Differential requirement for ROCK in dendritic cell migration within lymphatic capillaries in steady-state and inflammation

Dendritic cells (DCs) are important in the initiation of adaptive immune responses. In peripheral tissues, such as the skin, DCs take up antigen, mature, and migrate via lymphatic vessels (LVs) to draining lymph nodes (dLNs), where they present antigen to resting T cells. Although this migration pattern has been known for more than 20 years,¹  DC migration into LVs is only now starting to be unraveled at the cellular level using live imaging technologies.^2-4  Before transmigrating across the lymphatic endothelium, DCs first need to squeeze through preformed, narrow pores present in the lymphatic basement membrane (BM).³  DC entry into LVs is thought to occur at the level of primary lymphatic capillaries, which feature discontinuous cell junctions with “button”-like accumulations of cell adhesion molecules.⁵  At the sites of such buttons, lymphatic endothelial cells (LECs) partially overlap and generate open flaps, through which leukocytes enter into lymphatics.^3,5  The most prominent mediator of DC migration into afferent lymphatics is CCL21, a chemokine constitutively expressed in LVs.^6,7  More recently, also other LEC-expressed molecules that mediate DC migration via LVs to dLNs have been identified.^8-11  Notably, ICAM-1 and VCAM-1, which are up-regulated in LVs during inflammation,^8,12  have been implicated in this process.⁸  Intriguingly, experiments performed with pan-integrin knockout DCs revealed that DC migration to dLNs in steady-state was integrin-independent.²  This apparent discrepancy could suggest that inflammation modulates the requirement for integrins and their ligands in DC migration, but this has not been conclusively addressed to date.

Until recently, it was commonly assumed that DCs that have transmigrated into the LV lumen are passively transported by the lymph flow.¹³  This assumption was supported by intravital microscopy (IVM) performed in rat mesentery lymphatics, where freely flowing lymphocytes are rapidly propagated in a pulsatile fashion.¹⁴  Lymph flow in lymphatic capillaries of peripheral tissues is difficult to measure and values ranging from 1 to 30 μm/s have been reported.^15,16  Notably, these rates are considerably lower than the average lymph flow in large collecting LVs,¹⁴  making it questionable whether hydrodynamic forces present in the small lymphatic capillaries would be sufficiently strong to transport cells by flow. In fact, a recent IVM study revealed that DCs actively migrated within small lymphatic capillaries and were only propagated by lymph after they had reached larger collecting vessels.⁴  This suggests that leukocyte migration into and within LVs might involve similar adhesive interactions as leukocyte migration within and out of blood vessels (BVs). However, this possibility has been poorly explored and no molecules involved in active intralymphatic leukocyte migration have been identified so far.

Amoeboid cell movement is driven by 2 principal forces, namely actin polymerization and cellular extension occurring at the leading edge, and actomyosin-mediated cellular contraction and de-adhesion, which take place at the cell's rear.^17,18  An important mediator of the latter 2 processes is the Rho-associated protein kinase (ROCK). ROCK controls the activity of nonmuscle myosin II, thereby promoting the reorganization and contraction of cellular actomyosin filaments.¹⁹  ROCK was shown to support cell movement by mediating nuclear contraction, which is required for migration through dense 3-D meshworks,²  and by mediating cellular de-adhesion from integrin ligands during leukocyte crawling on 2-D surfaces or on ICAM-1 expressed by blood vascular endothelium.^20-22  However, the involvement of ROCK in DC migration in vivo has not been investigated to date.

In this study, we made use of a new transgenic mouse model with red-fluorescent vasculature to further investigate the role of ROCK in interstitial DC migration and migration into and within lymphatic capillaries. Performing IVM in the mouse ear skin we found that DC migration in the interstitium was ROCK-dependent. Interestingly, ROCK only played a minor role in intralymphatic DC migration at steady-state, whereas during tissue inflammation this process was highly ROCK-dependent. Further mechanistic analysis suggested that ROCK supports intralymphatic crawling during inflammation by mediating de-adhesion from inflammation-induced ICAM-1 on lymphatic endothelium. Overall, our findings indicate that inflammation alters the requirement for adhesion molecules in DC migration on lymphatic endothelium and represent the first description of molecules, which are involved in this process.

Wild-type (WT) C57BL/6 mice were purchased from Charles River laboratories or from Janvier. CCR7^−/−23  and vascular endothelial (VE)–cadherin-Cre mice²⁴  were from Jackson Laboratories. CD11c-YFP mice²⁵  were kindly provided by Michel Nussenzweig (Rockefeller Institute, NY). VE-cadherin-Cre × RFP were generated by crossing Rosa26-loxP-stop-loxP-tdRFP mice²⁶  with VE-cadherin-Cre mice. All experiments were approved by the Cantonal Veterinary Office Zurich.

Four- to 6-week-old VE-cadherin-Cre × RFP mice were lethally irradiated (950 rad) and reconstituted with 5 × 10⁶ BM cells isolated from CD11c-YFP mice. Chimeric mice were used at least 6 weeks after irradiation.

A contact hypersensitivity (CHS) response toward oxazolone was induced as described.¹²  Briefly, mice were anesthetized by intraperitoneal administration of medetomidine (1 mg/kg) and ketamine (75 mg/kg) and sensitized by topical application of 2% oxazolone (4-ethoxymethylene-2-phenyl-2-oxazoline-5-1; Sigma-Aldrich) in acetone/olive oil (4:1 vol/vol) on the shaved abdomen (50 μL) and on each paw (5 μL). Five days later, 10 μL of a 1% oxazolone solution was applied topically to each side of the ears (challenge phase). Experiments were performed 24 to 48 hours after CHS challenge.

See supplemental Methods (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Mice were killed and ears harvested after hair removal with Veet depilation cream. Ears were split into 2 halves along the cartilage and blocked at room temperature for 30 minutes in 12% bovine serum albumin (BSA)/phosphate-buffered saline (PBS). Subsequently, ears were incubated for 1 hour at room temperature with either rat anti–mouse LYVE 1 (clone ALY7²⁷ ), rabbit anti–mouse LYVE-1 (AngioBio), hamster anti–mouse podoplanin (clone 8.1.1¹² ), rat anti–mouse MECA-32 (BD Bioscience), rabbit anti–mouse laminin (Sigma-Aldrich), or with corresponding isotype controls. Ears were washed with PBS/0.1% Tween-20 and incubated for at least 30 minutes with Alexa-conjugated secondary antibodies (Invitrogen) in presence of Hoechst 33342 (Invitrogen). Ears were washed with PBS and mounted using Vectashield (Vector Laboratories).

Ear whole mounts were analyzed on a Zeiss LSM 710-FCS confocal microscope (Carl Zeiss AG). Images were acquired using Zeiss ZEN 2010 software (Carl Zeiss AG). Imaris Version 7.2.3 software (Bitplane AG) was used for offline image processing. In some cases γ-corrections were applied to better visualize LV morphology (stated in the legends for Figures 1,Figure 2–3).

DCs were generated as previously described.²⁸  A detailed description is provided in supplemental Methods.

See supplemental Methods.

CD11c-YFP DCs were adjusted to 50 × 10⁶/mL and transferred into an insulin syringe. For DC injection, mice were briefly anesthetized with isoflurane (Attane; Minrad), and cells in a total of 10 μL were injected into 2 to 3 regions in the ventral side of the mouse ear pinna. Four to 6 hours later, mice were anesthetized with medetomidine (1 mg/kg) and ketamine (75 mg/kg) and the ears depilated with Veet cream. Mice were transferred to a custom-made microscopy stage, lying on their back with the dorsal side of the ear facing the coverslip of the stage. Ears were moisturized with PBS and immobilized by carefully placing another coverslip on top. For imaging, mice were placed into a 37°C incubator chamber installed on the microscope platform. Imaging was performed on a Zeiss LSM 710 inverted confocal microscope (Carl Zeiss AG). All time-lapse videos were acquired using a 20 × 0.8 NA Plan Apochromat objective. An Argon laser (488 nm, for yellow fluorescent protein [YFP] excitation) and a solid-state laser (561 nm, for tandem dimer red fluorescent protein [tdRFP] excitation) were used for image acquisition. When tracking intralymphatic DC migration, a high-resolution image (1024 × 1024 pixel, signal averaging) was first acquired at the beginning of the experiment to prove intralymphatic localization of DCs. Subsequent image stacks were acquired with lower resolution in the red channel (512 × 512 pixel; with higher z-distance of images, no signal averaging, less laser power) to avoid photo damage and to allow for shorter stack intervals. z-dimensions of acquired image stacks typically ranged between 30 to 50 μm, and featured 20 to 30 slices with an overlap of ideally 50% (Nyquist Sampling). The individual slice depth typically was 3 to 4 μm, the acquired x/y-dimensions 707 × 707 μm.

To study the effect of Y27632 on DC motility, 2 consecutive, 60-minute time-lapse videos were acquired (30-second stack intervals) in 2 regions of interest. Subsequently, mice were treated by intraperitoneal injection of 10 mg/kg Y27632 (Sigma-Aldrich) or of PBS vehicle control. Mice were kept under anesthesia at 37°C for 3 hours. Subsequently, 2 consecutive 60-minute time-lapse videos were recorded in the 2 previously imaged regions.

Imaris Version 7.2.3 software (Bitplane) was used to analyze cell velocity, directionality and axis dimensions. Cell tracking was performed using automatic surface detection followed by manual cell tracking. Cells were included into statistical analysis if they displayed a motility > 1 μm/min and were trackable over at least 5 time points. In some experiments, the maximal cell length and width was measured using IMARIS.

See supplemental Methods.

See supplemental Methods.

Fluorescein isothiocyanate (FITC) painting data were analyzed using the Student t test (unpaired, 2-tailed) and are presented as mean ± SEM. Data from IVM or flow chamber experiments (velocities, axis-ratios) were analyzed using the Mann-Whitney test. Significant outliers, based on Grubbs test, were excluded. All experiments were repeated at least 2 times. Statistical analysis was performed with Prism 5 (GraphPad).

When studying the vasculature of the mouse ear skin by immunofluorescent analysis of whole mounts we noticed that BVs and LVs could easily be distinguished based on morphologic differences (Figure 1A). Most LVs, in particular lymphatic capillaries, were much wider than BVs. Furthermore, lymphatic capillaries could often be recognized by their blind beginnings, whereas lymphatic precollectors and the deeper lying collectors featured characteristic valve structures. We therefore reasoned that it should be possible to distinguish both vessel types in transgenic mice expressing 1 single fluorophore under the control of a pan-endothelial gene promoter, such as the VE-cadherin promoter.²⁴  Therefore, we bred VE-cadherin-Cre mice²⁴  with Rosa26-loxP-stop-loxP-tdRFP mice²⁶  (Figure 1B). On Cre-activation, the latter mice express a genetically modified version of red fluorescent protein (RFP) under the ubiquitously active Rosa26 promoter.^26,29,30  Analysis of the offspring, so-called VE-cadherin-Cre × RFP mice, revealed that the fluorescent signal in the vasculature was readily detectable in ear whole-mounts (Figure 1C). Costaining with LYVE-1 confirmed that BVs and LVs could be distinguished based on morphologic differences. Notably, virtually all LYVE-1–positive lymphatics also expressed RFP (Figure 1C). Blind beginnings of lymphatic capillaries and valve structures in precollecting and collecting LVs were readily detectable (Figure 1D). Interestingly, besides cytosolic expression, bright RFP expression was detected in cell nuclei of both blood vascular endothelial cells (BECs) and LECs (Figure 1E). Furthermore, some ear-resident leukocytes were RFP-positive (5%-12% of cells from different subsets, data not shown). Interestingly, further analysis revealed, that lymphatic collectors present in the deep dermis were RFP negative. By contrast, lymphatic capillaries present in the superficial dermis (ie, the presumed sites of leukocyte entry into lymphatics) were consistently RFP positive (supplemental Figure 1).

Confocal IVM performed in the ear skin of anesthetized mice confirmed that it was possible to detect fluorescent BVs and LVs in the dermis of VE-cadherin-Cre × RFP mice in vivo (Figure 2A-B). Imaging was performed at a depth of up to 70 μm, corresponding to the epidermis and most of the dermis in an uninflamed mouse ear. As expected from fluorescence-activated cell sorter (FACS) analysis, also RFP-expressing leukocytes were readily detectable, particularly in the epidermis (Figure 2C). At a depth of 15 to 25 μm, mainly superficial BVs were visible (Figure 2D), whereas LVs were predominately found in deeper regions of the dermis (Figure 2E). On intravenous injection of FITC-dextran (2000 kDa) only presumed RFP⁺ BVs, but no presumed RFP⁺ LVs filled with FITC-dextran (Figure 2F). Taken together, these findings confirmed the suitability of VE-cadherin-Cre × RFP mice for IVM studies.

To image DC migration into and within LVs, lipopolysaccharide (LPS)-matured DCs, generated from the bone marrow (BM) of CD11c-YFP mice²⁵  (Figure 3A-C), were adoptively transferred into the ear skin of VE-cadherin-Cre × RFP mice. Imaging was initiated 4 to 6 hours after transfer and performed in intact skin regions distant from the injection sites. At this time point, many YFP⁺ DCs were present in the interstitium or within LVs (Figure 3D-E). Notably, entry into LVs was CCR7-dependent, because adoptively transferred CCR7^−/− DCs were much less frequently observed within LVs than WT DCs (supplemental Figure 2). Occasionally, also DCs that were in the process of migrating into LVs were imaged (supplemental Video 1). Analysis of videos revealed that most adoptively transferred DCs actively migrated within lymphatic capillaries (Figure 3F, supplemental Videos 2-3). DCs crawling within LVs were significantly faster than DCs migrating in the interstitial tissue (velocity of 5.7 μm/min vs 4.4 μm/min, respectively, P < .0001; Figure 3G). Although DC migration displayed an overall directionality within the vessel (supplemental Figure 3), we frequently observed intralymphatic DCs that changed their migratory path by 180° to migrate for some time in reverse direction of the expected lymph drainage (supplemental Video 4). Interestingly, experimentally elevating lymph flow by injection of PBS into the ear pinna did not seem to enhance the frequency of DC migration in the direction of presumed lymph flow (supplemental Figure 3D). A similar pattern of active intralymphatic migration was also observed when imaging adoptively transferred splenic DCs (supplemental Figure 4, supplemental Video 5). A second mode of motion, which was only rarely observed in the superficial lymphatic capillaries imaged in our setup, was rapid cell movement (≫ 100 μm/min). Such DCs had a round shape, indicating no interaction/crawling along the endothelium (supplemental Video 6, supplemental Figure 5).

To investigate whether this mode of active intralymphatic migration could also be observed in endogenous, skin resident DCs, BM chimeras were generated by reconstituting irradiated VE-cadherin-Cre × RFP mice with BM from CD11c-YFP mice. When imaging in uninflamed ear skin of BM chimeras, virtually no DCs were found to colocalize or migrate within LVs. By contrast, when imaging in LPS-inflamed ears of BM chimeras, many DCs were observed to migrate within LVs, displaying a similar pattern of active migration as observed in the adoptive transfer settings (supplemental Video 7).

In agreement with a recent study,⁴  our data suggested that in addition to interstitial DC migration and transmigration into LVs, also DC movement within dermal lymphatic capillaries was dependent on active, amoeboid cell migration. The latter process generally requires Rac1/Rac2-mediated actin polymerization in the cellular front and concomitant actomyosin-mediated cellular contraction and de-adhesion occurring at the cellular rear.^17,18  Indeed, simultaneous deletion of the small GTPases Rac1/Rac2 was shown to effectively impair DC migration to dLNs.³¹  By contrast, nothing is known about the importance of cellular contraction and de-adhesion in the migration of DCs from skin to dLNs. Because the latter 2 processes reportedly are mediated by ROCK,²  we decided to investigate the role of this kinase by blocking its activity with the well-described ROCK inhibitor Y27632.^2,22  In a first analysis, FITC painting experiments were performed in the ears of Y27632-treated or PBS control-treated mice. In these experiments, significantly fewer FITC⁺ DCs were recovered from the ear-draining auricular LNs of Y27632-treated mice compared with the dLNs of control-treated mice (−41%, P = .028; Figure 4A-B).

To further explore the role of ROCK in interstitial and intralymphatic DC migration we studied these processes by IVM. For this, YFP⁺ DCs were adoptively transferred into the ears of VE-cadherin-Cre × RFP mice, and baseline motility of DCs in the interstitium and within lymphatics was recorded after 4 hours. Subsequently, mice were either treated with Y27632 or with PBS control and subjected to another round of imaging 3 hours later (Figure 4C). Intriguingly, treatment with Y27632 induced a striking shape change in interstitially migrating DCs, leading to significant cellular elongation and increased axis ratios (Figure 4D-F). This was probably because of the impairment of ROCK-mediated nuclear contraction, which caused the cells to get stuck with their nucleus in the dense network of the extracellular matrix (ECM).²  Accompanying these phenotypic changes, the average migratory speed of DCs within the interstitium was profoundly reduced by Y27632 treatment (−46% compared with PBS treatment; Figure 4G). Notably, the velocity of DC migration on PBS treatment was slightly increased compared with the velocity at baseline (Figure 4G). This could be because of heat or other manifestations of tissue damage, such as mild tissue inflammation induced by depilation of the ear, the cell injection process or prolonged imaging. Interestingly, Y27632 treatment also induced a significant, albeit very subtle reduction in the velocity of DC migration within LVs (−15% compared with PBS, P = .006; Figure 4H).

Several recent studies have shown that ROCK and nonmuscle myosin II are required for lymphocyte crawling on blood vascular endothelium, supposedly by mediating the de-adhesion of LFA-1 at the cellular uropod from endothelial cell–expressed ICAM-1,^20-22  but no reports on a similar role for ROCK in DCs exist. ICAM-1 reportedly is expressed at very low levels in uninflamed LVs but becomes up-regulated in response to inflammation.^8,12  Performing FACS analysis of ear single-cell suspensions (Figure 5A) we found that inflammation, induced by a CHS response to oxazolone, lead to an up-regulation of ICAM-1 in LECs, resulting in ICAM-1 expression levels similar to the ones found in uninflamed BECs (Figure 5B). To investigate whether inflammation-induced up-regulation of ICAM-1 in LECs would enhance the ROCK-dependence of intralymphatic DC migration, IVM was performed in CHS-inflamed ear skin. Similar to what we had observed in uninflamed conditions, ROCK blockade significantly reduced the speed of interstitial DC migration during inflammation (Figure 5C). Interestingly, interstitial as well as intralymphatic DC migration was slightly faster in inflamed compared with resting tissue (Figure 5C-D). In contrast to the only slight reduction observed under uninflamed conditions (−15%), Y27632 treatment profoundly reduced the speed of intralymphatic DC migration (−52%) in inflamed LVs (Figure 5D). Moreover, in presence of tissue inflammation, the migration velocity of Y27632-treated DCs was significantly reduced compared with the velocity of Y27632-treated DCs migrating in uninflamed LVs (−37%, P < .0001; Figure 5D). Thus, the contribution of ROCK to intralymphatic DC migration was more pronounced in inflamed compared with uninflamed LVs, supporting our hypothesis that ROCK mediates DC de-adhesion from LEC-expressed ICAM-1 during intralymphatic migration. Notably, ROCK blockade did not alter the expression levels of LFA-1 and Mac-1 in DCs (data not shown).

IVM performed in LPS-inflamed ears of BM chimeras revealed that interstitial and intralymphatic DC migration was similarly reduced on treatment with Y27632, indicating a similar requirement for ROCK in endogenous DCs (supplemental Figure 6). To investigate how ROCK blockade affected the overall process of DC migration from the skin to dLNs in the context of inflammation, we again performed FITC painting experiments. DC migration from CHS-inflamed ears to dLNs was significantly enhanced by inflammation, as previously reported.¹²  Also in the context of inflammation, ROCK blockade lead to a decrease in DC migration, but this effect did not reach statistical significance (Figure 5E).

Our findings described in the previous paragraph suggested that up-regulation of inflammation-induced ICAM-1 in LECs could be functionally linked to the increased ROCK dependence observed for intralymphatic DC migration in the context of inflammation. FACS analysis confirmed that BM-DCs expressed high levels of CD11a and CD11b, which form part of the ICAM-1 ligands LFA-1 and Mac1 (Figure 6A). To further study the role of ROCK in DC crawling on lymphatic endothelium, in vitro flow chamber experiments were performed with BM-DCs and conditionally immortalized LECs (imLECs).¹²  Confluent imLEC monolayers were cultured overnight in presence of TNFα and IFNγ to mimic tissue inflammation and to up-regulate ICAM-1 expression (Figure 6B).¹²  To simulate physiologic conditions in primary lymphatic capillaries in inflamed tissues^15,16  experiments were performed in presence of low flow. Under these conditions, DCs avidly crawled on imLECs (supplemental Video 8). Pre-treatment of LECs with 2 different ICAM-1 blocking antibodies significantly reduced the speed of DC migration, compared with pre-treatment with an unspecific control antibody (Figure 6C-D), thus confirming a role for ICAM-1 in DC crawling on lymphatic endothelium. Further experiments with blocking antibodies revealed that both LFA-1 and Mac-1 were involved in DC crawling on lymphatic endothelium (Figure 6E-F). Similarly to our IVM experiments, treatment of DCs with Y27632 resulted in a significantly slower speed of DC crawling on LECs (Figure 6G-H). Because DCs were pre-incubated with Y27633 and washed before the experiment, one can conclude from these findings that blockade of DC-expressed ROCK was sufficient to reduce DCs crawling on lymphatic endothelium.

In this study, we made use of IVM and of transgenic mice with red-fluorescent vasculature to analyze the role of ROCK in intralymphatic DC migration and in the overall process of DC migration from skin to dLNs.

Gene-targeted mice expressing a Cre-recombinase or a fluorophore under an LEC-specific promoter have recently been generated,^32-36  but only 2 IVM studies performed in such mice have been reported so far.^35,36  Therefore, to date, LVs have mainly been visualized in IVM by in vivo labeling with intracutaneously injected fluorescent antibodies.^4,37,38  However, antibody labeling is considered less optimal for IVM than endogenous expression of fluorophores; it is typically restricted to the injection site and downstream regions and could interfere with leukocyte-endothelium interactions or lead to unwanted immune activation. To bypass these potential problems, we bred mice expressing Cre-recombinase under the pan-endothelial VE-Cadherin promoter²⁴  with RFP reporter mice.²⁶  IVM experiments and confocal whole-mount analyses confirmed that it was possible to distinguish BVs and LVs in the ear skin of VE-cadherin-Cre × RFP mice based on morphologic differences. Furthermore, initial lymphatic capillaries, which are the main sites of leukocyte entry into LVs,^3,5  uniformly were RFP⁺ in VE-cadherin-cre × RFP mice, confirming the suitability of this transgenic model for imaging DC migration into and within lymphatic capillaries.

In agreement with a recent study by Tal and colleagues, our IVM experiments revealed that adoptively transferred as well as endogenous, skin resident DCs actively migrated within superficial lymphatic capillaries.⁴  Although the imaging of YFP⁺ skin-resident DCs in BM chimeras represented a more physiologic approach, it proved difficult to use this setup for an in-depth mechanistic analysis of intralymphatic DC migration. Even on injection of LPS into the ear pinna, to induce DC maturation and migration, the number of DCs found within a LV segment typically was low, making it difficult to quantitatively investigate intralymphatic DC migration. This is probably because of the much lower number and density of endogenous DCs in the ear skin, compared with high numbers of BM-DCs deposited at the injection site. Consequently, in the latter approach, the probability of finding many DCs in the same LV are much higher. As mentioned in “DCs actively migrate within LVs,” we only rarely observed intralymphatic DCs, which were round-shaped and were flushed through LVs by flow. This is probably explained by our confocal imaging setup, which did not allow us to image in deeper regions of the dermis, where collecting vessels are most abundant, in which DCs are mainly transported by flow.⁴ 

Our IVM experiments revealed that interstitial DC migration was significantly reduced on ROCK blockade. These in vivo findings for the first time confirm previous observations made in vitro with DCs migrating in dense 3-D collagen matrices.²  The fact that DCs in the ears of Y27632-treated mice adopted a more elongated shape compared with DCs in control treated mice suggests that interstitial movement requires actomyosin-mediated contraction of the cell nucleus, to navigate through the dense network of the dermal ECM. Given that the composition and density of the ECM is tissue-specific,³⁹  it is probable that also the requirement for ROCK-mediated nuclear contraction during interstitial DC migration varies in different tissues. Interestingly, in steady-state, inhibition of ROCK only marginally reduced the speed of DCs crawling within LVs, indicating that LVs provide a largely unconfined, 2-D–like environment for DC migration. However, initial capillaries often exhibit a rather flattened and collapsed phenotype with a compressed lumen in steady-state. Therefore, it is possible that some nuclear contraction and squeezing is still required for DC migration through collapsed lymphatic capillaries. In the context of tissue inflammation, which is accompanied by lymphatic vessel dilation and up-regulation of integrin ligands, such as ICAM-1, it is more probable that the role of ROCK is to mediate de-adhesion of DC-expressed integrins from their LEC-expressed ligands. Notably, ROCK inhibition did not lead to a complete block of interstitial or intralymphatic DC migration, indicating that besides ROCK, also other kinases might mediate nuclear contraction and integrin de-adhesion in DCs. However, although several kinases reportedly regulate nonmuscle myosin II activity, ROCK presently is the only one with a documented role in leukocyte de-adhesion.¹⁹ 

In vitro and in vivo studies have shown that ROCK is critical for leukocyte transmigration through blood vascular endothelium.^22,40  Furthermore, it was recently reported that actomyosin-mediated nuclear contraction, induced by LEC-expressed semaphorin 3A, was important for DC entry into LVs.¹⁰  Given these findings and the fact that ROCK is important for DC migration through dense collagen meshworks in vitro,²  it is rather probable that ROCK is required for DC transmigration through the perilymphatic BM and subsequent entry into lymphatic capillaries. Interestingly, when imaging DCs in Y27632-treated mice we observed one DC that seemed to be arrested in the transmigration process (supplemental Video 9). However, not sufficient transmigration events (supplemental Videos 6 and 9) could be recorded to allow for a statistical analysis of this phenomenon.

Although our IVM studies showed that in inflammation both interstitial and intralymphatic DC migration were strongly ROCK-dependent, the inhibitory effect of Y27632 on DC migration from skin to dLNs was less strong in inflamed compared with steady-state conditions. It is possible that inflammation-induced changes in the lymphatic vessel architecture or the surrounding BM might render the transmigration process less dependent on actomyosin-mediated nuclear contraction. However, when analyzing the expression of laminin in the perilymphatic BM we could not detect any clear differences in the pre-existing pores, through which DCs reportedly enter into lymphatics.³  By contrast, we observed that the overall staining intensity for laminin around lymphatic initials was significantly reduced in inflamed compared with steady-state ear skin (supplemental Figure 7). This could indicate that the BM becomes thinner and less rigid in inflammation, what could render the transmigration process less ROCK-dependent. Interestingly, a similar inflammation-induced thinning of the BM was already reported for BVs.⁴¹ 

It was recently shown that under steady-state conditions, integrin-deficient DCs homed as efficiently to dLNs as WT DCs, suggesting that interactions between DC-expressed integrins and LEC-expressed integrin ligands were not important in the process of DC migration in the interstitium and into or within LVs.²  By contrast, an earlier study identified a role for inflammation-induced ICAM-1 or VCAM-1 in mediating DC migration from inflamed skin to dLNs.⁸  The new data from our work helps to solve this discrepancy, as they provide evidence that the requirement for adhesion molecules in DC migration varies between steady-state and inflammation. A probable candidate for such inflammation-dependent effects is ICAM-1, which becomes highly up-regulated in inflamed LECs but is only expressed at low levels in steady-state. Several studies have documented the role of ICAM-1 in leukocyte crawling on blood vascular endothelium in vivo^42,43  and in vitro.⁴⁴  In our flow-chamber assays, blockade of ICAM-1 or its DC-expressed integrin ligands significantly reduced the speed of DC crawling on activated lymphatic endothelium, suggesting a similar role for ICAM-1 in intralymphatic DC migration. Inflammation-induced ICAM-1 on LECs therefore may support the migration of DCs, by providing anchoring residues at the leading edge. At the same time, DCs require ROCK to mediate de-adhesion from LEC-expressed ICAM-1 at the uropod. It is possible that in addition to ICAM-1, other LEC-expressed integrin ligands also participate in this process. Notably, VCAM-1 is also induced in inflamed LECs in vivo,^8,12  but its expression is lower compared with the one of ICAM-1 (data not shown).

What could be the biologic meaning of intralymphatic DC crawling? Measurements of lymphatic flow in lymphatic capillaries have indicated values in the order of a few μm/s.^15,16  Notably, these velocities are much higher than the speed of active intralymphatic DC crawling measured by us and by others.⁴  If, on entry into lymphatic capillaries, DCs were exclusively transported by flow, their movement from the ear skin or footpad to the respective dLN (distance of < 2 cm) should, in theory, take a bit more than 1 hour (assuming that cells were traveling with a median lymph flow velocity of 4.7 μm/s¹⁵ ). However, studies performed in mice to evaluate the kinetics of DC migration in response to contact sensitization have revealed that it takes much longer before the first DCs are detected in skin dLNs, and maximal numbers of migrated DCs are only reached after 24 hours.^45,46  Mobilization of dermal DCs from the interstitial space to the lymphatic vessel in response to contact sensitization was recently shown to occur within 1 hour of DC stimulation,³⁸  and DC transmigration into lymphatic capillaries reportedly occurs within 2 hours (Tal et al,⁴ Sen et al,³⁸ and supplemental Video 1). Thus, it is questionable whether the overall time needed for DC mobilization, transmigration, and passive intralymphatic transport by flow could explain the long time required for DC migration from skin to dLNs. Therefore, DC crawling within lymphatic capillaries might be a rate-limiting and essential step for mobilizing DCs from lymphatic capillaries to the downstream collectors, in which flow conditions are sufficiently strong to support passive transport.

Overall, our results provide first mechanistic insights into the molecules mediating DC crawling within lymphatic vessels. Our findings demonstrate that DC transport within lymphatic vessels is a more complex process than originally anticipated and probably regulated by similar adhesive mechanisms as the well-explored migration and extravasation of leukocytes out of blood vessels.

The authors thank Carlos Ochoa (ETH Zurich) and Mark Liebi (Theodor Kocher Institute, Bern) for excellent technical assistance and Peter Horvath (Light Microscopy Center of ETH Zurich) for precious help with image analysis. C.H. gratefully acknowledges financial support from the Professor Dr Max Cloëtta Foundation, ETH Zurich (TH-grant no. 05 07-3) and the Novartis Foundation of Medical and Biologic Research (Switzerland).

Contribution: M.N. and C.H. planned the experiments and wrote the paper; M.N., D.A., and B.V. performed experiments and analyzed data; M.A. and R.L. designed and performed flow chamber experiments; M.L. and S.H. analyzed data; O.B. gave technical advice for the microscopy experiments; and H.L. and H.J.F. provided mice.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Cornelia Halin, Institute of Pharmaceutical Sciences, ETH Zurich, Wolfgang-Pauli St 10, HCI H413, CH-8093 Zurich, Switzerland; e-mail: cornelia.halin@pharma.ethz.ch.