Interleukin 17 drives interstitial entrapment of tissue lipoproteins in experimental psoriasis

Impaired Passage of HDL through Skin in Experimental Psoriasis

Interstitial fluid and the lymph it subsequently forms is enriched in discoidal HDL (~5–10 nm Stokes radius). The skin is also one of the body’s largest reservoirs of extravascular interstitial fluid (Aukland and Reed, 1993) and contains a large pool of the total body HDL (Randolph and Miller, 2014). We thus wondered how HDL recirculation from the interstitium to lymph and then plasma was affected in skin and whether it was impacted in autoimmune skin disease, like psoriasis. We employed an imiquimod (IMQ)-induced experimental psoriasis model in WT C57BL/6 mice. In this model, daily application of the toll-like receptor 7 agonist imiquimod (IMQ) to ear skin for 14 days was associated with epidermal thickening (Fig. 1A) and local swelling (Fig. 1B). Distal skin sites like the back flank skin did not show histological changes like epidermal thickening or infiltrates (Fig. 1A). There was, nonetheless, altered HDL trafficking in distal back skin (Fig. 1C). Here, a mixture of human albumin and human interstitial HDL, the natural state of endogenous HDL particles, collected from human apoA-I transgenic (Tg-hApoA-I) mice was injected into back skin in a manner so as to minimize interstitial pressure from the injection (Swartz et al., 1999). Then its appearance in plasma, requiring that it first passes through skin tissue and then draining lymphatic vessels, was analyzed using ELISA specific for human albumin or human apoA-I, respectively. The experiment was designed in checkerboard-style, wherein interstitial HDL from IMQ-treated or control mice was injected, separately, into skin of either control or IMQ-treated mice. Albumin traveled with similar efficiency from skin to plasma in either control or IMQ-treated recipient mice (Fig. 1C). However, HDL was delayed in arrivingto plasma when injected into IMQ-treated mice, even though the skin site of injection was remote from the site of IMQ application on the ear where lesions were evident (Fig. 1C). By contrast, interstitial HDL from IMQ-treated mice mobilized normally in naïve recipients (Fig. 1C). These data suggested that systemic changes in the skin micro-environment, including distal nonlesional skin, induced by IMQ negatively affected HDL recirculation.

In contrast to HDL, dendritic cell (DC) migration from the back skin location was normal in IMQ-treated mice (Fig. 1D). Furthermore, intravital imaging of beads taken up into skin-draining lymphatic vessels (Fig. 1E, left; Movie 1), revealed that lymph flow overall was similar in the two conditions (Fig. 1E). Thus, experimental psoriasis impaired HDL recirculation from skin under conditions when DC migration to lymph nodes, fluid flow through lymphatics, and albumin transit from skin to plasma were normal. In keeping with the fact that extracellular matrix molecular arrangement and charge affect molecular transit, insoluble collagen / mg of skin was increased (Fig. 1F) in the absence of measurable changes in soluble collagen (Fig. S1A), hyaluronan (Fig. S1A), or sulfated proteoglycans (Fig. S1A), or loss of skin hydration (Fig. S1B). There was also a significant increase in lysyl pyridinoline residues, confirming an increase in crosslinked collagen in distal, nonlesional skin in IMQ-induced experimental psoriasis (Fig. 1F).

Generation of Photoactivatable ApoA-I Vectors and Knock-in Mice to Study HDL Trafficking

We next aimed to consider whether HDL trafficking in artery walls, a location highly relevant to vascular diseases, might be also impaired in experimental psoriasis. To address this technically challenging question, injection of HDL (such as done above) into thin-walled arteries would not be feasible. Moreover, we aimed to study HDL recirculation using techniques that did not perturb the microenvironment, including in skin, even as the most careful injection may. Thus, we constructed AAV vectors and knock-in mice that encoded the expression of human apoA-I bearing a photoactivatable GFP tag at the N-terminus (PGA1), as detailed in Fig. S2. Specifically, we constructed a technique wherein HDL particles, in whatever form they are found in tissues (Fig. S2A), could be tagged by exposure to light within the tissue microenvironment, and then followed from such interstitial spaces back to plasma (Fig. S2A). PGA1 was functionally active, both in its apoA1 domain and PA-GFP domain (Figs. S2–3) and it reconstituted HDL in WT and apoA1^−/− mice (Fig. S3).

With this new tool in hand, we quantified HDL transit out of skin and into plasma. WT mice were infected with PGA1 vectors. Then 3 weeks later, a region of the hind flank was shaved to expose approximately 4-cm² area of skin for subsequent photoactivation with 405 nm light generated from a handheld laser. PGA1 emerged in plasma by 1 h after photoactivation, reaching a plateau by 2 h (Fig. 2A). Simultaneous analysis of interstitial fluid revealed clearance of fluorescent PGA1 from the skin occurred steadily during the first 2 h (Fig. 2A). We also similarly light-activated WT mice that did not receive vector to ensure specificity (Figure 2B). We selected 2 h as the optimal time point at which to quantify transit of HDL from skin to plasma. Because reverse cholesterol transport depends on intact lymphatic vasculature in skin, as skin drains to plasma via intervening lymphatic vessels (Lim et al., 2013; Martel et al., 2013), we assessed PGA1 transport from skin of Chy mice compared with littermate controls (Figure 2C). The appearance of plasma fluorescence occurred by 2 h after skin photoactivation in PGA1-expressing WT mice, but not in PGA1-expressing Chy mice (Figure 2C), verifying that the increase in plasma fluorescence was dependent on an intact lymphatic passageway from skin.

We then generated a mouse line in which PGA1 followed by a stop codon was inserted in front of the first exon of mouse apoA-I. Direct sequencing of PCR-amplified 5’ and 3’ junction fragments verified the success of the PGA1 sequence insertion (Fig. S4, top left). PGA1 protein was robustly observed in PGA1^KI/+ mouse plasma, comparable to Tg-hApoA-I mice (Jax 001927) (Fig. S4A). Interstitial fluid PGA1 values were 20% of those observed in plasma, consistent with the established extent of reduction in apoA-I of human skin lymph compared with plasma (Miller et al., 2013b). N PGA1 expression supported generation of plasma HDL particles (Fig. S4B) bearing human apoA-I protein (inset, Fig. S4B). Photoactivation (Fig. S4C) led to elevated fluorescence in the same fractions, indicating that PGA1 was intact and functional in those fractions.

Experimental Psoriasis Traps Photoactivatable HDL in Skin

Having generated and validated use of the PGA1^KI/+ mice or PGA1 vectors, we returned to the condition of experimental psoriasis to determine if PGA1 might be useful to assess impaired HDL trafficking, as observed in HDL adoptive transfer experiments (Fig. 1). At 2 h after photoactivation of the nonlesional flank skin distal to the ear where IMQ was applied, fluorescent HDL failed to appear in plasma (Fig. 3A), with a much greater impairment than observed in the earlier injection approach. Remarkably, recovery of interstitial fluid from the sites of light activation revealed a strong retention of fluorescent PGA1 in the skin interstitial fluid (Fig. 3B), indicating that HDL was unable to efficiently pass through the skin microenvironment. An extended time course confirmed that appearance of fluorescent PGA1 in plasma after photoactivation in skin of IMQ-treated mice was greatly suppressed and delayed (Fig. 3C). Area under the curve corresponding to plasma-associated fluorescent HDL during 8 h following photoactivation reached only 40% of control values in IMQ-treated mice (Fig. 3C). Impaired skin to plasma trafficking of HDL took time to develop after IMQ treatment was initiated. Trafficking of HDL from nonlesional back skin to plasma was far more modestly impacted at one week after IMQ treatment on the ear compared with the profound inhibition observed at 2 weeks (Fig. 3D). Collagen accumulation in skin was not increased at week 1 after IMQ treatment, but was increased at week 2 (Fig. 3E).

To address the concept that HDL passage from skin to plasma might then allow passage from plasma to other target organs, we monitored the appearance of fluorescence in interstitial fluid prepared from the heart and skeletal muscle at the time of euthanasia. Fluorescence was diluted in these targets compared with plasma, as expected, but was nonetheless detected in the heart and skeletal muscle and was increased in these organs in mice not treated with IMQ as compared with those that were treated (Fig. 3F).

Experimental Psoriasis Negatively Affects Vascular Stiffness, Arterial Collagen Accumulation, HDL Transit through the Artery, and Atherosclerosis: Role of IL-17

Collagen increases earlier documented in skin were also observed in the heart tissue, with collagen levels in the heart returning to baseline after IMQ treatment ceased (Fig. 4A). Arterial collagen also increased remarkably, albeit with some delay, as arterial collagen was primarily elevated during a 2-week chase period following the initial 2-week IMQ treatment (Fig. 4A). This increase was primarily in the adventitia (Fig. 4A, micrograph panels). Consistent with these findings, IMQ treatment for two weeks on, then two weeks chase, promoted vascular stiffness, as assessed by a rise in the augmentation pressure after placement of a cardiac catheter (Fig. 4B).

Because the cytokine IL-17, produced by T cells, promotes vascular stiffness and accumulation of vascular collagen in angiotensin-induced hypertension (Guzik et al., 2007; Wu et al., 2014), we wondered if the arteries in IMQ-treated mice might, like skin, display impaired HDL trafficking and, if so, whether such impairment would be governed by IL-17. To that end, we treated mice with neutralizing anti-IL-17 mAb or isotype-matched control antibody during the IMQ treatment on skin to produce experimental psoriasis. Simultaneously, HDL trafficking from the carotid artery was assessed by shining light on surgically exposed carotid arteries in PGA1^KI/+ mice. When we photoactivated the carotid artery in PGA1^KI/+ mice to assess the influence of experimental psoriasis on HDL passage through the artery wall, we examined the plasma first within minutes after carotid photoactivation to ensure that we were not directly photoactivating HDL in the blood and then hourly thereafter (Fig. S5). We also examined the impact of IMQ treatment on this assessment of HDL trafficking from the artery wall. HDL trafficking was indeed suppressed (Fig. 4C), in the same time course that led to increased collagen deposition and augmentation pressure, the two-week on/two-week chase design (Fig. 4A; B). Neutralization of IL-17 during the experiment remarkably abrogated development of impaired HDL transit from the artery wall (Fig. 4C).

These findings in the artery wall prompted us to explore whether the conditions that promote sluggish HDL trafficking might be associated with increased cardiovascular disease. Thus, we studied atherosclerosis in apoE^−/− mice fed a cholesterol-enriched diet for 8 weeks, in which during the last 3 weeks of this treatment, some of the mice were treated on their ears daily with IMQ cream. As previously reported (Xie et al., 2017), this treatment also led to significantly reduced plasma cholesterol, lowering VLDL cholesterol in particular (Fig. S6). However, the effect was less dramatic than the effect we observed with the PCSK9 AAV vector. It is noteworthy that a recent study using IMQ treatment on mouse skin concluded that murine atherosclerosis was not enhanced by experimental psoriasis (Madsen et al., 2018). However, this study provided short (5 day), intermittent treatments with IMQ, which lead to dramatic skin inflammation, but may not allow for the increase in tissue collagen associated with HDL trapping, which only occurs over time (Fig. 4A). We would predict such collagen changes would be critical for affecting cardiovascular disease. Indeed, despite the reduction in VLDL observed in apoE^−/− mice (Fig. S6), which would favor disease protection, atherosclerosis was nonetheless enhanced in mice treated 3 weeks with IMQ on the ear (Fig. 4D).

Since HDL trafficking from the artery wall and from skin was restored to normal after loss of functional IL-17, we wondered if disease burden would likewise be reduced. Thus, we neutralized IL-17 during the 3 week period that we also administered IMQ to some of the mice. Such neutralization does not alter plasma cholesterol (Xie et al., 2017). Indeed, for this 3-week period of treatment, neutralizing IL-17 abrogated the extent of enhanced atherosclerosis increased by imiquimod (Fig. 4D), correlating with the impact of anti-IL-17 on preserving HDL trafficking through the artery wall (Fig. 4C).

Finally, considering that these changes in the artery wall correlated with collagen accumulation, we next set out to directly address whether inhibiting the crosslinking of any newly synthesized collagen by blocking lysyl oxidase activity would affect atherosclerosis in the context of experimental atherosclerosis. Indeed, the lysyl oxidase inhibitor and widely recognized anti-fibrotic agent beta-aminopropionitrile (BAPN) (Lampi and Reinhart-King, 2018) effectively blocked the enhanced disease associated with IMQ treatment, but did not impact baseline disease (Fig. 4E).

Experimental Psoriasis Impedes Recirculation of Multiple Larger Macromolecules in a Manner Dependent upon Increased Collagen

To clarify mechanisms associated with impaired HDL trafficking, we turned back to our studies in skin, where mechanistic work is more tractable. Elevations in collagen density were always observed in skin and arteries with impaired HDL recirculation. We wondered if a reduction in interstitial volume caused by increased collagen accounted for the failure of HDL to recirculate efficiently, in which case we would expect to observe (a) loss of interstitial space not occupied by collagen in dermis affected by impaired HDL trafficking, (b) more broadly impaired trafficking of interstitial molecules based at least roughly on Stokes radius and not specific to HDL, (c) and reversal of impaired trafficking under conditions when increased collagen was prevented.

First, we determined if HDL lipid loading (which affects its overall size) differed enough between control and IMQ-treated groups to affect molecular size, we analyzed HDL particles on nondenaturing gels after interstitial HDL was collected from back skin of PGA1^KI/+ mice, treated or not with IMQ for 2 weeks, compared with interstitial fluid from Tg-hApoA-I mice. The latter was observed in a range of sizes similar to that of human plasma HDL (Fig. 5A), all greater than 7 nm Stokes radius indicative of lipid loading. PGA1^KI/+ HDL was also observed at a size indicative of lipid loading in vivo (Fig. 5A), consistent with in vitro results (Fig. S2H), and this fusion protein ran as a larger mass overall. However, there was no size shift in HDL particle size associated with IMQ treatment (Fig. 5A).

To examine possible loss of interstitial space between collagen fibrils in skin from IMQ-treated mice that retained HDL, we carried out focused ion beam scanning electron microscopy (FIB-SEM), examining the back skin dermis just beneath the papillary ridges in mice treated or not with IMQ on both ears daily for 2 weeks (Figs. 5B, C). There was a striking change in the arrangement and packing density of collagen fibrils in the nonlesional back skin of IMQ-treated mice (Figs. 5B, C; Supplemental Movies 2), with a clear reduction in the space between collagen fibrils (Fig. 5D). There were also dark rims of contrast around collagen fibrils in replicates of the IMQ-treated skin (Fig. 5C) but not in replicates of control skin (Fig. 5B). This contrast difference may relate to collagen fibril association with other as of yet undefined molecules in the interstitium.

The increased collagen density in nonlesional skin raised the general possibility that typically larger macromolecules (> albumin) might generally be less capable of moving through a tightly woven interstitium with a higher density of collagen. Thus, using the injection approach of different molecules in skin (from Fig. 1), we compared the ability of additional molecules to exit the skin and mobilize to plasma in control and IMQ-treated mice. IgA, with a Stokes radius of ~6.5 nm, below that of even the smallest interstitial HDL observed herein, mobilized from skin similarly in the control and IMQ-treated mice (Fig. 5E), reminiscent of albumin. However, pentameric IgM, at ~12.65 Stokes radius, was reduced in recirculation to plasma if injected into the back skin of IMQ-treated mice (Fig. 5E), consistent with the concept that size is a key determinant in affecting interstitial trafficking in experimental psoriasis. Beyond IgM, few naturally occurring extracellular macromolecules are larger in size than lipoproteins, suggesting that the mechanism we have uncovered might particularly impact trafficking and trapping of lipoproteins. Indeed, LDL mobilization to plasma was also markedly suppressed from the back skin of mice with experimental psoriasis (Fig. 5E).

Finally, we prevented the accumulation of additional dermal collagen in experimental psoriasis by treating with BAPN (Lampi and Reinhart-King, 2018) during the period of IMQ application. The use of this inhibitor did not affect baseline collagen levels but indeed prevented the increase in dermal collagen, quantified by hydroxyproline assessments (Fig. 5F). In PGA1^KI/+ mice, photoactivation of the back skin showed that the IMQ-induced impairment in HDL mobilization to plasma was completely reversed by BAPN treatment (Fig. 5G), again with no effect of BAPN on baseline transit of HDL (Fig. 5G). Taken together, these data support the concept that experimental psoriasis alters tissue microenvironment distally, such that due to a loss in interstitial space between matrix fibrils, particularly collagen fibrils, molecules in the size range of lipoproteins are unable to mobilize effectively.

HDL Trafficking and Induction of Procollagen in Experimental Psoriasis is Regulated by CD4⁺ T Cells and IL-17

Current clinical treatments to combat psoriasis include neutralizing the cytokines IL-23 and IL-17. Depletion of CD4⁺ T cells or neutralization of IL-17 during the course of IMQ treatment prevented impairment of HDL trafficking from the back skin (Fig. 6A). However, neutralization of the p40 subunit shared by IL-23 and IL-12 was unable to restore HDL trafficking during experimental psoriasis (Fig. 6A). Collagen accumulation in the skin was prevented in the treatments that restored HDL transit (Fig. 6B). Thus, here again, collagen accumulation in the skin correlated with impaired HDL mobilization. Since inhibitors of collagen crosslinking, BAPN or IL-17, abrogated the enhancing effect of IMQ on atherosclerosis (Figs. 4E, 4F), we also tested whether depletion of CD4⁺ T cells affected atherosclerosis. Indeed, the increase in atherosclerotic plaque observedin response to IMQ was reversed by anti-CD4 T cell depletion, although in the short time frame of depletion (3 weeks), baseline disease was not affected by loss of CD4⁺ T cells (Fig. S6C-D).

In skin, by day 9 following IMQ treatment, CD4⁺ RORγτ⁺ IL-17-producing T cells were evident at low levels in single cell suspensions of back skin (Fig. 6C). Analysis of skin tissue by quantitative PCR for cytokine mRNA revealed the expected strong induction at the site IMQ application of IL-17A, IL-17F, IL-22, TGFβ, IL-1β, and IFNγ (Fig. 6D). By contrast, in the distal back skin where HDL trafficking was impaired at day 14, IL-17A mRNA was the only cytokine mRNA elevated at day 14, while other cytokine mRNAs were not or were scarcely induced at either day 7 or day 14 (Fig. 6D). Neutralization of IL-12/IL-23 p40 led to a marked reduction in several cytokine mRNAs at the lesional site of topical IMQ application on the ear (Fig. 6D), suggesting that the neutralizing antibody was at least partially efficacious to quell local lesion immunity. However, anti-p40 mAb was unable to abrogate appearance of IL-17 mRNA in the nonlesional distal back skin observed at day 14, in contrast to the loss of CD4⁺ T cells or IL-17 blockade (Fig. 6D). We then wondered if the role of CD4⁺ T cells and the cytokine IL-17 in regulating the accumulation of insoluble collagen was mechanistically linked to changes in the fragmentation, or turnover, of collagen or new collagen production in the form of procollagen. While insoluble collagen cannot be reliably extracted in acid, collagen fragments and procollagens are readily extracted in acidic conditions and can be analyzed by immunoblot (Voorhees et al., 2015). This approach revealed little to no evidence of collagen fragments, since no bands smaller than soluble processed collagen I (150 kD) were detected. However, bands of higher molecular weight corresponding to procollagen I were increased in the back skin when IMQ was applied to the ear for 14 days (Fig. 6E), indicating an induction of new collagen synthesis, that upon crosslinking by lysyl oxidase likely accounts for the rise in stable, insoluble collagen. This rise in production of procollagen was prevented by loss of CD4⁺ T cells or neutralization of IL-17, but was not affected by neutralizing IL-12/23 p40 (Fig. 6E). These data suggest that Th cells producing IL-17 regulate collagen synthesis in distal back skin.

These findings, indicating that CD4⁺ T cells regulate collagen synthesis and HDL trafficking in back skin remote from the site of IMQ application, raised the possibility that T cells programmed in the lymph nodes draining the site of IMQ application might home to distal skin, as previously demonstrated in IMQ-induced experimental psoriasis for γδ T cells (Ramirez-Valle et al., 2015). Thus, we treated some mice receiving IMQ with the S1P1 agonist FTY720 to impede the ability of lymph node T cells to enter the blood and home to distal sites (Cyster and Schwab, 2012). This treatment blocked the collagen increase in distal skin (Fig. 6F, left graph; Fig. S7) and partially restored HDL trafficking from skin to plasma (Fig. 6F, right graph). These data are consistent with a model in which T cells are not needed for local inflammation at the site of the IMQ-induced psoriasis-like lesion, as previously indicated (Naik et al., 2017), but that nonetheless an induction of a T cell response in lymph nodes allows for the programming of widespread T cell homing back to distal skin, where subtle changes are induced that increase collagen levels without evidence of overt inflammation.

Genetic Absence of T Cells Differentially Affects Local and Distal skin Sequelae in IMQ-induced Psoriasis

Next, we studied RAG1^−/− mice and TCRβ^−/− × TCRδ^−/− mice compared with WT control mice with respect to the impact on lesional ear skin swelling (ear thickness) or HDL trafficking from distal skin. Consistent with other literature finding no necessity for conventional T cells in local lesions of experimental psoriasis (Pantelyushin et al., 2012), histological features of psoriasis were augmented in mice lacking T cells for the entire 14 days of IMQ treatment (Fig. 7A, compare to Fig. 1B), and this was borne out by increases in overall ear skin thickness after 14 days of daily IMQ application (Fig. 7B). Depletion of CD4⁺ cells at day 7 ameliorated disease but this effect was lost by day 14 (Fig. 7B), when the genetic absence of T cells enhances disease. Neutralizing IL-17, by contrast to loss of T cells, was able to suppress lesional changes like ear thickening even in the 14-day IMQ regimen. These findings are consistent with the presence of local alternative sources of IL-17, such as innate lymphocytes (Fig. 7B).

Despite the lack of T cell dependence on the inflammatory features of the lesion site itself, HDL trafficking from the nonlesional back skin did not succumb to IMQ-induced blockade in the genetic absence of T cells (Fig. 7C), fitting with the data obtained using anti-CD4 depleting mAb that CD4⁺ T cells drive impaired HDL trafficking from nonlesional skin in experimental psoriasis (from Fig. 6). Thus, the role of T cells in this model of experimental psoriasis is more prominent in nonlesional than in lesional skin.

Limitations of study

Limitations of the study include the possibility that BAPN had off target effects. It is also important to keep in mind that PGA1 may not fully recapitulate endogenous HDL. Furthermore, while we suggest that the heighted atherosclerosis observed in concert with experimental psoriasis is due to changes in lipoprotein trafficking, we cannot eliminate other mechanisms at play in the atherosclerosis experiment. While anti-CD4 T cell depletion actually heightened inflammation in the skin following prolonged experimental psoriasis, and thus might be expected to enhance atherosclerosis rather than inhibit it in response to IMQ, we cannot rule out that a major effect of anti-CD4 mAb in the atherosclerosis model, along with the effects of anti-IL-17 and BAPN, were to block inflammation in the artery wall, rather than affect lipoprotein entrapment.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to, and will be fulfilled by, the corresponding author Gwendalyn J. Randolph (gjrandolph@wustl.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

GraphPad Prism version 7 was used for statistical analysis. Statistical analyses utilized two-tailed t test or one-way ANOVA followed by analysis of variance with Tukey correction when comparing with multiple groups.

DATA AND SOFTWARE AVAILABILITY

https://data.mendeley.com/datasets/jp56vt7dcc/1

ADDITIONAL RESOURCES