also interesting to note that IFNα-stimulated LSECs promote 447 microvascular alterations like those typifying pathological conditions (e.g., initial hepatic 448 capillarization and liver fibrosis (Pandey et al., 2020;Wohlfeil et al., 2019)) associated with 449 impaired immune cell extravasation and reduced immune surveillance (Guidotti et al., 2015) 450 and reduction of hepatic metastases from solid tumors including CRCs (Wohlfeil et al., 451 2019). This fits with the evidence that CRC patients suffering from chronic viral liver fibrotic 452 diseases characterized by hepatic endogenous type I interferon production display lower 453 HECs/LSECs to shape a vascular antimetastatic barrier preventing the interaction between 463 tumor cells and endothelial cells that are known to promote the extravasation of the former 464 cells (Glinskii et al., 2005;Mitchell & King, 2014;Wilkinson et al., 2020). Accordingly, the 465 enhanced expression of "pro-migratory" adhesion molecules and integrins that we observed 466 in the liver of animals bearing IFNα-responsive LSECs appear to be efficiently counteracted 467 by the creation of such vascular barrier. IL-6RA, B2m, Tap1, Psmb-8/9 and H2-d1, H2-k1/2) (Bottcher et al., 2014;Katz et al., 2004;487 Montoya et al, 2002;Rodriguez et al, 1999). Moreover, our results also suggest that IFNα-488 stimulated LSECs may play a key role in antitumor immunity, as mice were protected from 489 secondary tumor rechallenge even after discontinuation of IFNα treatment. The fact that the 490 same IFNα therapy also significantly increased the overall number of central memory T cells 491 in the spleen while decreasing that of naïve T cells (Sallusto et al., 2004;Yu et al., 2019)  sinusoidally migrate into the liver parenchyma and develop micrometastases that will 525 eventually grow overtime, promoting the generation of an immunosuppressive 526 microenvironment. Continuous therapy with well-tolerated doses of recombinant IFNα, 527 stimulates HECs/LSECs to limit CRC trans-sinusoidal migration and parenchymal invasion 528 by building up a vascular barrier typified by the reduction of LSECs porosity, the increased 529 thickness of GCX and the appearance of a basal membrane. Continuous IFNα therapy also 530 promotes long-term antitumor immunity in cured mice and protection from secondary tumor 531 challenge, by stimulating LSECs to efficiently cross-prime tumor antigens to naïve CD8 + T 532 cells ( Fig 7F). and IFNα cell/gene therapy approaches (Catarinella et al., 2016), which could quickly 542 translate our results into clinical practice. Of note, the use of clinically approved doses of 543 pegylated-IFNα has shown improved serum stability and clinical efficacy and reduced side 544 effects, with serum IFNα concentrations similar to those achieved in our system (Foser et 545 al., 2003;Glue et al., 2000). 546 547 All in all, the results of this study support the use of continuous low doses of IFNα as an 548 antimetastatic drug during the perioperative period, due to its ability to transform a 549 metastases-prone liver into a metastases-resistant organ. 550 554

Methods and Protocols 555
Animal studies. Eight-to ten-week-old C57BL/6J and BALB/c mice were purchased from 556 Charles River Laboratory, Calco, Italy. Alb Ifnar1_KO and CD11c Ifnar1_KO and Ifnar1 fl/fl littermates, mice from each litter and cage were 585 randomly allocated into the experimental groups and were co-housed or systematically 586 exposed to beddings of the other groups to ensure the same exposure to the microbiota. anti-coagulated blood of MOP-NaCl and MOP-IFNα-treated mice was collected from the 713 retro-orbital plexus of anesthetized animals (isoflurane, 5% for induction and 2% for 714 maintenance in 2l/minute oxygen) using Na-heparin coated capillaries (Hirschmann 715 Laborgeräte GmbH, Germany) and vials (Microvette, Sarstedt, Germany). Hematologic 716 parameters were evaluated using an automated cell counter (ProCyte Dx, IDEXX 717 Laboratories, USA). The extent of hepatocellular injury was monitored by measuring serum 718 ALT (sALT) activity at several time points after IFNα treatment, as previously described (Sitia 719 et al., 2012). (Avertin), a Y incision was made in the abdomen to expose the liver and the portal vein. The 817 portal vein was cannulated with an appropriately sized IV cannula of 22G and the liver was 818 perfused with 15 ml of warm PBS at a constant rate of 3 rpm using a peristaltic pump (Peri-819 Star™ Pro, 2Biological Instruments) as previously reported (Guidotti et al., 2015). In situ 820 fixation was achieved by perfusing EM fixative for approximately 5 min at 3 rpm. Fixed liver 821 was harvested and cut into 5 mm blocks using a scalpel. Liver blocks were ulteriorly 822 immersed in EM fixative for 24-72 h at 4°C and finally EM fixative was replaced with 0.1 M 823 sodium cacodylate buffer and stored at 4°C until processed for TEM or SEM analysis. Liver 824 blocks were post-fixed in 1% osmium tetroxide (OsO4), 1,5% potassium After 40 days of IFNα discontinuation, 2 randomly selected liver micrographs of IFNα-838 Ifnar1 fl/fl (n=3) were analyzed to determine LSEC fenestra measurements and a total area of 839 approximately 300 µm 2 of sinusoidal surface was analyzed. All measurements were made 840 using the ImageJ software, as previously described (Cogger et al, 2015). Briefly, the 841 flattened area of the liver sinusoid was selected and longest diameter of each fenestrae was 842 measured. Gaps larger than 250nm were excluded from the analysis. The average 843 fenestration diameter (defined as the average of all fenestrae diameters excluding gaps 844 area), the fenestration area (π r2, where r, the radius, was calculated from the individual 845 fenestrae diameter r=d/2, without gaps area), the porosity [ Σ(π r2)/(total area analyzed -846 Σ(area of gaps)) × 100; expressed as a percentage], and the fenestration frequency [total 847 number of fenestrations/(total sinusoidal area analyzed -Σ(area of gaps)); expressed as 848 μm2] have been calculated. Surface roughness analysis of endothelial GCX was determined 849 using Image J software. The flattened area of the liver sinusoid was analyzed, with the 850 strainer on a petri dish containing 10 ml of plain RPMI, ground with a syringe plunger to 886 obtain a cell suspension and washed 3 times with plain RPMI as previously described (Sitia 887 et al., 2012). The cell suspension was centrifugated at 400g for 5 min at room temperature, 888 the resuspended pellet was incubated for 30 sec with ACK lysis buffer and neutralized with 889 RPMI. Resultant splenocytes were counted and processed for FACS analysis. Splenocytes 890 derived from OT1 mice were prepared as described above and naïve CD8 + T cells were 891 isolated using EasySep TM Mouse Naïve CD8 + T cell isolation kit (STEMCELL Technologies) 892 following manufacturer's instructions.

Isolation of HECs, including LSECs and splenic DCs for in vitro studies. Mouse liver 925
perfusion was performed as described in the section Electron Microscopy. After PBS 926 perfusion, liver digestion was achieved in situ by perfusing warm digestion medium at 4 rpm 927 for 10 min. The cava vein was squeezed tight several times to build up some pressure within 928 the liver in order to fill all liver lobes with digestion medium. The resultant digested liver was 929 excised, placed on a petri dish containing digestion medium and the Glisson's capsule was 930 removed. Disaggregated tissue was filtered using a 70 µm cell strainer and centrifuged at 931 50g for 3 min to discard hepatocytes. The supernatant containing NPCs was counted and 932 Kupffer cells (KC) were removed using anti-F4/80 Ultrapure microbeads (Miltenyi Biotec) 933 following manufacturer's guidelines. The flow-through of unlabeled NPCs was placed on top 934 of a 25%-50% Percoll gradient and centrifuged at 850g for 20 min at RT without brake and 935 the LSECs located at the 25/50% interface were collected. LSECs were counted and 10 5 936 cells were seeded in a 48-well plate and cultured in collagenated plates with EGM-2 medium 937 (Lonza) for 3 days. For the isolation of splenic DC, spleens were slowly injected with 1ml of 938 digestion medium until they changed from dark maroon to reddish-orange color. Then, the 939 spleen was minced and pipetted vigorously several times in digestion medium. The cell 940 suspension was filtered using a 70 µm cell strainer, and larger undigested fragments were 941 ulteriorly incubated with digestion medium at 37°C for 30 min. After tissue digestion, 942 splenocytes were centrifuged at 500 g for 5 min and washed 3 times with plain RPMI 943 supplemented with 5mM EDTA to disrupt DC-T cell complexes. Red blood cells were lysed 944 with ACK lysis buffer and DCs cells were isolated using anti-CD11c microbeads UltraPure 945 (Miltenyi) according to manufacturer's instructions. CD11c + DCs were counted and seeded 946 at 10 5 cells in a 48 well-plate using RPMI GlutaMAX supplemented with 10% FBS, 1% P/S, 947 1x Na Pyruvate (Gibco), 1x MEM Non-essential Amino Acid Solution (Gibco), 20 µM β-948 mercaptoethanol and 40ng/ml GM-CSF (BioXcell). Every 48h 200 µl of fresh medium were 949 added to cultured cells and DCs were grown for 7 days to stimulate DCs differentiation. 950 951 Antigen cross-priming. Prior to naïve CD8 + T cell co-culture, HECs, including LSECs, and 952 sDCs were stimulated for 18 h with 1µg/ml of SIINFEKL peptide (OVA 257-264, Proimmune) 953 or with 1mg/ml of low endotoxin soluble ovalbumin protein (sOVA; Sigma-Aldrich) in 954 combination with NaCl or 5 ng/ml of IFNα. Subsequently, after extensive washes, cells were 955 co-culture with 5x10 5 naïve CD8 + T cells isolated from OT-I mice in complete RPMI 956 (containing 10% FBS and 50 µM β-mercaptoethanol) in combination with NaCl or 5 ng/ml of 957 IFNα. After 3 days, CD8 + T cells were stimulated for 4 h with 1µg/ml of SIINFEKL peptide, 958 5 µg/ml of Brefeldin A (Sigma-Aldrich) and 2.5% EL-4 supernatant in complete RPMI. 959 Finally, the production of IFNγ and the expression of activation markers, such as CD44, 960 were measured by FACS. Statistical significance was estimated by two-tailed non-parametric Mann-Whitney test (e.g. 995 to evaluate differences generated as a consequence of tumor growth) or by one-way 996 ANOVA with Tukey's multiple comparison test when more than two groups were analyzed. 997           NaCl IFNα