Dendritic cells overcome Cre/Lox induced gene deficiency by siphoning cytosolic material from surrounding cells

Summary In a previous report, keratinocytes were shown to share their gene expression profile with surrounding Langerhans cells (LCs), influencing LC biology. Here, we investigated whether transferred material could substitute for lost gene products in cells subjected to Cre/Lox conditional gene deletion. We found that in human Langerin-Cre mice, epidermal LCs and CD11b+CD103+ mesenteric DCs overcome gene deletion if the deleted gene was expressed by neighboring cells. The mechanism of material transfer differed from traditional antigen uptake routes, relying on calcium and PI3K, being susceptible to polyguanylic acid inhibition, and remaining unaffected by inflammation. Termed intracellular monitoring, this process was specific to DCs, occurring in all murine DC subsets tested and human monocyte-derived DCs. The transferred material was presented on MHC-I and MHC-II, suggesting a role in regulating immune responses.


INTRODUCTION
Dendritic cells (DCs) are the critical link between innate and adaptive immunity.At steady state, DCs in peripheral tissue scavenge their surroundings for antigens in the form of apoptotic bodies, cell debris, and extracellular vesicles.If they encounter molecules that ligate pattern recognition receptors, they become activated, upregulate costimulatory molecules and MHC-II on their surface, and migrate to lymph nodes to present antigens to adaptive immune cells. 1,2When cells die or release material into the extracellular environment, DCs are thought to acquire it through multiple endocytic processes, including receptor-mediated endocytosis, phagocytosis, and macropinocytosis, which they conduct at high rates. 3However, more recent literature has challenged this notion of DCs as scavengers by showing that DCs acquire and cross-present antigens equally well from live cells as they do apoptotic. 4,5Further, compared to DCs that acquire antigen from apoptotic cells, DCs acquiring antigen from live cells generate larger CD8 + T cell responses and increased protection from lethal tumor challenge when injected in vivo. 6Separating DCs from live donor cells with a 0.45 mm pore size transmembrane insert prevents cross-presentation, indicating live cell contact-dependent antigen uptake is critical for inducing an adaptive response. 4mportantly, by acquiring material from live cells, DCs can interact with a variety of molecules that are not usually present in scavenged material.For example, mRNA is degraded as cells go through apoptosis, 7 and the total volume of material that can be transferred through extracellular vesicles is restricted by their small size.By circumventing these restrictions, DCs can contain large, functionally relevant quantities of RNA and protein from their surroundings.The immunological impact of such transfer has been observed in numerous contexts.Antigen acquired by metallophilic marginal zone macrophages in the spleen is actively transferred to DCs, which can promote or suppress adaptive immunity depending on context. 8A similar transfer is seen between CXCR1+ macrophages and CD103+ DCs in the context of oral tolerance, 9 and macrophages are known to siphon cytosolic material from stem cells during quality control checks. 10Our lab found that epidermal Langerhans cells (LCs) contain certain KC-derived mRNA at nearly 50% of the level present in KCs themselves. 11Aside from its biological importance, high volume transfer of material from one cell type to another is of concern for researchers utilizing conditional knockout animals.
Cre/Lox animal models are a standard tool for deleting genetic regions in specific cell types.This is accomplished by placing the expression of the bacterial recombinase Cre under the control of a cell type-specific promoter.When expressed, Cre will act on two short LoxP sequences that have been inserted into the gene of interest, resulting in cell type-specific gene disruption.However, if DCs can acquire a large enough quantity of material from neighboring cells, the efficacy of Cre/Lox models may be undermined.Our lab has already shown that DCs can acquire Cre expressed by neighboring cells, potentially resulting in off-target effects, 11 but it remains to be seen whether the material transfer can overcome DC-specific gene deletion.If so, many DC-specific conditional knockout models may be non-functional at the protein level despite successful genetic recombination.Considering the broad usage of such models, further investigation of this concern is warranted.

Intracellular material acquisition from surrounding cells is specific to DCs and universal among all DC subsets tested
The data above suggest that DCs can acquire cytosolic material from different cell types.To determine which cell types DCs monitor, we used a modified in vitro co-culture system that we developed to image RNA acquisition by LCs from epidermal KCs. 11Briefly, we co-cultured GFPexpressing MutuDC1 cells (DC cell line with cDC1 phenotypical and functional characteristics) 27 with SYTO62-labeled (nucleic acid dye) B cells, T cells, peritoneal macrophages, or dermal CD45 À cells sorted from adult wild-type naive C57BL/6 mice (Figures S2A and S2B).Results show that while DCs acquire RNA from all cell types tested, they most efficiently siphon from macrophages and non-hematopoietic CD45 À stromal cells (Figure 2A).Thus, DCs can monitor both hematopoietic and non-hematopoietic cells.
To better understand the role of intracellular material acquisition, we sought to determine if it is exclusive to DCs, or a property of many cell types.To do that, we co-cultured the sorted T cells, B cells, macrophages and CD45 À stromal cells with RNA labeled keratinocytes from the murine cell line COCA. 28We observed that RNA transfer ranged from significantly less efficient (macrophage and stromal cells) to almost entirely absent (T and B cells) when using cell types other than DCs as recipients (Figure 2B).In addition, we observe minimal RNA transfer when MutuDC1 cells are used as donors, indicating that transfer is mostly unidirectional in favor of DCs (Figure S2C).These findings support that intracellular monitoring is a unique DC property.We previously reported that DC subsets harbor mRNAs specific to their tissue of residence. 11To bring experimental evidence that other DC subsets can acquire RNA from the surrounding cells, we sorted DC subsets from single-cell suspensions generated from the epidermis, skin-draining lymph nodes, and mesenteric lymph nodes of adult naive mice (Figure S2D) and co-cultured them with RNA labeled COCA KCs.We found that all DC subsets of the epidermis and skin-draining lymph nodes were able to acquire RNA from KCs to varying degrees (Figure 2C), while mesenteric CD11b+CD103+ DCs acquired some, but significantly less RNA-aligning well with their lesser ability to overcome CX43 knockout relative to eLCs.Therefore, these data support that the material acquisition from the target cells by DCs is widespread and not limited to LCs or MutuDC1s.
We previously showed that human LCs, similar to mouse LCs, contain KC-derived keratins. 11To provide support that intracellular material acquisition is conserved in humans, we determined whether closely representative human monocyte-derived DCs (moDCs) can acquire RNA from other cells.We differentiated DCs from CD14 + blood monocytes (Figure S2E) and incubated them with autologous PBMCs labeled with RNA dye at 37 C or on ice.We found that the moDCs were efficient in acquiring RNA from the autologous PBMCs at 37 C (Figures 2D and  S2F).Thus, these data support that intracellular material acquisition by DCs exists in humans.

Intracellular material transfer is contact-dependent, but independent of known antigen acquisition pathways
We previously observed that LCs separated from KCs using 0.4 mm Transwell membrane are unable to acquire detectable levels of KC-derived RNA. 11These data suggest that free RNA, exosomes, other forms of extracellular vesicles and/or cell debris that could cross the membrane might not play a significant role in RNA transfer from KCs to LCs.However, some RNA containing microvesicles are larger than 0.4 mm, 29 and the Transwell membrane may nonspecifically bind some vesicles and therefore hinder their access to LCs.To overcome these caveats and confirm the requirement for physical interaction for RNA transfer from KCs to DCs, we adapted an in vitro co-culture system 30 where donor cells (KCs) are suspended above recipient cells (DCs) (Figure 3A).In this system, the two cell types are facing each other and are only separated by a thin layer (1.5 mm) of cell culture media.This setup provides DCs unobstructed access to KC-derived exosomes, vesicles, cell debris, and apoptotic cells, while still preventing direct contact to live adherent cells.For our system, MutuDC1 cells 27 were seeded at the bottom of a 48-well plate and murine COCA keratinocytes were grown on round coverslips, labeled with the nucleic acid dye SYTO62 and protein dye cell trace violet (CTV), and suspended above the DCs using silicone O-rings throughout the entire culture period (''physical separation'') (Figures 3A and S3A).Physical separation again prevented transfer of RNA from KCs to DCs as measured by flow cytometry and confocal microscopy, whereas DCs that had direct physical contact with the KCs at 37 C, but not on ice, acquired KC-derived RNA (Figures 3B, 3C,  and S3).Time-lapse imaging shows RNA and protein signal intensity increasing within DCs over time during direct contact (Figure S3B).Imaging also revealed that only DCs in direct contact with donor cells acquired RNA (Video S1).Together, these data strongly support that released cell vesicles, debris, and apoptotic bodies do not play a significant role in RNA transfer, and they further underpin the need for physical contact for RNA and protein transfer.
To gain insight into the physical interaction and mechanism that allow DCs to acquire cytosolic material from other cells, we co-cultured COCA keratinocytes with MutuDC1 cells and took confocal images and time-lapse videos of these cells interacting with one another.The physical interaction between the DCs and KCs was diverse in nature, ranging from superficial-looking touching/screening all the way to DC dendrites pressing into the KC plasma membrane (Figure 3D and Videos S2, S3, S4, and S5).Occasionally, DCs formed ring structures when contacting KCs, resulting in RNA containing vesicles.(Figure S3C and Video S3).
Having established physical interaction as a requirement for material transfer, we next tested whether previously described, standard routes of antigen acquisition, such as phagocytosis, tunneling nanotubes, macropinocytosis, and gap junctions are involved in cytosolic material acquisition by DCs.To do this, we measured RNA transfer by flow cytometry from RNA labeled COCA KCs to MutuDC1 cells during coculture under different conditions.Transfer was significantly inhibited if they were treated with the ATP synthase inhibitor Oligomycin A, 31 demonstrating that the process is energy-intensive, rather than a passive transfer (Figures 3E and S3D).It was recently established that tunneling nanotubes (TNTs) enable significant RNA transfer between stationary cells, 30 and we have observed structures resembling TNTs between DCs in some of our long-term (more than 45 min) co-cultures (Figure S3E).TNTs require intact F-actin, 32 and their formation can be inhibited with low concentrations (50 nM) of cytochalasin D, an F-actin inhibitor.Phagocytosis of detached cells or cell debris was not prevented in our physical separation system, so phagocytosis is unlikely to be contributing substantially to material transfer, however, at micromolar concentrations cytochalasin D also inhibits phagocytosis, 33,34 so its use can further exclude phagocytosis as the dominant mechanism of transfer.Indeed, while cytochalasin D was highly effective at disrupting F-actin (Figure S3F) and inhibiting the uptake of 2 mm beads (Figure S3G), we only observed a minor inhibition (19%) of RNA transfer (Figure 3E).Increasing doses of cytochalasin D up to 100 mM did not further inhibit transfer (Figure S3H).We next sought to evaluate the contribution of macropinocytosis.5-(N-ethyl-N-isopropyl)-Amiloride (EIPA), a Na + channel inhibitor known to block macropinocytosis, 35 did not inhibit RNA acquisition (Figure 3E), but did block fluorescent dextran uptake by DCs (Figure S3I), which is mediated partially by macropinocytosis. 36The gap junction inhibitor 1-heptanol 9 also failed to inhibit RNA acquisition (Figure 3E), but did significantly reduce transfer of Calcein dye between COCA KCs (Figure S3J), which is partially mediated by gap junctions. 37Our findings were not limited to RNA acquisition from KCs by MutuDC1s.We observed roughly similar responses with MutuDC1s or primary splenic DCs combined with an unrelated cancer cell line, B16 (Figures S3K and S3L).Therefore, known mechanisms of material uptake such as macropinocytosis, TNTs, phagocytosis, and gap junctions do not appear to play a major role in RNA transfer.

DCs acquire cytosolic material from other cells through a mechanism dependent on calcium and PolyG-blockable receptors
Intercellular interactions are mediated by surface receptors that often rely on Ca 2+ for binding. 38Thus, we next tested whether extracellular Ca 2+ plays a role in intracellular material acquisition by DCs.We supplemented the DC/KC or DC/B16 co-cultures with 5 mM EDTA to chelate extracellular Ca 2+ .We observed a significant inhibition of material transfer for both DC/KC (Figures 4A and S4A) and DC/B16 (Figure S4A) co-cultures.The inhibition reached maximum with 5 mM EDTA (Figure S4B).To determine whether intracellular Ca 2+ also plays a role in intracellular monitoring, we supplemented the EDTA treated DC/B16 co-cultures or the Ca 2+ -free media with thapsigargin or BAPTA-AM.Adding thapsigargin, a non-competitive irreversible inhibitor of the endoplasmic reticular Ca 2+ ATPase that is often used to deplete intracellular Ca 2+ , 39 or BAPTA-AM, a cell membrane permeable Ca 2+ chelator, to Ca 2+ -free media had additive effects, leading to an overall 60-70% inhibition of RNA transfer (Figure 4B).Thus, these data suggest a mechanism partially dependent on extracellular and intracellular Ca 2+ .Cadherins and integrins play an essential role in cell adhesion, synapse formation, and intercellular interactions in general, and some are Ca 2+ dependent. 38Therefore, we next tested the contribution of certain, well-characterized cadherins and integrins to the RNA transfer.The DC/target cell co-cultures were supplemented with blocking antibodies to E-cadherin, CD11b, CD11c, RGD peptides (to block RGDbinding integrins), or ADH-1 (small molecule inhibitor of N-cadherin).We found no significant inhibition with any of the reagents tested (Figure S4C).The binding and potency of the antibodies and RGD peptides were confirmed prior to use (Figures S4D-S4F).These data suggest that the integrins and cadherins tested here do not play a substantial role in material transfer.
Protease mixtures, such as Pronase, that can digest a wide range of proteins, are often used to confirm the involvement of cell surface proteins in cellular interactions. 40To test whether RNA acquisition by DCs is Pronase sensitive, we treated the MutuDC1s with Pronase as previously described. 40The effect of Pronase digestion on cell surface proteins was confirmed by flow cytometry using markers such as CD8 (sensitive), CD11c, CD11b (partially sensitive), and MHC-II (resistant) (Figure S4D).Pronase treatment of the DCs caused slight (roughly 30%) but significant inhibition of RNA transfer (Figure 4C), supporting the involvement of a Pronase sensitive DC surface protein in material transfer.Pronase treatment of DCs has been reported to inhibit trogocytosis of target cell membrane by degrading class A scavenger receptor CD204, which can also be blocked by the molecule polyguanylic acid (PolyG). 40,41Adding PolyG to intact DC/B16 co-cultures led to roughly 50% inhibition of RNA transfer (Figure 4C).The difference in percent inhibition between Pronase and PolyG indicates that the two treatments act through different receptors, and that PolyG likely acts through a receptor other than CD204, as this receptor is Pronase sensitive 40 and not expressed by the MutuDC1 cell line used here (Figure S4G).Blocking other scavenger receptors that are expressed by MutuDC1s, such as DEC205 and CD36L1 (SR-B1), 27 resulted in negligible inhibition (Figure S4H).Pronase treatment in combination with PolyG caused near complete inhibition of RNA transfer (Figure 4D), supporting that these two treatments act through different receptors with partially redundant functions.Considering this redundancy, it is possible that the cadherins, integrins, and other receptors tested above do mediate material transfer, but that this effect can only be observed when they are blocked in combination with PolyG.After testing, apart from CD11c, we found that blocking candidate receptors in conjunction with PolyG had no effect over PolyG alone (Figures S4I-S4J).PolyG in combination with antibody binding MHC-II, an abundant surface protein that is not degraded by Pronase (Figure S4D), resulted in modest additive inhibition similar to anti-CD11c, indicating this inhibition likely reflects general steric effects and not a specific mechanism (Figure S4K).
Next, we probed whether Ca 2+ works in concert with the Pronase-sensitive or PolyG-sensitive receptor.Interestingly, in combination with PolyG, but not with Pronase-treated DCs, EDTA almost completely inhibited RNA transfer (Figure 4D).PolyG+EDTA inhibition of RNA acquisition was effective with either KCs or B16s as donors (Figure S4L), and also inhibited the majority of protein transfer from B16s (Figure S4M).PolyG+EDTA also inhibited RNA transfer to human DCs from PBMCs (Figure S4N).Combining LY294, a PI3K inhibitor, with PolyG, but not EDTA showed additive effect in mouse cell cultures (Figures 4E and S4O).Thus, these data suggest a mechanism dependent on Ca 2+ and that the Pronase-sensitive receptor on DCs is likely Ca 2+ -and PI3K-dependent.Overall, these data support that at least two sets of receptors mediate the RNA and protein acquisition by DCs.

DCs present the antigen acquired through intracellular monitoring on both MHC-I and MHC-II
We found that PolyG in concert with EDTA blocked monitoring with high efficiency, and that previously described antigen acquisition routes did not substantially contribute to RNA and protein transfer from target cells to DCs in our model.PolyG in combination with EDTA did not inhibit macropinocytosis (Figure S5A) but did inhibit phagocytosis in peritoneal macrophages (Figure S5B).However, because MutuDCs showed very poor phagocytic capability (Figure S5B), and inhibition of phagocytosis by cytochalasin D did not prevent RNA transfer (Figures 3E and S3H), PolyG/EDTA can be considered a specific inhibitor of intracellular monitoring in this system and used to address the immunological role of this unique antigen acquisition pathway.We first determined whether the acquired protein is presented on MHC-I.For this purpose, we co-cultured MutuDC1s, known to efficiently cross-present, 27 or MutuDC2s, unable to cross-present, 42 with B16 or B16-OVA cells for different time points.Then, we determined the presentation of the SIINFEKL peptide by DCs using peptide/ MHC-I-specific antibody (Figure 5A).We found detectable levels of SIINFEKL peptides on the MutuDC1s co-cultured with B16-OVA, but not B16, as early as 1 h after co-incubation, which increased with time (Figure 5B).In contrast, with MutuDC2s we failed to detect any significant SIINFEKL presentation at any of the time points tested (Figure 5B).Inclusion of PolyG/EDTA in co-cultures significantly blocked SIINFEKL presentation by MutuDC1s (Figure 5C).The lack of cross-presentation was not due to the failure of MutuDC2 to perform intracellular monitoring; MutuDC2s, albeit less efficient than MutuDC1s, acquired significant amounts of RNA from both B16 and B16-OVA cells (Figure 5D).Thus, these data support that specific DC subsets specialized in cross-presentation can process and present antigen acquired through intracellular monitoring on MHC-I.
To determine whether the acquired antigens can be presented on MHC-II, we took advantage of the YAe antibody.The YAe antibody recognizes the Ea peptide presented in the context of I-Ab expressed by B6 mice.Ea peptide is derived from BALB/c MHC-II.Thus, we flow-sorted T and B cells from BALB/c skin-draining lymph nodes and co-cultured them with B6-derived MutuDC1 and MutuDC2 in the presence or absence of PolyG/EDTA.The rationale behind this setting was that the cells from BALB/c mice would serve as a source of Ea peptide.If the B6 DCs can take up MHC-II from the BALB/c cells, process, and present the resulting Ea on their MHC-II, then they should turn YAe positive.We found that MutuDC1s could present detectable amounts of Ea when co-cultured with B cells, but not with T cells, and that this presentation was significantly reduced in the presence of PolyG/EDTA (Figures 6A and 6B).In contrast, MutuDC2s did not present detectable amounts of Ea when co-cultured with either B or T cells (Figures 6A and 6B).

DC maturation induced by inflammatory signals do not affect intracellular monitoring
4][45] To test whether acquisition of cytosolic material through intracellular monitoring is affected by maturation signals, we exposed the MutuDC1s and MutuDC2s for 12 h to 1 mg/mL LPS or 10 mg/mL IFNa or 5 mg/mL PolyI:C, or 0.5 mM CpG.Both cell lines express the receptors for these ligands, and they can respond to these stimuli by upregulating co-stimulatory markers. 27,42We also confirmed maturation through morphological changes and upregulation of specific markers.Representative data can be found in Figure S6.Then, the exposed and non-exposed DCs were compared side-by-side in an intracellular monitoring assay.We found no significant differences between treated and non-treated DCs in acquiring RNA from the target cells and between different treatments (Figure 7A).To mimic tissue inflammation and determine whether DCs entering an inflamed tissue could perform intracellular monitoring, we exposed the B16 target cells, known to express TLR-4, 46 to LPS or IFNa for 12 h, then used them as target cells in our assay.The MutuDC1s were equally able to monitor both steady-state B16s and B16 exposed to inflammatory stimuli (Figure 7B).Thus, these data support that inflammatory conditions do not affect intracellular monitoring, which further separates it from other antigen acquisition routes.

DISCUSSION
Herein, we show that epidermal LCs overcome specific gene deletion when neighboring cells contain the missing gene product.Whereas MHC-II gene deletion in LCs results in a depletion of the corresponding protein, deleting genes coding for Cx43 and MyD88, expressed by neighboring KCs, does not decrease the quantity of gene products in knockout LCs.CXCR5 and MHC-II deletions are, however, overcome after LC migration to the CXCR5 and MHC-II rich skin draining lymph nodes. 24CD11b+CD103+ mLN DCs similarly overcome Cx43 conditional deletion, demonstrating that this trait is likely shared with DCs.After in vitro co-culture with RNA labeled donor cells, all primary DC subsets tested acquired RNA from neighboring cells to some extent.Co-cultures with different cell types revealed that DCs acquire RNA from a broad range of cell types, but cell types other than DCs acquire RNA at substantially lower rates.Human moDCs were also able to acquire RNA from autologous donor cells.Investigation into the mechanism of RNA transfer revealed it to be dependent on close contact and an active process, as physical separation, or direct contact while on ice results in near complete inhibition of material transfer.Live cell time-lapse confocal imaging shows DCs pressing dendrites into the membrane of donor cells and maintaining close contact.Actin cytoskeletal inhibition with cytochalasin D, which is known to prevent most forms of phagocytosis, endocytosis, trogocytosis, and the formation of tunneling nanotubes, 34,[47][48][49][50][51] does not substantially prevent transfer.RGD peptides or blocking antibodies against integrins commonly involved in endocytic process such as CD11c and CD11b 48 also have no effect on transfer.EIPA, an inhibitor of macropinocytosis, and 1-heptanol, a gap junction inhibitor, also fail to prevent transfer.Instead, we find that transfer is partially inhibited by removing calcium from the media, PI3K inhibition, or by the introduction of PolyG into co-cultures.Combining PolyG with EDTA or Pronase treated DCs, but not EDTA with Pronase-treated DCs, is sufficient to block most of the transfer.Transferred material is successfully presented and cross-presented on MHC-II and MHC-I, and occurs between allogenic donor and acceptor cells.Inducing DC maturation with various inflammatory stimuli did not influence the material transfer observed here.Due to its discordance with conventional means of antigen uptake, we termed this route intracellular monitoring (ICM).Our finding that LCs and DCs can overcome gene deficiencies has important implications for researchers utilizing conditional knockout models targeting DCs.Specifically, it emphasizes the importance of verifying protein depletion in addition to genetic recombination.Failing to do so may increase the likelihood of type II error, as the acquisition of protein from neighboring cells may lead researchers to incorrectly conclude that depleting the protein of interest has no effect, when in reality, lost protein was simply replaced through ICM.Furthermore, these data raise serious concerns regarding gene expression databases on DCs, which, based on our data, likely represent a mixture of mRNA from DCs and local cells.This highlights a need for the curation of RNA-seq data.
Aside from the immediate practical concerns surrounding ICM and conditional knockouts, ICM may also be relevant to important biological functions such as microenvironmental adaptation, immunosurveillance, and tolerance.Immune cell adaptation to the local microenvironment is a concept that has been extensively studied in macrophages, and refers to the dramatic shift in the chromatin landscape of macrophages in response to environmental queues such as retinoic acid or heme. 52These changes endow macrophages with functions necessary to operate properly in their local niche, and contribute to, instead of interrupt, the function of their resident organ. 53ICM may be providing a similar benefit to LCs.While we did not directly test the functionality of transferred protein in this study, one out of many viable explanations for LC's possession of Cx43 is to prevent the disruption of wound healing, which is dependent on the direct transfer of Ca 2+ , IP 3 , and ATP through gap junctions and the ensuing calcium waves. 54,55These waves are projected to travel through LCs as well as KCs, 56 supporting that LCs might acquire functional protein to help them adapt to their environment.Further studies investigating the fate and function of transferred material will help elucidate the roles of ICM in microenvironmental adaptation.Considering DCs overcome the deficiency of multiple proteins, including ones they express on their own, it is possible that DCs continually and non-specifically conduct ICM.Among cell types tested, ICM was specific for DCs, and, to a lesser degree, macrophages.Its specificity for DCs and macrophages, combined with the finding that acquired protein is presented and cross-presented, points toward ICM being highly relevant to typical DC functions such as immunosurveillance and tolerance, and may explain the long-standing mystery of how DCs receive material from other cells for cross-presentation. 4,5,57DCs can monitor all donor cells tested, but more efficiently monitor CD45 À cells and macrophages.If DCs use ICM to detect pathogens, monitoring macrophages with high efficiency would provide an evolutionary advantage, as these cells are often the first to encounter pathogens, and are more likely to contain a diverse pool of antigens.The low monitoring efficiency of the CD103+ cDC2 (CD11b+CD103+) mesenteric DCs, which migrate from the predominantly tolerogenic environment of the gut and are involved in T reg and T h17 cell induction, 2,58 supports this and argues against a role for ICM in maintaining tolerance, though this cannot be ruled out.On the other hand, it is also possible that while they are in the lamina propria of the gut, the very same DCs might possess high ICM capability, then downregulate it by the time they reach the mesenteric LNs to protect the cargo that requires tolerance induction.While this remains to be experimentally tested, we found that the opposite is true for LCs.LCs that have migrated to LNs are more efficient in ICM than their peripheral counterparts in the epidermis.Whether these site-specific differences have evolved to better serve tolerance induction or simply reflect that ICM is a tool for pathogen detection or that the monitored cell type and environment in the periphery will imprint a downstream program in the DCs, remains to be addressed.
From an evolutionary standpoint, it is logical that DCs would use ICM to detect pathogens.0][61] Relying on material released or presented by infected cells is, therefore, not a dependable way to detect meddling pathogens.Direct presentation after infection also cannot be relied on, as not all viruses are DC tropic or highly cytopathic, and even if they are, DCs themselves could be subjected to pathogen immune evasion mechanisms, resulting in inefficient presentation.Thus, ICM may be a counter to the evading mechanisms developed by the pathogens.Our finding that ICM efficiency is unaffected by inflammatory stimuli or maturation is sensible in this context.A continuous and invariable monitoring system would be more difficult to distort than one that is regularly modulated.Further, if ICM facilitates inflammatory immune responses, our observation that DCs perform it in allogeneic and xenogeneic (unpublished observation) settings suggests it may play a role in organ rejection and be a valid therapeutic target.
While the exact mechanism of ICM remains to be determined, experiments conducted herein sufficiently differentiate from known processes of material transfer.Its contact-dependent nature rules out the uptake of extracellular material as a major contributing factor of RNA and protein transfer.Phagocytosis of dead or dying cells can be ruled out as dying cells are not prevented from detaching and coming in contact with DCs in our physical separation experiments.Cultured cells also maintained high viability throughout experimentation, and no donor cells or debris were observed in contact with DCs in physical separation experiments.TNTs are notoriously fragile and can be eliminated with doses as low as 50 nM Cytochalasin D, 62 ruling out their involvement in material transfer.While some aspects of intracellular monitoring are reminiscent of trogocytosis, our findings are not consistent with this mechanism.Trogocytosis has been successfully inhibited by PolyG in DCs 40 and LY294002 in other cell types, 47 similar to what we observed, however, Harshyne et al. show that PolyG inhibits trogocytosis through blockade of Scavenger receptor A (CD204), 40 which the MutuDC1 cells used in this study do not express.Further, the actin cytoskeleton is required for trogocytosis, whereas cytochalasin D fails to substantially prevent intracellular monitoring.Finally, trogocytosis almost exclusively refers to the transfer of membrane between cells, 63 not the transfer of cytosolic material, further differentiating our observation of RNA and protein transfer from trogocytosis.Considering their size limitations, it is very unlikely that gap junctions would enable substantial RNA and protein transfer.Interestingly, in their study of oral tolerance, Mazzini et al., find that gap junctions only partially mediate transfer from macrophages to DCs, and note that ''still-unknown mechanisms'' may be contributing. 9In retrospect, it is likely that at least some of the material transfer from macrophages to DCs reported by Mazzini et al., was through ICM.ICM is further separated from other routes of antigen acquisition, such as phagocytosis, endocytosis, and macropinocytosis in that it is not significantly altered by the inflammatory signals that induce DCs maturation (Figure 7).4][45] However, our data aligns with some of the in vivo findings showing that matured DCs remain efficient in acquiring soluble antigens. 64The fact that in our platform, we rarely detected phagocytic DCs (Figure S5B), and that ICM was minimally affected by actin cytoskeletal drug, further support that ICM might be the dominant route of acquisition of soluble cytosolic antigens both in vitro and in vivo.
We previously reported that human LCs, like their mouse counterpart, also contain detectable levels of Krt14 mRNA. 11Here, we further showed that human moDCs differentiated from CD14 + monocytes efficiently acquire RNA from PBMCs, and that RNA transfer can be significantly inhibited by PolyG/EDTA.These data support the translatability of our mouse data and indicate ICM may be a conserved process.
In summary, we show that a widely used research tool-Cre/Lox conditional gene knockout-may be inherently flawed when applied to DCs due to their ability to acquire material from neighboring cells through intracellular monitoring.

Limitations of the study
Co-culture assays measuring the ability of various cell types to act as RNA donors or acceptors (Figure 2) would benefit from an expanded panel of cell types.Investigating alternative donor cells to the COCA KCs used here would be particularly helpful to ensure the trends identified in this manuscript are conserved in other settings.Some functional ICM assays were only performed on cell lines.Further experiments will be needed to establish whether ICM has different functional characteristics in primary mouse and human DC subsets.suspended coverslips were completely submerged.After a 45 min incubation at either 37 C or on ice, MutuDC1 cells were resuspended by pipetting up and down, and transferred RNA was measured by flow cytometry.The same protocol was conducted using ibidi 8 chamber slides (Fisher Scientific) to allow for confocal imaging with the exception that a positive control condition was included (Dye in Media) where SYTO62 was not washed out.Images were taken on a Nikon A1R Confocal microscope using a Plan Fluor 403 Oil objective at the end of the 45 min incubation.The amount of transferred RNA was measured in ImageJ by calculating the mean far-red pixel intensity (SYTO62 signal) contained within regions of high green channel signal (representative GFP + MutuDC1s).Briefly, green channel images were converted to 8-bit and thresholded appropriately.Watersheding was used to parse clumped cells, then the analyze particle's function was used to identify Regions of Interests corresponding to area within MutuDC1 cells.These ROIs were then applied to the far-red channel and mean pixel intensity was calculated.

Study of RNA transfer in the presence of inhibitors
Calcein labeled cells were combined in the wells of a 96 well flat bottom plate in media containing either 5 mM 1-Heptanol or vehicle control and incubated for 2 h at 37 C. Calcein transfer to Tag-it Violet+ cells was measured by flow cytometry by comparing to unstained controls.

RGD peptide adherence assay
B16 cells were suspended in cold B16 media with or without 1 mg/mL RGD peptides.Two hundred and fifty thousand cells were then plated per chamber of an 8-chamber collagen I-coated microscope slide (Corning).Cells were incubated for 45 min at 37 C and 5% CO 2 .After incubation, the media was removed and replaced with PBS, and cells were imaged.The PBS was then removed and combined with initial as the non-adherent fraction, and 200 mL warmed Trypsin-EDTA (0.25%) was added to each well and incubated for 5 min.Two hundred microliters of media were then added to each chamber and pipetted up and down to ensure the removal of all adherent cells.This was considered the adherent fraction.

QUANTIFICATION AND STATISTICAL ANALYSIS
Data were analyzed by unpaired two-tailed Student's t test for parametric data, and one-way analysis of variance (ANOVA) followed by Tukey's post hoc test for multiple comparisons.Data normality was determined using the Shapiro-Wilks test.Data displayed as Mean G SD.Additional information is reported in figure legends.GraphPad Prism software was used for the analyses (GraphPad Software, La Jolla, CA).

Figure 1 .
Figure 1.DCs can overcome gene deficiency (A) Gating strategy for identifying keratinocytes and Langerhans cells in an epidermal cell suspension.Epidermal cell suspension from a WT mouse was stained for MHC-II (top) or Cx43 (bottom) or corresponding isotype controls.KCs: keratinocytes, LCs: Langerhans cells.Isotype control signal shown for keratinocytes.Representative flow plots.(B) Expression of the indicated proteins by epidermal Langerhans cells derived from MHC-II or Cx43 hLangCre conditional knockout mice.Representative flow plots and summary graphs.Dots represent individual mice from one of three independent experiments.(C) myd88 mRNA level relative to housekeeping gene gapdh in Cre-or Cre+ LCs sorted from hLangCre MyD88 f/f or global myd88 knockout epidermis, quantified by qPCR.Dots represent individual mice from one out of three independent experiments.MyD88 KO mice included in one of three repeat experiments.(D) Representative CXCR5 flow staining of splenic DCs from a global CXCR5 knockout mouse (KO), or skin draining lymph node migratory LCs from either Cre-(blue) or Cre+ (green) hLangCre CXCR5 f/f mice, or isotype control staining of skin draining lymph node migratory LCs from Cre-hLangCre CXCR5 f/f mice, and summary graphs.Dotted line represents isotype control.Dots represent individual mice from one of two independent experiments.(E) Cx43 flow staining of CD11b+CD103+ mesenteric lymph node DCs derived from either Cre-or Cre+ hLangCre Cx43 f/f mice, or isotype control staining of Cre-hLangCre Cx43 f/f mesenteric lymph node DCs.Dots represent individual mice from one of three independent experiments.Data are represented as mean G SD.

Figure 2 .
Figure 2. Intracellular material acquisition from surrounding cells is specific to DCs and universal among all DC subsets tested (A) Splenic B cells (B), T cells (T), peritoneal macrophages (Mac), or dermal CD45 À cells (CD45 À ) were sorted from wild type C57BL/6 mice, labeled with SYTO62 and co-cultured with MutuDC1 cells for 45 min.Dots represent individual mice.Data combined from two experiments.(B) Sorted cells were labeled with CFSE and co-cultured for 45 min with SYTO62 labeled COCA keratinocytes (KC).Points represent individual mice.Data normalized and combined from two experiments.(C) Epidermal Langerhans cells (eLC), skin draining lymph node migratory Langerhans cells (sdLN LC), sdLN cDC1, sdLN cDC2, resident DCs (rDC), and mesenteric lymph node migratory DCs (mLN DC) were sorted from hLangCre-YFP f/f mice, labeled with CFSE, and co-cultured with RNA labeled COCA KCs for 45 min.Dots represent individual mice.Data pooled from three experiments.(D) moDCs were differentiated from human CD14 + monocytes and labeled with CFSE, then co-cultured with RNA labeled PBMCs on ice or 37 C for 45 min.Dots represent individual replicates.One representative experiment of two is shown.Data are represented as mean G SD.

Figure 3 .
Figure 3. Dendritic cells siphon RNA from neighboring cells through a contact dependent mechanism that does not resemble conventional means of antigen uptake (A) Outline of experiment to measure RNA and protein transfer to MutuDC1s with or without direct contact.(B) Flow cytometric analysis of SYTO62 RNA signal measured in MutuDC1s after 45 min incubation with RNA labeled COCA KCs.Results are from a single experiment.(C) Representative images and quantification of SYTO62 signal contained within MutuDC1s after keratinocyte:DC co-cultures.Mean pixel intensity of far-red channel (SYTO62) was calculated within the area occupied by GFP+ MutuDC1s on a per cell basis.Images acquired with a Nikon A1R confocal microscope using a Plan Fluor 403 Oil objective.Dots represent individual cells.Results are from a single experiment.(D) MutuDC1s (green) interacting with COCA KCs.Max projections of z stack images taken on a Nikon A1R confocal microscope using a Plan Fluor 403 Oil objective plus 10x scanner zoom.(E) Transfer of SYTO62 labeled RNA to MutuDC1s from keratinocytes relative to vehicle controls in the presence of an ATPsynthase inhibitor (1 mM Oligomycin A), an inhibitor of F-actin formation (8 mM Cytochalasin D), a macropinocytosis inhibitor (32 mM 5-(N-Ethyl-N-isopropyl) amiloride (EIPA), and a gap junction inhibitor (5 mM 1-Heptanol).Data normalized and pooled from three experiments.Data are represented as mean G SD.

Figure 4 .
Figure 4. RNA transfer is dependent on calcium and can be partially blocked with the scavenger receptor inhibitor Polyguanylic acid (A) Representative histograms of MutuDC1 cells after incubation in direct contact with RNA labeled COCA keratinocytes in the presence or absence of 5 mM EDTA.(B) RNA signal in MutuDC1 cells after incubation with RNA labeled B16 cells.MutuDC1s were treated with thapsigargin (2 mM), BAPTA-AM (50 mM), or both for 30 min on ice prior to co-culture in media containing Ca 2+ , Ca 2+ free media, or 5 mM EDTA.(C) RNA dye signal relative to control measured in MutuDC1 cells after incubation with RNA labeled B16 cells.MutuDC1s treated with 32 mg/mL Pronase or cocultured with B16 cells in the presence or absence of 500 mg/mL PolyG, or 5 mM EDTA as indicated.(D) RNA dye signal relative to control measured in MutuDC1 cells after incubation with RNA labeled B16 cells.MutuDC1s treated with 32 mg/mL Pronase or cocultured with B16 cells in the presence or absence of 500 mg/mL PolyG, or 5 mM EDTA as indicated.(E) RNA dye signal relative to control measured in MutuDC1 cells after incubation with RNA labeled B16 cells.MutuDC1s treated for 30 min with 50 mM LY294002, 5 mM EDTA, or 500 mg/mL PolyG as indicated.All experiments repeated at least three times.Representative results from a single experiment shown.Data are represented as mean G SD.

Figure 5 .
Figure 5. MutuDC1, but not MutuDC2 can cross-present the acquired ovalbumin (A) Control staining with SIINFEKL-MHC-I-specific antibody of B16, B16-OVA, MutuDC1, MutuDC1 pulsed with SIINFEKL, MutuDC2 and MutuDC2 pulsed with SIINFEKL.(B) MutuDC1 and MutuDC2 were co-cultured for the indicated time with B16 or B16-OVA and then the SIINFEKL-MHC-I levels determined by flow cytometry.Representative flow plots and summary graph (left lower corner) from one out of two experiments are shown with 2-3 technical replicates.(C) As in (B), but some of the MutuDC1 co-cultured with B16 or B16-OVA for 3 h were supplemented with PolyG/EDTA.Data from two independent experiments with 3 technical replicates were pooled.(D) MutuDC1 and MutuDC2 were co-cultured with B16 or B16-OVA labeled with SYTO62 for 45 min and then the transferred RNA signals (SYTO62) determined by flow cytometry.Data pooled from 3 independent experiments for B16, and one experiment for B16-OVA, with 2-3 technical replicates.Data are represented as mean G SD.

Figure 6 .
Figure 6.Materials acquired are presented on MHC-II (A) Top row, representative flow plots for YAe staining of MutuDC1 unmanipulated, pulsed with Ea peptide, or co-cultured for 3 h with BALB/c B cells either with or without PolyG/EDTA treatment.Second row, YAe staining of MutuDC2 unmanipulated, pulsed with Ea peptide, or co-cultured for 3 h with BALB/c B cells either with or without PolyG/EDTA treatment.Right: summary graph for MutuDC1 (top) and MutuDC2 (bottom).(B) As in (A), but BALB/c T cell were used.Data from three independent experiments with 2-3 technical replicates were pooled.Relative levels to DCs are shown.Data are represented as mean G SD.

Figure 7 .
Figure 7. Inflammation does not alter RNA acquisition by DCs (A) MutuDC1s and MutuDC2s were exposed to 1 mg/mL LPS, 10 mg/mL IFNa, 5 mg/mL PolyI:C, or 0.5 mM CpG for 12 h and co-cultured with SYTO62-labeled B16 cells.The acquired RNA signal in DCs was determined by flow cytometry.(B) Like (A), but the B16 cells were treated as indicated.Data were pooled from two independent experiments, with 2-3 technical replicates.Data are represented as mean G SD.