Rapid Lymphatic Dissemination of Encapsulated Group A Streptococci via Lymphatic Vessel Endothelial Receptor-1 Interaction

The host lymphatic network represents an important conduit for pathogen dissemination. Indeed, the lethal human pathogen group A streptococcus has a predilection to induce pathology in the lymphatic system and draining lymph nodes, however the underlying basis and subsequent consequences for disease outcome are currently unknown. Here we report that the hyaluronan capsule of group A streptococci is a crucial virulence determinant for lymphatic tropism in vivo, and further, we identify the lymphatic vessel endothelial receptor-1 as the critical host receptor for capsular hyaluronan in the lymphatic system. Interference with this interaction in vivo impeded bacterial dissemination to local draining lymph nodes and, in the case of a hyper-encapsulated M18 strain, redirected streptococcal entry into the blood circulation, suggesting a pivotal role in the manifestation of streptococcal infections. Our results reveal a novel function for bacterial capsular polysaccharide in directing lymphatic tropism, with potential implications for disease pathology.


Introduction
Lymphatic dissemination of intracellular bacteria and viruses is a well characterized mechanism of pathogenic invasion in the host, which occurs independently of transit through blood [1][2][3]. In contrast, exploitation of lymphatics by extracellular bacterial pathogens has received scant attention, despite clear clinical evidence that such pathogens can induce pathology within the lymphatic system [4,5].
Group A streptococcus (GAS) is one such important, exclusively human, extracellular pathogen. Pathology in the host is initiated by breach of mucosal surfaces and subsequent tissue destruction, resulting in a diverse disease spectrum spanning the superficial (pharyngitis, pyoderma) to the systemic (necrotizing fasciitis, toxic shock syndrome) as well as subsequent postinfection immune sequelae (rheumatic fever) [6]. Combined, these manifestations of GAS disease account for the 4 th highest mortality rate amongst bacterial pathogens [7].
The onset of invasive bacterial disease is dependent on the appropriately timed expression of an arsenal of virulence factors which facilitate pathogen dissemination within the host. The group A streptococci produce over 40 virulence factors, of which the hyaluronan (HA; (GlcNAc β1-4 -GlcUA) n ) capsule [8] expressed by almost all serotypes is critical in defense against neutrophil-mediated opsonophagocytosis [9]. Elaboration of this capsule further contributes to GAS pathogenesis by facilitating colonization of the host through adhesive interactions between HA and its well characterized receptor, CD44, in pharyngeal epithelium [10,11]. This interaction results in bacterial translocation and subsequent invasion of host tissue via the paracellular route [12]. We and others have previously shown that GAS have a particular ability to disseminate from a focus of infection to locally draining lymph nodes [13,14], a feature that, experimentally, has been associated with systemic dissemination. Clinically, GAS has been associated with lymphatic pathologies such as lymphadenitis and lymphangitis [4,5]. At present however, the host factors underlying this apparent lymph tropism and lymphoid tissue dissemination are completely unknown.
In addition to CD44, a number of other HA binding proteins are expressed in the human host [15,16]. Lymphatic vessel endothelial receptor (LYVE)-1 is a member of the HA binding Link protein superfamily that bears 41% amino acid sequence similarity to CD44 [16]. LYVE-1 is almost exclusively expressed in lymph vessels and lymph node sinuses, a system where CD44 is completely absent, and has been implicated in binding of HA from interstitial matrix and in HA-mediated leukocyte trafficking [16,17]. However, the propensity for HA-producing pathogens to hijack this pathway as a means of lymphatic tropism has not been explored.
In this work, we provide the first evidence to implicate the HA capsule as a critical bacterial determinant of GAS lymphoid tropism. We further demonstrate that a specific adhesive interaction between capsular HA and LYVE-1 is pivotal for GAS dissemination to draining lymph nodes in vivo, thus defining a novel conduit for GAS translocation, with potentially important implications for the establishment of systemic disease.

Capsular HA is a critical factor for lymphatic trafficking of GAS in vivo
The streptococcal virulence factors which govern bacterial transit to lymph nodes remain poorly understood. We hypothesized that the HA capsule could play an important role in this tropism, due to both its anti-phagocytic and adhesive properties. Since the degree of encapsulation varies greatly between GAS serotypes [8], and capsular HA is known to occlude other surface expressed adhesins [18,19], we generated isogenic hyaluronan synthase (hasA) mutant strains in two important serotypes; the hyper-encapsulated GAS serotype M18, and the lower density capsule serotype M89, to allow for unambiguous assessment of capsule involvement ( Fig 1A). Importantly, the total ablation of HA capsule expression had no discernible effect on in vitro bacterial growth rate in either GAS strain (Fig 1A).
Soft tissue infections represent an important GAS disease manifestation, which can readily be modelled in mice. Such models allow reproducible measurement of spread to locally draining lymph nodes at early time points following thigh muscle infection, prior to development of any severe signs of sepsis. Accordingly, we utilized a murine thigh muscle infection model to investigate the role of capsular HA in bacterial lymphatic tropism. Significantly, expression of capsular HA was clearly associated with enhanced bacterial transit to draining lymph nodes in vivo. Three hours after intra-muscular infection, the bacterial burden in the draining ipsilateral inguinal lymph node was markedly greater in mice infected with wildtype encapsulated GAS than that observed for mice infected with isogenic acapsular mutant GAS (Fig 1B). This was true for both the M18 and M89 isogenic strain pairs. Importantly, there was virtually no spread to the contralateral node, indicating that bacterial transit to draining lymph node was via afferent lymphatics and not blood circulation.
Loss of capsule expression is known to increase sensitivity of GAS to opsonophagocytic clearance by neutrophils in a serotype dependent manner, a phenomenon to which M18 strains are particularly sensitive [9]. Consistent with this, we detected a 1.5 log-fold reduction in bacterial burden at the site of infection in mice infected with isogenic acapsular M18 Δcapsule compared with wildtype M18 (Fig 1B), which corresponded with a significant increase in neutrophil-mediated uptake of bacterial cocci in an in vitro opsonophagocytosis assay (S1 Fig). Despite the observed differences in clearance from muscle between M18 and M18 Δcapsule , an almost three log-fold difference was observed between these isogenic strains upon dissemination to the draining ipsilateral lymph node (Fig 1B). In the case of serotype M89, loss of capsule had barely any effect on bacterial clearance from muscle, consistent with in vitro opsonophagocytosis data (S1 Fig). Similar to serotype M18 however, this loss of capsule led to a 3-log fold reduction in transit to the draining inguinal lymph node. These results provide direct evidence of a new and previously unanticipated role for the GAS HA capsule in directing lymph tropism and lymphatic spread.

Capsular HA interacts directly with human LYVE-1
Having established a link between HA capsule expression and the capacity of GAS to disseminate to draining lymph nodes via the lymphatics (Fig 1), we considered the possibility that the process was mediated by interaction with LYVE-1, the major HA receptor expressed in human and murine lymphatic vessels, which do not express CD44 [16,20]. We have previously reported that endogenous LYVE-1 displays only residual levels of constitutive HA binding, both in vitro and in vivo, due in part to heavy modification with auto-inhibitory sialylated Olinked glycans [21]. Hence, in preliminary experiments we performed adhesion assays with recombinant human (hu) and murine (m) LYVE-1 expressed by transfected COS 7 fibroblasts, a cell type with reduced sialyltransferase activity that permits efficient binding of soluble high molecular weight HA. Although GAS showed no ability to bind to untransfected COS 7 cells, encapsulated GAS bound both species of LYVE-1 transfectant avidly, and in a manner that was specifically inhibited by LYVE-1 blocking mAb (S2 Fig).
To determine whether native sialylated LYVE-1 can support binding of capsular HA we performed quantitative adhesion experiments with monolayers of primary lymphatic endothelial cells isolated from human dermis (HDLEC), which express abundant levels of the receptor on the cell surface [22]. Encapsulated GAS bound efficiently to HDLEC, and the adherent streptococci could be seen to outline the cell borders in a distribution that closely matched that of LYVE-1 (Fig 2A). Moreover, encapsulated GAS also showed increased binding to HDLEC that had been engineered to over-express human LYVE-1 by lentivirus-mediated gene transduction ( Fig 2B). Binding to native and lentivirus transfected HDLEC was decreased 3 and 8 fold respectively following incubation with LYVE-1 HA-blocking mAbs (Fig 2A and 2B), and up to 3-fold by competition with a soluble recombinant human LYVE-1 full ectodomain fragment ( Fig 2C). Further to this we demonstrated that free high molecular weight HA (>1 x 10 6 Da) could compete with encapsulated GAS for LYVE-1 binding in a dose-dependent manner, resulting in an almost 5-fold reduction in bacterial adhesion to lentivirus transfected HDLEC, whereas low molecular weight HA (7.5 x 10 3 Da) could not ( Fig 2D). Importantly, addition of a CD44 HA-blocking mAb had no effect on adhesion (S3 Fig). Together these data indicate that the interaction between GAS and the lymphatic endothelium was indeed specific and LYVE-1 mediated.
Intriguingly, higher magnification images of HA encapsulated GAS adhering to the surface of native untransfected HDLEC revealed the individual streptococcal chains to be anchored to variable sized patches of LYVE-1 tethered by single or multiple cocci ( Fig 2E). This suggests that the configuration of HA within the bacterial capsule, unlike free high molecular weight HA, can engage the natively sialylated normally quiescent LYVE-1 in lymphatic endothelium through receptor clustering.

Acapsular GAS do not interact with LYVE-1
The HA capsule constitutes the outermost surface of GAS, and has been demonstrated to mask both fibronectin binding proteins and M-protein [18,19,23]. Indeed, ablation of capsular HA can enhance GAS adhesion to a variety of cell types [19], likely due to the subsequent exposure of addtional bacterial surface proteins, able to interact with host receptors. Consistent with this, isogenic acapsular GAS displayed enhanced adhesion to both native and LYVE-1 supertranfected HDLEC when compared with parental encapsulated strains (S4 Fig), as has been described previously for other primary human cells [11]. In stark contrast to the preference shown by HA encapsulated GAS for adherence at endothelial junctions, acapsular cocci aggregated randomly over the endothelial cell bodies, and in a manner that was wholly independent of LYVE-1 adhesion (S4 Fig). The observed clumping may well result from increased GAS-GAS interaction upon loss of capsule and the interaction with LECs is most likely mediated by integrins as described previously for the analogous interaction of non-encapsulated GAS with epithelial cells [11].

HA encapsulated GAS bind murine lymphatic endothelium
The structures of human and murine LYVE-1 are very similar, and as such we hypothesized that capsular HA would be sufficient to mediate specific interactions between GAS and native murine LYVE-1. We assessed this by performing adhesion experiments with monolayers of primary lymphatic endothelial cells isolated from murine dermis (MDLEC), which express native LYVE-1 on the cell surface [22]. In common with HDLECs, MDLECs were able to support binding of encapsulated M18 GAS via the LYVE-1 receptor, as demonstrated by inhibition of this interaction by inclusion of a specific murine LYVE-1 HA blocking mAb (Fig 3). Similar to HDLECs, acapsular GAS exhibited clumping and HA-independent adhesion that did not involve LYVE-1 binding. These results, from both human and murine primary cells firmly establish LYVE-1 as a specific lymphatic endothelial adhesion receptor for HA encapsulated GAS in vitro.
In order to visualize lymph-borne dissemination of GAS in vivo, mice were infected intradermally in the ear pinna, and confocal microscopy of whole mount sections was utilized to image bacterial translocation through surrounding lymphatic vessels to the cervical lymph node draining the site of infection (Fig 4A-4C). Although few images captured bacteria within the lumen of LYVE-1 + dermal lymph vessels, we could nevertheless observe that GAS were concentrated in LYVE-1 rich regions, particularly the paracortical regions that receives afferent lymph from the ear and adjacent dermis (Fig 4D and 4E).

GAS translocation to draining lymph nodes is impaired in LYVE-1 deficient mice
Having shown that the HA capsule is critical for lymphatic dissemination of GAS in the mouse and identified LYVE-1 as the main adhesion receptor for capsular GAS in primary HDLEC and MDLEC, we next determined whether LYVE-1 plays a role in GAS dissemination in vivo.
Accordingly, to explore the consequences of total LYVE-1 deletion on the course of lymphborne GAS infection we measured streptococcal transit to draining lymph nodes in constitutive LYVE-1 -/mice. As previously reported, these mice have no obvious developmental defects, and have apparently normal lymphatic architecture, fluid clearance and immune function [24]. Following infection via an intra-muscular route, the dissemination of encapsulated M18 GAS was compared between LYVE-1 -/mice and age and sex matched wildtype controls by quantitative culture. Importantly no difference was observed in bacterial burden at the site of infection ( Fig 5A). However, considerably fewer viable streptococci were recovered from the draining ipsilateral inguinal lymph nodes of LYVE-1 -/mice compared with wildtype mice (Fig  5B), consistent with a functional role for LYVE-1 in streptococcal uptake and / or nodal transit through lymphatic vessels. Interestingly, there was a corresponding rise in the number of bacteria recovered from the blood of LYVE-1 -/animals compared with wildtype (Fig 5C and 5D). This suggests HA-encapsulated M18 GAS can switch to the systemic circulation when access to the lymphatics is impeded.
In order to visualize the location of M18 GAS in the lymph nodes draining the site of infection, the same intra-muscular thigh infection model was used, and inguinal lymph nodes from both wildtype C57Bl/6 and LYVE-1 -/were dissected and sectioned at the 3 hour time-point. Hyper-encapsulated M18 GAS were readily visible in 3/5 of the inguinal lymph nodes resected from infected wildtype mice (Fig 5E-5G), consistent with the results of quantitative culture experiments ( Fig 5B). Moreover, while discrete clumps of streptococci could be seen in contact with LYVE-1 + sinuses within these nodes, the bacteria were particularly prominent in the paracortical regions ( Fig 5G). Importantly, when lymph nodes from LYVE-1 -/mice were imaged for comparison, these in all cases (0/4) contained very few bacteria ( Fig 5H) and hardly any were detectable within the paracortical region ( Fig 5I). These findings further underline the functional importance of LYVE-1 for lymphatic dissemination of encapsulated M18 GAS and support the conclusion that LYVE-1 may regulate the initial entry of bacteria to the afferent lymphatic vessels.
Functional blockade of LYVE-1 in mice impairs GAS dissemination to draining lymph nodes Finally, we explored the consequences of LYVE-1 functional blockade in a mouse model of soft tissue infection, with pre-administration of either LYVE-1 HA blocking mAb or control immunoglobulin. Dissemination of M18 GAS to the draining ipsilateral inguinal lymph node and persistence of bacteria within the thigh muscle injection site were quantified three hours postinfection. There was no difference in bacterial burden at the site of infection at this time point (Fig 6A), however quantification of bacteria recovered from the ipsilateral lymph node revealed a ten-fold reduction in GAS dissemination following pre-treatment with LYVE-1 mAb ( Fig  6B). A similar result was also observed in mice pre-treated with LYVE-1 mAb following infection with M89 GAS (S5 Fig), confirming the same key role for LYVE-1 in the translocation of encapsulated GAS to draining lymph nodes indicated from the studies carried out with LYVE-1 -/mice. Furthermore, in common with LYVE-1 -/mice (Fig 5B), the reduction in bacterial burden in the draining lymph node of mice infected with M18 GAS was accompanied by a rise in bacterial burden in spleen (Fig 6C), and blood in LYVE-1 mAb pre-treated mice (Fig 6D). This effect was not seen in mice infected with M89 GAS and therefore may be specific to M18 GAS.
In order to better understand the influence of the M18 GAS LYVE-1 interaction on disease outcome and severity, we assessed the impact of LYVE-1 functional blockade on the induction of sepsis in mice following intra-muscular bacterial challenge over a 24 hour period (S6 Fig).  outcome, we measured the levels of serum IL-6, a sensitive marker of tissue inflammation. Although not significantly different, there was nevertheless a clear trend towards increased IL-6 levels in mice subjected to functional LYVE-1 blockade, (control IgG-treated mean: 1534, (SD: 372.9); LYVE-1 mAb-treated mean: 1848, (SD: 612.7)). Together these data provide evidence that LYVE-1 mediates interactions between the GAS capsule and lymphatic vessel endothelium that are critical for lymphatic tropism in vivo. Furthermore, in the case of hyperencapsulated M18 GAS, this tropism may also limit systemic spread via the blood.

Discussion
The host lymphatic network represents an ideal conduit for extracellular pathogens. Although recovery of GAS from draining lymph nodes in experimental disease and clinical syndromes are consistent with lymphatic spread [5], the underlying mechanisms for this potentially critical process and subsequent impact on disease outcome remain poorly characterized. In this investigation we demonstrate for the first time that the Gram positive extracellular pathogen, Group A streptococcus, can rapidly disseminate to draining lymph nodes via the lymphatic network, and that the process occurs through a specific interaction between the bacterial HA capsule and the lymphatic endothelial receptor, LYVE-1. We show by means of in vitro binding studies that expression of LYVE-1 is both necessary and sufficient to support adhesion of encapsulated GAS in vitro, and provide several lines of evidence indicating that the interaction between HA capsule and LYVE-1 is critical in vivo for GAS transit to draining lymph nodes. This was in stark contrast to isogenic acapsular GAS, which, although able to adhere to both human and murine LECs, did so in a non-specific and LYVE-1 independent manner and displayed a reduced ability to disseminate to draining lymph nodes in vivo.
Our studies have demonstrated a clear requirement for capsular HA in the interaction between GAS and LYVE-1. As documented for many pathogenic bacteria, GAS is known to vary capsule expression both in vitro and in vivo, a phenomenon known to be regulated by multiple factors including the key two-component regulatory system CovR/S [25,26]. In addition to this, capsule expression can be phenotypically fixed by genetic mutation, as exemplified by the hyper-encapsulated serotype M18, strains of which harbor a premature stop codon in the CovR/S regulator RocA [8], and acapsular serotypes M4 and M22, which completely lack the capsule operon [27]. The impact of serotype-specific variation in capsule expression on GAS adhesion to LYVE-1 and subsequent transit to draining lymph nodes is a subject that clearly warrants further study. While we report a similar reduction in transit of M18 and M89 GAS to draining lymph nodes following LYVE-1 receptor blockade, an accompanying increase in dissemination to the blood circulation and distant organs was only seen following M18 GAS infection, suggesting serotype specific variation in the role of the capsule LYVE-1 interaction during infection. We speculate that the interaction between M18 GAS and LYVE-1 augments association with lymphoid tissue and limits the spread to blood, thus providing a possible explanation for the infrequent association of M18 GAS with invasive disease and closer association with rheumatic fever [7].
It has previously been reported that mutations in the global GAS regulator CovR that normally represses capsule expression, can result in increased GAS dissemination to draining lymph nodes, a finding that can now be explained by the observations reported herein [28]. Previous data also demonstrate a role for the chemokine-cleaving protease SpyCEP in transit of GAS to draining lymph nodes, but the mechanism by which this occurs is still unknown [13]. As such the significance of lymphatic spread on the pathogenesis of individual GAS serotypes and other streptococcal species clearly warrants further investigation.
It is interesting to consider the extent to which lymphatic access via the LYVE-1:HA axis might contribute to the pathogenesis of other extracellular bacteria in addition to GAS. Several streptococcal species such as the zoonotic S. zooepidemicus and animal pathogens S. equi and S. uberis are known to express capsular HA [29]. Interestingly virulent S. equi displays a mucoid phenotype similar to that of hyper-encapsulated serotype M18 GAS, and is associated with the suppurative equine disease "strangles" that affects primarily the lymph nodes [30]. This further implicates the lymphatic system as a major conduit and focus for infection in domestic animals, possibly through an interaction with equine LYVE-1. Additional bacterial pathogens such as Bacillus cereus also express a capsule comprised of HA, while the closely related B. anthracis exhibits a single nucleotide polymorphism which results in a frame-shift that precludes the expression of HA [31]. One might speculate that such evolutionary loss of HA expression impairs access to the lymphatic network, resulting in increased transit through the blood circulation, as reported here, and may even be a critical determinant in the choice made by pathogens between these different vascular routes for establishing invasive disease. The observations made herein may well also be applicable to intracellular bacterial pathogens. It was recently reported that the Gram negative intracellular pathogen Yersinia pestis disseminates to draining lymph nodes, but that this process may not require an intracellular stage as previously thought [2]. These data suggest that the extracellular trafficking of pathogenic bacteria via the lymphatics may represent a universal mechanism for regulating systemic dissemination, in which LYVE-1 and other host receptors may play critical and previously unrecognized roles. In the case of GAS, the lymphatic tropism of heavily encapsulated bacteria, may act to restrict spread to the blood.
We have yet to determine exactly how LYVE-1:GAS interactions mediate lymphatic dissemination. The simplest explanation is that they promote streptococcal translocation into the lumen of lymphatic vessels and establishment of a reservoir in lymphoid tissue, much in the same way as GAS interactions with CD44 on the epithelial cell surface promote longevity of nasopharyngeal carriage and GAS persistence in the nasopharynx [10]. This hypothesis is supported by our observation that relatively few M18 GAS were recovered from lymph nodes draining the muscle infection site in LYVE-1 -/mice compared with their wildtype counterparts. It is also notable that LYVE-1 is most abundant within the button-like junctions of initial lymphatics where migrating leukocytes enter from the surrounding tissues [32], and where the receptor may well facilitate their transit through hyaluronan-mediated interactions. Hence, it is tempting to speculate that HA encapsulated GAS hijack this normal immune mechanism to mediate their transit into lymphatics and then to draining lymph nodes. Whether some bacteria exit or-re-enter lymphatic vessels before reaching the lymph node is unknown; it seems possible that the clinical syndrome of lymphangitis could reflect either the presence of GAS in, or egress from, lymphatic vessels. GAS interaction with LYVE-1 may influence this and any subsequent inflammation.
The observed interaction between GAS and LYVE-1, and subsequent transit to draining lymph nodes can be regarded as benefitting either the pathogen or the host. It is widely accepted that the expression of capsular HA protects GAS against opsonization and phagocytosis by host immune cells, providing a mechanism by which encapsulated GAS could subvert any specialized innate response elicited by infected lymph nodes [33]. Results herein suggest that interaction with LYVE-1 or transit to lymph nodes, at least in the case of serotype M18 GAS, may prevent spread to the blood, insofar as blockade or genetic deletion of LYVE-1 not only prevented lymph node infection but promoted emergence in peripheral blood and distant organs. However, the interaction of GAS HA capsule with LYVE-1 may also provide an adhesive niche for the establishment of a persistent streptococcal reservoir within draining lymph nodes.
In conclusion, we describe a novel interaction between GAS capsular HA and the lymphatic endothelial receptor LYVE-1 that provides an important conduit by which encapsulated streptococci can disseminate into the lymphatic vasculature. To our knowledge this represents the first identification of a molecular mechanism for entry into the lymphatic system by an extracellular pathogen. Moreover, our findings provide a molecular explanation for the clinical observation that local lymphatic drainage plays a major role in limiting immediate systemic spread by GAS. The dissemination of extracellular bacteria in lymphatics may have far reaching relevance in the pathogenesis of both acute bacterial infection and post-infection sequelae.

Ethics statement
Normal donor cells were acquired from an approved sub collection of the Imperial College Tissue Bank. Use of anonymized human tissue was approved through the Oxford Regional Ethics Committee. In vivo experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986, with appropriate UK Home Office licenses according to established institutional guidelines.

Cell culture and transfections
Primary HDLEC and MDLEC were derived from surgically resected human skin tissue and the skin of BALB/C pups respectively, using mAb MACS bead magnetic selection as described previously [22]. Purified cells were cultured in EGM-2 MV on gelatin-coated plastic (0.1%, Invitrogen). All cell lines were cultured at 37°C, 5% CO 2 under humidified conditions and all experiments performed using confluent cell monolayers. Transfection of COS 7 with fulllength human and mouse LYVE-1 cDNA and super-transfection of HDLEC with human LYVE-1 cDNA was achieved by lentiviral transduction.

Lentiviral transduction of COS 7 and HDLEC with human and murine LYVE-1
Coding sequences of lyve-1 were amplified from the appropriate cDNA using pfu Ultra AD (Agilent) with the following primers: The amplified product was cloned into the lentiviral vector pHR Sin, based on the HIV retrovirus. 293T cells cultured in DMEM supplemented with 10% FCS were transiently transfected with pHR Sin plasmids together with pMD.G and p8.91 [34] using Genejuice (Merck, Darmstadt, Germany) according to the manufacturer's instructions. Supernatant was harvested at 72hrs post-transfection and passed through a 0.45 micron filter to remove cell debris before transferring to COS 7 cells or HDLEC.

Preparation of soluble recombinant human LYVE-1 ectodomains
The extracellular domain of human LYVE-1 (terminating after residue 238) in a 10 x histidine tag followed by a stop codon was amplified from a plasmid carrying full-length human LYVE-1 cDNA using the high fidelity polymerase pfu Ultra AD (Agilent). The construct was transfected into CHO cells and a clone with high expression levels was selected (kindly carried out by Janet Fennelly and Simon Davis; HIU, WIMM, Oxford). His-tagged protein was purified from culture supernatants using a 5 ml His Trap column (GE Healthcare). Following elution, the protein was buffer exchanged into PBS.

Capsule quantification
GAS strains were cultured in THB and capsule quantified as described previously [8] using the hyaluronan DuoSet (R&D).

GAS adhesion assays
HDLECs, MDLECs and COS 7 cells were seeded at 5x10 5 cells/well and once confluent, pretreated with R10 media (RPMI, 10% FCS, L-glutamine) containing either control or blocking antibodies against LYVE-1 or CD44, high (>950 kDa) or ultralow (4-8 kDa) molecular weight HA (R&D Systems, UK), or recombinant soluble human LYVE-1 ectodomain for 20 minutes at room temperature. GAS were cultured on CBA and resuspended in PBS before being inoculated into each well at an MOI of 10. Adhesion assays were carried out following incubation for 30 minutes at 37°C, 5% CO 2 . Cells were washed 3 times with 1x PBS (Gibco) to remove nonadherent bacteria and cell-associated bacteria were quantified following lysis of the cell monolayer in sterile water. Replica wells of an 8-well chamber slide (Becton-Dickinson) inoculated with GAS pre-labelled with FITC-conjugated anti group A antibody (Abcam) were counterstained with polyclonal anti-human or anti-mouse LYVE-1 antibodies as appropriate and imaged using a Zeiss Axiovert inverted epi-fluorescence microscope or a Zeiss LSM780 laser scanning inverted confocal microscope.

Immunofluorescence antibody staining of mouse tissues
C57Bl/6 female mice (n = 3) were infected intra-dermally with 5×10 7 M18 GAS and whole mount tissue sections and draining lymph nodes imaged 6 hours post infection. Whole mount tissue staining was carried out as described previously [2]. Frozen sections of lymph node were prepared by cryostat and fixed for 5 minutes in ice-cold acetone. Samples were stained with polyclonal anti-mouse LYVE-1 Ab and FITC-conjugated anti-group A carbohydrate antibody (Abcam). Whole mount tissue samples were viewed by Zeiss LSM780 confocal microscope, using either a Plan-Apochromat 10x/0.3DIC M27 (total magnification: 100x) or Plan-Apochromat 63x/1.4 oil (total magnification: 630x, resolution: 0.24 μm). For frozen lymph node sections, nuclei were counterstained with DAPI and sections were viewed on a Zeiss Axiovert S100 microscope equipped with epi-fluorescence.
Neutrophil phagocytosis assays GAS pre-labelled with FITC-conjugated group A carbohydrate antibody (Abcam) were assayed for ability to resist uptake by purified human neutrophils as described previously [35].

Statistics
All statistical analyses were performed with GraphPad Prism 5.0. Comparison of two datasets was carried out using unpaired Mann-Whitney and three or more data sets were analyzed by Kruskal-Wallis followed by Dunn's multiple comparison test. A p-value of 0.05 was considered significant.