Research ArticleVascular Biology

Methicillin-resistant Staphylococcus aureus causes sustained collecting lymphatic vessel dysfunction

See allHide authors and affiliations

Science Translational Medicine  17 Jan 2018:
Vol. 10, Issue 424, eaam7964
DOI: 10.1126/scitranslmed.aam7964

Lymphatics limp along after MRSA

Lymphedema is associated with skin and soft tissue infections, and both can be recurring, causing continual suffering in affected patients. To better understand the relationship between bacterial infections and lymphedema, Jones et al. used intravital imaging to examine the lymphatics of mice infected with MRSA. They observed lymphatic muscle cell death, which led to prolonged dysfunction months after the bacteria had been cleared. In vitro experiments with human cells indicated that bacterial toxins were responsible for damaging the lymphatic muscle cells, suggesting that the toxins could be targeted in patients to interrupt this brutal cycle.


Methicillin-resistant Staphylococcus aureus (MRSA) is a major cause of morbidity and mortality worldwide and is a frequent cause of skin and soft tissue infections (SSTIs). Lymphedema—fluid accumulation in tissue caused by impaired lymphatic vessel function—is a strong risk factor for SSTIs. SSTIs also frequently recur in patients and sometimes lead to acquired lymphedema. However, the mechanism of how SSTIs can be both the consequence and the cause of lymphatic vessel dysfunction is not known. Intravital imaging in mice revealed an acute reduction in both lymphatic vessel contractility and lymph flow after localized MRSA infection. Moreover, chronic lymphatic impairment is observed long after MRSA is cleared and inflammation is resolved. Associated with decreased collecting lymphatic vessel function was the loss and disorganization of lymphatic muscle cells (LMCs), which are critical for lymphatic contraction. In vitro, incubation with MRSA-conditioned supernatant led to LMC death. Proteomic analysis identified several accessory gene regulator (agr)–controlled MRSA exotoxins that contribute to LMC death. Infection with agr mutant MRSA resulted in sustained lymphatic function compared to animals infected with wild-type MRSA. Our findings suggest that agr is a promising target to preserve lymphatic vessel function and promote immunity during SSTIs.


Skin and soft tissue infections (SSTIs) are caused by microbial invasion of the epidermis, for which there are 14 million outpatient visits annually in the United States (1). Patients are generally treated with oral and topical antibiotics. However, complicated infections may additionally require intravenous antibiotics, leading to nearly 500,000 hospital admissions each year in the United States (2, 3). The most common pathogens associated with bacterial infections of the skin and underlying tissues are β-hemolytic Streptococci and Staphylococcus aureus, including strains of methicillin-resistant S. aureus (MRSA) (4). The resistance of MRSA to antibiotics, along with an increasing incidence of infections, has become a growing public health concern for individuals in the community and hospital settings (5), and underscores the need for treatments that target both MRSA colonization and its pathogenesis (6). These infections can often recur and are frequent in patients with impaired lymphatic function, including those with lymphedema (7, 8). Further, SSTIs can also lead to secondary lymphedema (7), suggesting that SSTIs can be both the consequence and the cause of impaired lymphatic function. However, little is known about how bacteria interact with lymphatic vessels. Understanding this relationship will be critical to developing future therapies and vaccination strategies that maintain lymph flow to improve clearance of MRSA and toxins as well as promote antigen delivery to the lymph node for robust immune responses.

The lymphatic vasculature consists of a network of vessels critical for maintenance of tissue homeostasis, fluid balance, immune function, absorption of dietary fat, and lipid transport (9). Disruption or malfunction of lymphatic vessels can result in lymphedema, presented as local fluid retention and tissue swelling (10). Primary lymphedema has a genetic etiology that leads to lymphatic insufficiency (11). In contrast, secondary lymphedema is more common and is triggered by obstruction or damage to lymphatic vessels caused by iatrogenic damage, inflammation, or infection—either bacterial or parasitic (12). Individuals with lymphedema are prone to develop SSTIs, both erysipelas and underlying tissue cellulitis (13).

Lymph flow is controlled by a combination of factors that propel lymph to the lymph nodes for immune surveillance and filtration (14). Primarily, the organized lymphatic muscle cells (LMCs) that cover collecting lymphatic vessels (CLVs) periodically contract to drive lymph forward (15). Calcium (Ca2+) signaling in LMCs drives contraction in response to chemical and mechanical stimuli such as endothelium-derived factors and lymph flow, respectively (16, 17). Nitric oxide, a vasodilator produced by the endothelium in normal and inflammatory conditions, has been shown to have a context-dependent effect on lymphatic contraction. In normal physiology, loss of nitric oxide has been shown to strengthen, blunt, or have no effect on lymphatic contraction (1820). Perturbations in either Ca2+ or nitric oxide signaling, such as blocking Ca2+ channels, increasing nitric oxide production, or causing sterile inflammation, have been shown to significantly compromise lymphatic contraction (18, 19, 21, 22). Here, we present a murine infection model to determine whether MRSA infections inhibit lymphatic vessel function and to define the molecular mechanisms related to this response.


Acute inhibition of lymphatic function in a localized MRSA infection model

To examine a potential relationship between MRSA infection and lymphatic function, we established an MRSA infection model of the mouse hindlimb, a location analogous to the commonly affected lower extremities in human SSTIs (23). Neutrophils infiltrated the hindlimb and draining lymph node (fig. S1, A to D) within 24 hours after infection. Lymphangiography revealed dilated afferent lymphatic vessels to the popliteal lymph node (PLN), which we designate here as PLVs, compared to PLVs of uninfected mice (Fig. 1, A and B, and fig. S1, A and E).

Fig. 1 Collecting popliteal lymphatic vessels exhibit diminished contraction and lymph velocity after MRSA infection.

C57BL/6 mice were infected subcutaneously in the hindlimb with 2 × 106 to 4 × 106 colony-forming units (CFU) of MRSA for 4 days or left uninfected. (A) Representative sequential images from intravital microscopy of PLVs perfused with FITC-dextran (2 million molecular weight). Scale bar, 50 μm (see movies S1 and S2). (B) Representative traces from PLV wall measurements show lymphatic diameter and contraction over time (n = 8 to 11, each group). (C) Ejection fraction, indicating strength of lymphatic vessel contraction, in uninfected mice and mice infected with WT MRSA after 4 days (n = 8 to 11, each group). (D) Frequency indicates the counts per minute (cpm) of PLV contractions in uninfected mice and mice infected with WT MRSA after 4 days (n = 7 to 10, each group). (E) Top: Representative depth versus time image of the optical coherence tomography intensity signal (grayscale) from a fixed transverse location that is used to identify the upper and lower boundaries of PLVs in uninfected animals and during active infection (day 4). LV, lymphatic vessel. Instantaneous lymph velocity (color) in the PLV in uninfected animals and during active infection (day 4) is overlaid on structural OCT depth scan. Bottom: Representative trace of instantaneous velocity of lymph flow averaged over the PLV cross-sectional area in uninfected animals and during active infection (day 4). (F) Time-averaged mean velocity of lymph flow in uninfected animals and during active infection (n = 3 to 11 measurements in each group). For (C), (D), and (F), statistical analysis was performed using Student’s unpaired two-sided t test. Error bars show SEM. *P < 0.05.

Because dilated lymphatic vessels are associated with lymphatic dysfunction (24), we measured the ability of PLVs to contract using intravital fluorescence microscopy (Fig. 1A) (25). MRSA infection impaired lymphatic contraction compared to mock-infected animals (Fig. 1, A and B, and movies S1 and S2). The strength of PLV contraction, as measured by ejection fraction—the theoretical fractional volume of lymph expelled with each vessel contraction—declined significantly after infection (Fig. 1C) (P < 0.05). PLVs from infected animals also showed a significant decrease in phasic contraction frequency compared to uninfected controls (Fig. 1D) (P < 0.05).

To determine whether lymph flow inside PLVs was also decreased, we used a recently developed label-free method to identify PLVs (Fig. 1E) and directly measure instantaneous total lymph velocity (basal and pulsatile flow) in vivo by Doppler optical coherence tomography (DOCT) (26). In uninfected mice, 90% of the lymph velocity measurements showed distinct peaks in flow in the PLV (Fig. 1E). Four days after infection, only 33% of the lymph velocity measurements were pulsatile, with others showing constant yet reduced lymph velocity compared to uninfected animals (Fig. 1, E and F). Together, mice infected with MRSA demonstrated significant inhibition of both lymphatic vessel contraction and lymph velocity during active infection (P < 0.05).

Chronic inhibition of lymphatic function after clearance of MRSA and resolution of inflammation

Subcutaneous infection of mice with MRSA led to a transient infection, with the peak of bacterial burden in the infected skin and underlying muscle, as well as in the lymph node, at 4 days after infection (Fig. 2A). Consistent with previous results for subcutaneous MRSA infections in mice, MRSA in skin (27) and ipsilateral PLN was not detected by colony formation assays 30 days after infection (Fig. 2A). Similar results were obtained using quantitative polymerase chain reaction (qPCR) to detect mecA and thermonuclease—both Staphylococcal-specific genes (fig. S2, A and B). Concomitant with the presence of inflammatory cells during early infection (fig. S2, A to D), a gene array measured an increase in several key genes that mediate the host inflammatory response (Fig. 2B). Notably, the transcript for interleukin-1β (IL-1β)—a protein essential for host defense against S. aureus (28)—was the most highly up-regulated gene 4 days after infection. Il-1β up-regulation was separately confirmed by qPCR (fig. S2C). In contrast, 60 days after infection, the inflammatory gene profile (Fig. 2B), including il-1β (fig. S2C), was comparable to the expression profile of uninfected mice. These data show MRSA clearance and resolution of inflammation by 30 and 60 days after infection, respectively.

Fig. 2 CLV dysfunction persists after clearance of MRSA.

(A) Infectious burden measured by colony formation assay in the draining PLN (left) at the indicated time points (n = 6 to 14, all groups). Infectious burden in the skin and underlying muscle tissue (right) from the site of infection at the indicated time points (n = 4 to 5, all groups). (B) Gene expression array of mouse inflammatory cytokines and receptors comparing tissue from the site of infection collected at day 4 (left) and day 60 (right) after MRSA infection. ΔΔCT was calculated with normalization of raw data to housekeeping genes. The graph plots normalized gene expression levels (log10) from control (uninfected) skin and muscle tissue (n = 2) on the x axis versus normalized gene expression levels (log10) of day 4 (n = 2) or day 60 (n = 2) post-infection skin and muscle tissue on the y axis. Circles above the upper gray line identify significantly up-regulated genes (P < 0.05); the blue line and lower gray line represent the boundary for genes unchanged and significantly down-regulated, respectively. Individual samples were loaded into each qPCR array plate and normalized to skin from uninfected mice. (C) Ejection fraction (n = 3 to 8) and (D) frequency (n = 3 to 10) show the strength and number, respectively, of PLV contractions over time. Each time point represents a different cohort of mice; days 1, 2, 4, 30, 60, and 120 were measured. (E) Representative depth versus time image of the depth-resolved optical coherence tomography intensity from a fixed transverse location that is used to identify the upper and lower boundaries of PLVs in uninfected animals and after clearance of infection (day 35). Instantaneous lymph velocity (color) in the lymphatic vessel in uninfected animals and after clearance of infection (day 35) is overlaid on structural OCT depth scan. (F) Time-averaged mean velocity of lymph in uninfected animals and after clearance of infection (day 35; n = 11 to 17 measurements in each group). Statistical analysis was performed using Student’s unpaired two-sided t test. Error bars show SEM. *P < 0.05.

By 120 days after infection, lymphatic vessel diameter had returned to the normal range found in most uninfected animals (fig. S2D). Surprisingly, the strength (Fig. 2C) and frequency (Fig. 2D) of PLV contraction remained impaired after clearance of infection at day 30, and this impairment was sustained 120 days after infection (P < 0.05). Using DOCT to measure lymph flow (Fig. 2E), we found pulsatile lymph flow in only 50% of the mice measured after MRSA clearance (35 days after MRSA infection) compared to 90% of mice in the control group (Fig. 2E). In addition, measurements made 35 days after MRSA infection showed limited net flow velocity (Fig. 2F), comparable to flow velocity measurements 4 days after infection. Moreover, we found that only two of seven mice at 260 days after MRSA infection showed pulsatile lymph flow. Consistent with this finding, the net lymph velocity remained decreased in MRSA-infected mice relative to age-matched controls 260 days after infection (P < 0.05) (fig. S2E).

Effect of nitric oxide inhibition on lymphatic function during infection

We hypothesized that nitric oxide produced by infiltrating cells during MRSA infection could disrupt the temporal and spatial nitric oxide gradients established by lymphatic endothelial cells (LECs) (19, 29), similar to sterile inflammation (19). Using immunofluorescence, inducible nitric oxide synthase (iNOS) was identified in inflammatory cells in the skin and within the abscess (Fig. 3A). iNOS-positive cells were often found in proximity to lymphatic vessels (Fig. 3A), suggesting a potential role for iNOS in inhibiting lymphatic contraction. Unexpectedly, most of the iNOS-positive cells did not stain positively for Gr-1, a marker of neutrophils and myeloid-derived suppressor cells (Fig. 3B). Depletion of Gr-1+ cells led to greater reduction of lymphatic contraction (fig. S3A), likely due to the critical role of Gr-1+ cells in host defense during acute infection (30). These findings are in contrast to the role of Gr-1+ cells in a model of sterile inflammation (19). Next, we tested whether MRSA-derived lipoteichoic acid (LTA), through production of iNOS (31), inhibited lymphatic contraction during MRSA infection. Purified LTA injected into the mouse hindlimb decreased lymphatic contraction in a dose-dependent manner (Fig. 3C) but did not impair lymphatic contraction in iNOS-deficient mice [iNOS knockout (KO)] (Fig. 3C). This suggests that LTA acts through iNOS to reduce the strength of lymphatic function, similar to sterile inflammation (19). Although the concentration of LTA at day 4 of MRSA infection was comparable to purified LTA at the same time point (fig. S3B), iNOS KO mice infected with MRSA maintained a significant reduction in lymphatic contraction (P < 0.05) (Fig. 3D). Notably, 60 days after MRSA infection, iNOS was undetectable in hindlimb tissue compared to 4 days after infection (fig. S3C), and animals injected with LTA showed normal lymphatic contraction 30 and 60 days after LTA injection (fig. S3D).

Fig. 3 Inhibition of iNOS is not sufficient to restore CLV contraction during MRSA infection.

(A) Representative sections of control skin (from uninfected mice) and skin 4 days after MRSA infection. Lymphatic vessels were identified with anti–LYVE-1 antibody (red) and nuclei with 4′,6-diamidino-2-phenylindole (DAPI) (blue). Anti-iNOS antibody (green) identified iNOS-positive cells; n = 4. Scale bar, 300 μm. (B) Serial sections to the images in (A) were stained with anti–Gr-1 antibody (blue), anti-iNOS antibody (green), and anti-Ly6G antibody (red); n = 4. Scale bar, 50 μm. (C) Sterile PBS (50 μl) or LTA (50 μl) suspended in sterile PBS was injected into the hindlimb of C57BL/6 and C57BL/6-iNOS−/− mice. On day 4 after injection, the ejection fraction shows the strength of PLV contraction from the indicated concentrations of LTA (n = 3 to 4, each group). (D) C57BL/6 mice were injected subcutaneously in the hindlimb with PBS (uninfected) or infected subcutaneously in the hindlimb with 2 × 106 to 4 × 106 CFU of WT MRSA or MRSA deficient in nitric oxide production (ΔsaNOS). C57BL/6-iNOS−/− mice were infected subcutaneously in the hindlimb with 2 × 106 to 4 × 106 CFU of WT MRSA or ΔsaNOS MRSA. Ejection fraction 4 days after infection shows the strength of PLV contraction among respective groups (n = 8 to 11, each group). For (C) and (D), one-way analysis of variance (ANOVA) comparison with Fisher’s least significant difference post hoc analysis was used to determine significance. Error bars show SEM. *P < 0.05.

Another source of nitric oxide is MRSA itself, which uses nitric oxide synthase oxygenase (saNOS) to catalyze nitric oxide production from l-arginine (32). Infection of wild-type (WT) mice with MRSA lacking saNOS (ΔsaNOS) resulted in reduced lymphatic vessel contraction (Fig. 3D). Similarly, infection of iNOS KO mice with ΔsaNOS MRSA did not improve PLV contraction (Fig. 3D). These results suggest that nitric oxide—produced by iNOS or bacterial saNOS—is not the sole inhibitor of lymphatic contraction during and after MRSA infection.

Effect of inhibiting inflammation on lymphatic function during infection

The inflammatory cytokines IL-1β, TNF-α (tumor necrosis factor–α) (both elevated in Fig. 2B), and IL-6 have been previously shown to decrease lymphatic contraction and lymph flow (33). Animals null for MyD88 (MyD88 KO), an intracellular adaptor protein for inflammatory signaling pathways downstream of Toll-like receptors, show attenuated production of these cytokines after intravenous S. aureus challenge (34, 35). However, lymphatic contraction in MRSA-infected MyD88 KO mice revealed no improvement in lymphatic function (measured by intravital microscopy) 4 days after infection, compared to infected WT mice (fig. S3E). In addition, prophylactic and therapeutic neutralization of TNF-α yielded no benefit in lymphatic function during MRSA infection (fig. S3F). Finally, mice treated prophylactically and during active infection with nonsteroidal anti-inflammatory drugs—the cyclooxygenase inhibitor acetylsalicylic acid (aspirin) or etodolac—did not exhibit enhanced lymphatic contraction (fig. S3G) relative to vehicle control.

Loss of LMCs after MRSA infection

Because reducing host inflammation and inflammatory cytokines did not improve lymphatic function (Fig. 3 and fig. S3, A, E, and G), we analyzed the cellular integrity of PLVs after MRSA infection. To this end, we performed immunohistochemical staining on hindlimb tissue. Using CD31 to identify blood and lymphatic vessels (fig. S4A), we found decreased α-smooth muscle actin (αSMA) staining of the PLV and adjacent posterior tibial artery 36 and 96 hours after infection (Fig. 4A and fig. S4C). However, no apoptotic αSMA+ cells were detected at these times using TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) staining (Fig. 4A and fig. S4C). Because this finding suggested decreased lymphatic muscle and smooth muscle cell (SMC) coverage, we next used mice that express DsRed under the control of the αSMA gene promoter (αSMA-DsRed) (19) to identify the LMCs on lymphatic vessels highlighted by fluorescein isothiocyanate (FITC)–dextran lymphangiography (Fig. 4B). LMC coverage of the PLV was decreased and DsRed localization was altered 4 days after infection (P < 0.05) (Fig. 4, B and C). In addition, loss of SMC coverage was measured in the posterior tibial artery (Fig. 4, A and B, and fig. S4C). LMC coverage remained abnormal 30 days after infection (Fig. 4, C and D), and LMCs showed an elongated pattern of αSMA distribution after infection (Fig. 4C). To confirm LMC loss, we analyzed the expression of αSMA by immunofluorescence using optically cleared hindlimb tissue. In agreement with findings from αSMA-DsRed reporter mice, infection of WT mice led to reduced LMC coverage 4 days after infection (P < 0.05) with partial recovery 30 days after infection (Fig. 4, E and F). Surprisingly, 260 days after infection, decreased LMC coverage persisted (P < 0.05) and LMCs displayed abnormal morphology and distribution of αSMA compared to the corresponding uninfected contralateral leg of the same animal (Fig. 4, G and H).

Fig. 4 Infection-induced CLV dysfunction is associated with decreased LMC coverage.

(A) Representative immunohistochemical images of uninfected mouse hindlimb and mouse hindlimb 36 hours after infection stained with αSMA (crimson, arrowhead) and TUNEL (brown); n = 4 mice per group. Scale bar, 40 μm. (B) Representative intravital image from segment of PLV of an αSMAPDsRed/C57BL/6 mouse interstitially injected with FITC-dextran; n = 4 mice per group. Arrowhead indicates lymphatic vessel. Arrow points to blood vessel. Scale bar, 100 μm. (C) Representative intravital image from PLV segments of αSMAPDsRed/C57BL/6 mice on the indicated days. Scale bar, 100 μm. (D) Quantification of αSMA-positive cells per 100 μm of PLV; n = 4 to 8 mice per group. (E) Representative image from PLV segments of C57BL/6 mice, stained with anti-αSMA, on the indicated days. Scale bar, 100 μm. (F) Computer-automated quantification of % αSMA-positive area of PLV; n = 3 to 4 mice for each group. (G) Representative image from a segment of PLV stained with anti-αSMA (red) 260 days after infection and corresponding control (contralateral) PLV. Scale bar, 25 μm. Arrowheads indicate adjacent blood vessel (posterior tibial artery). (H) Computer-automated quantification of % αSMA coverage of PLV 260 days after infection or corresponding age-matched control (contralateral) popliteal lymphatic vessel (n = 5, both groups). For (D) and (F), statistical analysis was performed using one-way ANOVA comparison with Fisher’s least significant difference post hoc analysis. For (H), statistical analysis was performed using Student’s unpaired two-sided t test. Error bars show SEM. *P < 0.05.

MRSA-induced LMC death

MRSA can damage biological membranes in several cell types, leading to cell death (36). We investigated whether MRSA could cause death of LMCs. To this end, we isolated LMCs from murine PLVs and cultured them in vitro (fig. S5A). After only 6 hours of incubation with MRSA-conditioned supernatant, LMC viability in vitro was significantly diminished compared to control medium [tryptic soy broth (TSB)] (P < 0.05) (Fig. 5A). Because LMCs are critical for lymphatic contraction, we sought to further assess the mechanism of LMC death. To test the relevance to human disease mechanisms, we also added MRSA-conditioned supernatant to human SMCs (hSMCs) (Fig. 5B). To determine whether the relevant secreted factors in the MRSA-conditioned supernatant were proteins, we pretreated MRSA-conditioned supernatant with trypsin. Trypsin abolished the lethality of MRSA-conditioned supernatant on hSMCs (Fig. 5B), suggesting that proteins mediate the observed cell death. Within abscesses, we found Gr-1+ cells that were also TUNEL-positive (fig. S6, A and B). Because we did not identify TUNEL-positive LMCs (Fig. 4A), we next asked whether cellular lysis represented a mechanism of MRSA-induced LMC death by measuring the release of lactate dehydrogenase (LDH) into the medium. The LDH release assay indicated MRSA cytolysis of hSMCs as early as 6 hours after exposure to MRSA-conditioned medium (Fig. 5C).

Fig. 5 MRSA protein causes muscle cell death.

(A) Cell viability [relative luciferase units (RLU)] analysis of hSMCs after exposure for 24 hours to control medium (TSB), MRSA-conditioned medium, or MRSA-conditioned medium incubated for 18 to 24 hours at 37°C with trypsin; n = 3 independent experiments. (B) Graph depicting LDH release from hSMCs (left y axis) and cell viability (RLU, right y axis) after exposure for 6 hours to control medium (TSB), MRSA-conditioned medium, or doxycycline; n = 3 independent experiments. (C) Mouse LMC viability analysis (RLU) after exposure for 6 hours to control medium (TSB) or (WT) MRSA-conditioned medium; n = 3 independent experiments. For (A), one-way ANOVA comparison with Fisher’s least significant difference post hoc analysis was performed. For (B), statistical analysis was performed using multiple t tests for comparators indicated. For (C), statistical analysis was performed using Student’s unpaired two-sided t test compared to the control group. Error bars show SEM. *P < 0.05. N.D., not detected.

Death of LMCs from MRSA toxins

Because MRSA-conditioned supernatant was able to kill LMCs (Fig. 5A), we next used mass spectrometry to analyze the cell-free conditioned medium from the MRSA strain USA300 JE2 and found 233 proteins that were highly enriched in the conditioned medium compared to control medium (Fig. 6A). Toxin expression in the MRSA-conditioned supernatant was abundant (see full list of proteins identified in table S1), accounting for 9% of identified proteins with known function (fig. S7A). Expression of many S. aureus toxins is primarily dependent on the accessory gene regulator (agr) operon (Fig. 6B) (36). On the basis of their high expression in MRSA supernatant, we chose to investigate the effect of δ-hemolysin, α-hemolysin, and phenol-soluble modulin α-1 (PSMα1) on LMCs.

Fig. 6 Effect of MRSA toxins on LMCs and CLV function.

(A) Identification by mass spectrometry of the relative abundance of proteins (area) from <100-kDa fraction of MRSA-conditioned medium. (B) Heat map of the most highly expressed toxins, as identified by mass spectrometry; also depicted is the corresponding toxin expression level in the absence of agr. (C) Graph depicting LDH release from LMCs (left y axis) and cell viability (RLU, right y axis) after incubation for 24 hours with normal growth medium only (−), α-hemolysin (1.0 μg/ml), δ-hemolysin (100 μg/ml), or PSMα1 (100 μg/ml). Statistical analysis was performed using multiple t tests relative to untreated control for cytolysis and cell viability. *P < 0.05; n = 3 independent experiments. (D) Cell viability analysis of murine LMCs (mLMCs) after exposure for 24 hours to control medium (TSB), MRSA-conditioned medium, or conditioned medium of isogenic MRSA mutants, as indicated; n = 3 independent experiments. *P < 0.05 relative to control medium. hla, α-hemolysin; hld, δ-hemolysin. (E) Lymphatic ejection fraction (n = 4 to 5 mice for each group) and (F) frequency (n = 5 to 6 mice for each group) of lymphatic vessel contraction was measured in mice infected with WT or the indicated mutant MRSA strains. (G) Computer-automated quantification of % αSMA-positive area of PLV in animals 30 days after infection with WT or the indicated mutant MRSA strain; n = 3 to 4 mice for each group. For (D) to (G), statistical analysis was performed by one-way ANOVA comparison with Fisher’s least significance difference post hoc analysis. Error bars show SEM. *P < 0.05.

To this end, we added recombinant δ-hemolysin, α-hemolysin, or PSMα1 to LMCs in vitro. Incubation of each toxin with LMCs led to cell lysis, with PSMα1 causing the greatest degree of cytolysis (Fig. 6C). However, genetic deletion of δ-hemolysin (hld) or α-hemolysin (hla) did not render LMCs viable after incubation with conditioned medium from these respective mutants (Fig. 6D). Incubation of conditioned medium from a genetic deletion of psmα (including psmα1 to psmα4) resulted in minimal improvement in LMC viability, also seen in a mutant with combined disruption of the psmα, psmβ, and hld (combined deletion) loci (Fig. 6D). In vivo, the strength (Fig. 6E) and frequency (Fig. 6F) of lymphatic contraction in mice infected with hld, psmα, and combined deletion mutants remained significantly reduced after 30 days relative to phosphate-buffered saline (PBS) sham-infected mice (P < 0.05). LMC coverage after infection with mutants was comparable to that of WT MRSA (Fig. 6G). Together, these data suggest a role for additional agr-dependent factors toward LMC lethality. To this point, an α-hemolysin–neutralizing antibody was sufficient to prevent hSMC viability loss from recombinant α-hemolysin (fig. S7B) (P < 0.05) but had no effect on the lethality from MRSA-conditioned supernatant (P > 0.05) (fig. S7C).

Effect of MRSA on additional cells associated with blood and lymphatic vasculature

The loss of SMCs on blood vessels during acute MRSA infections (Fig. 4, A and B) suggested that additional cell types may be affected by MRSA toxins. We tested the ability of MRSA-conditioned medium to kill other vascular cell types, including both murine and human vascular SMCs, as well as human LECs. MRSA-conditioned supernatant killed hSMCs and murine SMCs (mSMCs) (fig. S7, D and E) and cultured primary human LECs (fig. S7F). Together, these data suggest that MRSA cytotoxicity is not limited to LMCs and explain the loss of SMCs of the posterior tibial artery. The requirement of α-hemolysin, δ-hemolysin, and PSMs for cell toxicity was cell-specific, because loss of δ-hemolysin and PSMα led to less killing of hSMCs (fig. S7, D and G) and mSMCs (fig. S7E) relative to LMCs (Fig. 6D) in vitro. Further, deletion of hla and psmβ in MRSA resulted in improved viability of LECs (fig. S7F).

agr-dependent inhibition of CLV function

Next, we used an agr mutant of the USA300 strain of MRSA, which has diminished production of several virulence-associated genes including hld and hla (37). Further, the agr mutant has no detectable PSM production (38). Unlike WT MRSA-conditioned supernatant, agr mutant–conditioned supernatant was not cytotoxic and did not cause death of LMCs, SMCs, or LECs (Fig. 7, A and B, and fig. S7, B to D). In vivo, mice infected with agr mutant MRSA had LMC coverage comparable to uninfected mice (Fig. 7C). Further, agr mutant–infected mice showed stronger lymphatic function relative to mice infected with the parental strain at day 30, which is after clearance of MRSA (Fig. 7D and movies S3 and S4). The strength of PLV contraction (ejection fraction) in agr mutant–infected mice was 58% stronger relative to the contraction of animals infected with WT MRSA (P < 0.05). Several agr mutant–infected animals showed contraction strength similar to that of uninfected mice (Fig. 7E). Moreover, the frequency of lymphatic vessel contraction was significantly faster in agr mutant–infected mice compared to WT MRSA (Fig. 7F) (P < 0.05). Notably, the contraction frequency of agr mutant–infected mice was increased even relative to uninfected mice, potentially compensating for the moderately lower ejection fraction. Together, these data suggest that agr-dependent MRSA toxins are primarily responsible for long-term inhibition of CLV function.

Fig. 7 MRSA virulence proteins cause LMC death and diminished CLV function.

(A) mLMC viability analysis after exposure for 24 hours to control medium (TSB), agr mutant–conditioned medium, or MRSA-conditioned medium; n = 3 independent experiments. (B) LDH release from LMCs incubated with conditioned medium from WT or agr mutant MRSA; n = 3 independent experiments. (C) Computer-automated quantification of % αSMA-positive area of PLV in uninfected animals and animals 30 days after agr mutant MRSA infection; n = 3 to 4 mice for each group. (D) Representative traces from lymphatic vessel wall measurements show lymphatic diameter and contraction over time in uninfected mice and mice infected with WT or agr mutant MRSA for 30 days. (E) Ejection fraction shows the strength of lymphatic vessel contraction in uninfected mice and mice infected with WT or agr mutant MRSA for 30 days; n = 8 to 17. (F) Frequency indicates the cpm of lymphatic vessel contractions in uninfected mice and mice infected with WT or agr mutant MRSA for 30 days; n = 8 to 15. For (E) and (F), statistical analysis was performed by one-way ANOVA comparison with Fisher’s least significance difference post hoc analysis. Error bars show SEM. *P < 0.05.


MRSA compromises and evades host immune responses by several mechanisms, including, but not limited to, suppressing chemotaxis of leukocytes, toxin-mediated killing of leukocytes, and the production of superantigens to impair the immune response (39). Here, we used the endemic USA300 strain of community-associated MRSA (CA-MRSA), a common cause of serious bacterial infections in the United States, to investigate its effect on lymphatic vessels in a murine model of infection. Our data show that a single MRSA infection inhibits lymphatic vessel contraction and lymph flow long after the infection clears, providing another mechanism of how MRSA compromises immune responses.

Several mechanisms have been shown to alter the strength of lymphatic contraction. These include physical, neural, and humoral influences (15). In particular, nitric oxide and inflammatory cytokine concentrations are able to dictate lymphatic contraction frequency and amplitude (19, 40, 41). Although we measured a marked increase in inflammatory cytokine production and nitric oxide–producing infiltrating cells at the peak of MRSA infection, several lines of evidence support the conclusion that nitric oxide and other inflammatory signaling mediators are not the cause of long-term lymphatic dysfunction after MRSA infections. The absence of inflammatory molecules, even though CLV function was still impaired after MRSA clearance, suggested underlying damage to CLVs established during the acute stage of infection.

From this study, we report a new mechanism of lymphatic impairment. In vitro and in vivo, we show that MRSA causes LMC cell death, suggesting a causal connection between LMC loss and CLV dysfunction. We found that agr-dependent proteins, including α-hemolysin, δ-hemolysin, and PSMα (38), are able to kill LMCs. Consistent with protection of LMCs from cytotoxicity caused by agr mutant–conditioned medium in vitro, animals infected with an agr null mutant exhibited normal CLV function.

In agreement with other studies (42), the β-type PSM peptides were less cytolytic toward LMCs and SMCs in our study than the α-type PSM peptides. Several PSMs are able to lyse a variety of eukaryotic cell types (42). It has been proposed that this is due to membrane perturbation of cells in a receptor-independent manner (43). α-Hemolysin also showed the ability to kill LMCs in vitro, but deletion of α-hemolysin was not sufficient to rescue LMCs from MRSA-induced death in vitro and in vivo. α-Hemolysin is known to cause indiscriminate cell death at high concentrations (44), and this may explain the discrepancy in our results with recombinant α-hemolysin and mutants deficient in α-hemolysin. Together, our results suggest that agr-dependent production of α-hemolysin, δ-hemolysin, and PSMα cause LMC death in vitro. However, we cannot rule out synergism or redundancy with other agr-controlled toxins in contributing to LMC death in vivo, particularly because inhibiting the production of many toxins by mutating agr led to the maintenance of lymphatic function.

The death of LMCs and the apparent slow regeneration of LMCs explain how CLV function can be impaired long after MRSA have been cleared and the toxins are no longer present in the system. Further, the coupling of slow LMC regeneration with MRSA recurrence may explain the cyclical pattern of lymphatic deterioration and reinfection in some SSTI patients. Investigations into therapeutic approaches to accelerate LMC regeneration may provide an alternative treatment to restore lymphatic function in patients with recurrent SSTIs.

The identification of a bacterial-derived mechanism for LMC death represents a critical first step in a mechanistic analysis of how MRSA impairs CLV function. It is unknown whether the altered morphology of LMCs that regenerate after infection is indicative of functional differences compared to LMCs of uninfected mice. We also cannot rule out effects of MRSA on LECs in vivo, which also contribute to the maintenance of lymphatic vessel contraction. However, prophylactic treatment of recurrent cellulitis infections (45) by targeting agr signaling may overcome the effects exerted by MRSA toxins on multiple different cell types and lead to the preservation of CLV function and immunity. Several studies have focused on identifying chemical inhibitors of agr to combat the virulence of bacterial infections and to limit antibiotic use (4650). Although these inhibitors broadly limit pathological tissue damage, it is unknown whether they also help maintain lymphatic function.

Although the ability to evaluate CLV function provides insight into the effect of MRSA after infection, our model is limited in that it is not a recurrent model of infection. Further, animals are maintained in a controlled sterile environment, and mice do not develop secondary chronic lymphedema. Another limitation of this study is that it does not address a role, if any, for Streptococci and methicillin-susceptible S. aureus (MSSA) on lymphatic impairment, because group A Streptococci and MSSA are also common pathogens associated with SSTIs (4). It is likely that MSSA can blunt CLV function through agr-mediated toxin production, albeit MSSA toxin production is lower than that of MRSA (37).

In lymphedema patients, recurring episodes of SSTIs lead to further damage of the lymphatic system, resulting in lymph flow deterioration (5153). Emerging evidence suggests that both T and B lymphocyte–mediated adaptive immune responses are generated after S. aureus infection (54). In patients with preexisting lymphedema or progressive lymphedema as a result of recurring infections, we expect local immune deficiency due to the ineffective transport of antigen to the local draining lymph nodes, representing another mechanism by which S. aureus inhibits the adaptive immune response. When lymphatic function is impaired, the protein-rich and stagnant lymphatic fluid may facilitate bacterial growth. Further, impaired lymphatic function also inhibits bacterial clearance through the lymphatic system, resulting in bacterial and toxin accumulation that will impair both innate and adaptive immunity and increase the risk of recurrent infections. Our data provide a mechanistic framework for the chronic reduction of lymph flow in patients after MRSA infection. Neutralization of MRSA pathogenicity as part of the treatment of MRSA-induced cellulitis may help intervene in the vicious cycle of SSTI, lymph flow deterioration, lymphedema, and recurrent infection in these patients.


Study design

The goal of this study was to determine the effect of MRSA infection on CLV function. Using a model of CA-MRSA infection in the hindlimb of mice, we assessed CLV function by performing lymphatic contraction experiments previously established (19, 25). We deemed a 25% difference in lymphatic contraction or lymph flow with an anticipated SD of 20% as biologically significant. At an α of 0.05 and a β of 0.80, we determined sample size based on previous and gained experience regarding the success rate of lymphatic contraction and lymph flow, respectively. No animals were excluded from lymphatic contraction studies pending successful surgical exposure of lymphatic vessel. The criteria for exclusion of animals for lymph flow experiments with DOCT are described in Materials and Methods and involved the ability to obtain a clear image of the vessel and calculate a Doppler angle of less than 85°. αSMA-DsRed transgenic mice were visualized and immunofluorescence staining of αSMA+ LMCs of WT mice was performed to analyze LMC coverage of lymphatic vessels on multiple days after infection. Each analysis represents a terminal experiment, not a longitudinal study of the same animal. Researchers were blinded to the details of the infection during image analysis. Both male and female mice were used and were age-matched between all comparator groups. Cell viability and cell lysis assays were used to measure killing of LMCs, hSMCs, mSMCs, and LECs by MRSA-conditioned supernatant from strain(s) simultaneously grown overnight in similar conditions. Cell culture experiments to investigate the effects of MRSA-conditioned supernatant or recombinant S. aureus toxins on cells were performed in duplicate or triplicate in three to five independent experiments. All data are means ± SEM, unless stated otherwise in the figure legends. The methods used to support these findings are described in Materials and Methods/Supplementary Materials and Methods. Statistical tests used to establish significance and associated P values are described in the figure legends. Primary data are located in table S2.


In vivo studies were initiated in 5- to 8-week-old male and female C57BL/6, iNOS−/− C57BL/6, MyD88−/− C57BL/6 (gift from A. Fassano), LysMgfp/gfp C57BL/6 (gift from T. Graf), and αSMAP-DsRed/C57BL/6 mice. All mice were bred and maintained in our Cox-7 gnotobiotic animal colony or at the Center for Comparative Medicine and Animal Laboratory Resources at Massachusetts General Hospital. All procedures were performed following the guidelines of the Institutional Animal Care and Use Committee of the Massachusetts General Hospital.

S. aureus strains

The clinical MRSA isolate, CA-MRSA USA300 LAC, had three nonessential plasmids removed to create the CA-MRSA USA300 JE2 strain (55). In addition, the NE27 mutant (lacking functional saNOS) was created by mutagenesis from mariner-based transposon bursa aurealis, resulting in an erythromycin-resistant deletion strain of JE2. Both isolates were obtained from the Network on Antimicrobial Resistance in S. aureus. WT and isogenic deletion mutants of the agr regulatory system, α-hemolysin, δ-hemolysin, psmα, psmβ, and psmαβhld (TKO) in the LAC (USA300) strain are described previously (37, 38, 5658).


MRSA strains were subcultured until they reached their exponential growth phase, washed with sterile PBS once, and resuspended in sterile PBS. Mice were anesthetized with intraperitoneal administration of ketamine (10 mg/kg)/xylazine (100 mg/kg). Fur was shaved from the hindlimb. MRSA (2 × 106 to 4 × 106 CFU) was injected subcutaneously (using a 29-gauge 1/2-inch insulin syringe) into the hindlimb, proximal to the ankle, in a 50-μl volume of sterile PBS. A similar procedure was performed with LTA (Invivogen, catalog no. tlrl-pslta) after resuspension in sterile PBS.

Isolation of mLMCs and blood SMCs

While in solution of Hanks’ balanced salt solution, fat and surrounding connective tissue were removed from the posterior tibial artery and afferent lymphatic vessels adjacent to the PLN. To make 5 ml of enzyme solution, Collagenase Type II and Soybean Trypsin Inhibitor (Worthington) were added to Hanks’ balanced salt solution at 1 mg/ml. Elastase (35 μl; Worthington) and penicillin/streptomycin (50 μl) were added. Next, the enzyme solution was added to minced fragments of popliteal artery or lymphatic vessels and allowed to incubate at 37°C, 5% CO2 for 4 hours with occasional agitation. Cell outgrowth was allowed to proceed for 5 days.

Preparation of MRSA-conditioned supernatant

Planktonic culture-conditioned medium was prepared by growing an overnight culture of the indicated MRSA strain in TSB (BD Biosciences) under shaking conditions (200 rpm) at 37°C for 12 to 18 hours. Afterward, the bacteria were centrifuged at 6000 rpm for 10 min, and subsequently, the supernatant was filtered through a 0.22-μm pore-size polyethersulfone filter (Olympus Plastics) to produce cell-free conditioned medium.

The cell-free conditioned MRSA medium was mixed at a 1:1 dilution with muscle cell culture medium. A 1:1 dilution of cell culture medium with TSB medium was used as a control in parallel. Where indicated, MRSA-conditioned medium was treated with 0.0625% trypsin (final percentage in solution) for 18 to 24 hours at 37°C to degrade proteins. Trypsin was heat-inactivated at 100°C for 1 hour and allowed to cool to room temperature before supernatant was incubated with cells.

Lymphatic contraction model

Mice were anesthetized with ketamine/xylazine, and 3 μl of 2% FITC-dextran (2 million molecular weight; Thermo Fisher Scientific, catalog no. D7137) was interstitially injected in the footpad. Surgical preparation and intravital imaging were performed as described (19, 25). Briefly, the leg skin and underlying connective tissue near the afferent lymphatic vessel to the PLN (PLV) were carefully excised. Time-lapse images (360 images separated by 210 ms) of the exposed afferent lymphatic vessels to the PLN were captured while imaging with an inverted fluorescence microscope. For each mouse, we imaged two to three lymphatic segments and took images every 15 min for four time points. Using in-house MATLAB codes, time-lapse images were analyzed to track the position of the vessel wall and to measure frequency using the peak-and-valley method as described previously (19, 25). For most analyses, we collected 6 to 10 measurements of ejection fraction values from a single mouse. The mean of the measurements was used to represent lymphatic contraction for that animal.

Measuring lymphatic flow

To measure lymph flow inside murine CLVs, we used a recently developed label-free method for directly measuring lymph flow velocity in vivo by DOCT (26). This technique measures the velocity of lymph fluid, both basal and pulsatile, and relies only upon the absolute movement of lymph relative to the laser beam. Mice were anesthetized with ketamine/xylazine. The surgical procedure to expose the PLV was performed as described previously (19, 25). Anesthetized mice were transferred to a heating pad to maintain body temperature at 37°C.

Hydration of the PLV and surrounding tissue was maintained with physiological saline while measuring lymph flow. Repeated 5-min DOCT measurements were acquired from a fixed location within a lymphangion, in between lymphatic valves. Next, the measurements were processed to produce a structural image, and the scattering signals within the lymphatic vessel were analyzed using an algorithm designed specifically to measure the lymph flow signal. Three-dimensional imaging was used to measure the lymphatic vessel orientation and thereby the Doppler angle of the flow, the latter being mandatory to calculate flow speed (from Doppler shift). For these measurements, a structural image showing a clear vessel-tissue boundary and absence of signal artifact within the vessel lumen was of paramount importance. In addition, a Doppler angle less than 85° and a lymphatic vessel within 100 μm from the tissue surface were needed to guarantee trustworthy results. Data not adhering to these criteria were excluded for analysis.

Twenty-four 8- to 12-week-old C57BL/6 female mice were infected with 2 × 106 to 4 × 106 CFU of MRSA USA300 JE2, and lymphatic flow was measured at either 4 days (n = 8), 35 days (n = 16), or 260 days after infection (n = 7). Eight age-matched mice served as uninfected controls for each time point. Lymphatic flow was measured using M-mode acquisition until two measurements per mouse were obtained without consequential x, y, or z drift, as judged by the investigators. Mouse limbs were positioned lower than the abdomen on a 15° tilted stage, and measurements were acquired and subsequently analyzed as described previously (26). In the control group, 19 of 25 measurements were successful in six of eight mice. Four days after CA-MRSA infection, four of eight mice could not be imaged because of excessive postoperative bleeding resulting from severe inflammation during active bacterial infection. In the remaining four mice, 3 of 13 measurements were successful, all from the same mouse. After 35 days, 17 of 46 measurements were successful in 6 of 12 mice. In this group, four additional mice did not show intact collecting vessels within 100 μm from the tissue surface, and therefore, no measurements could be acquired.

Cell viability assays

hSMCs (10,000 to 20,000) or mLMCs (1000 to 5000) delivered in 200 μl of respective medium were allowed to adhere overnight to the bottom of a 48-well plate. After 24 to 48 hours of exposure to MRSA-conditioned supernatant or TSB, the CellTiter-Glo Luminescent Cell Viability Assay (Promega) was performed according to instructions. Briefly, 100 μl of medium was removed from the wells of the plates containing medium, and 100 μl of room temperature CellTiter-Glo Reagent was added. Plates were rocked for 2 min in the dark and incubated for an additional 8 min or longer (while protected from light) before reading luminescent output on a FLUOstar Omega series luminescence microplate reader.

Statistical analysis

Statistical analyses were completed with Prism 7 (GraphPad). Statistical significance was determined with either a one-way ANOVA with the Fisher’s least significance difference post hoc test or two-tailed unpaired Student’s t test, as appropriate. Statistical significance was set at P < 0.05. All statistical tests were two-sided.


Materials and Methods

Fig. S1. MRSA infection leads to acute inflammation and CLV dilation.

Fig. S2. CLV function remains impaired after the clearance of MRSA.

Fig. S3. Attenuation of host inflammation is not sufficient to restore CLV contraction during MRSA infection.

Fig. S4. Effect of MRSA infection on the blood vasculature.

Fig. S5. Characterization of LMCs, vascular muscle cells, and LECs.

Fig. S6. Identification of apoptotic cells in skin abscesses caused by MRSA infection.

Fig. S7. MRSA toxins contribute to reduced hSMC and LEC viability.

Movie S1. Representative lymphatic contraction of afferent CLV to the PLN in uninfected mouse.

Movie S2. Representative lymphatic contraction of afferent CLV to the PLN in MRSA-infected (CA-MRSA USA300 JE2 strain) mouse 4 days after infection.

Movie S3. Representative lymphatic contraction of afferent CLV to the PLN in MRSA-infected (CA-MRSA USA300 strain) mouse 30 days after infection.

Movie S4. Representative lymphatic contraction of popliteal lymphatic vessel afferent CLV to the PLN in agr mutant–infected mouse 30 days after infection.

Table S1. List of proteins identified by mass spectrometry of supernatant from WT CA-MRSA USA300 JE2 strain (provided as an Excel file).

Table S2. Primary data (provided as an Excel file).

Reference (59)


Acknowledgments: We thank J. C. Lee (Harvard Medical School) for critical discussion and J. Bubeck-Wardenburg (University of Chicago) for supplying the α-hemolysin mutant MRSA strain. We also thank Janssen Biotech Inc. for contributing the blocking anti–TNF-α monoclonal antibody. We thank the Harvard University Mass Spectrometry and Proteomics Resource Laboratory for sample analysis. Funding: This work was supported by the NIH under award numbers R21AI097745 (T.P.P.), DP2OD008780 (T.P.P.), R01CA214913 (T.P.P.), and R01HL128168 (T.P.P. and L.L.M.). Research reported in this publication was supported in part by the Center for Biomedical OCT Research and Translation through grant number P41EB015903, awarded by the National Institute of Biomedical Imaging and Bioengineering of the NIH. This work as supported in part by the National Cancer Institute Federal Share of Proton Income (CA059267; T.P.P. and B.J.V.), National Cancer Institute (R01CA163528; B.J.V.), Massachusetts General Hospital Executive Committee on Research ISF (T.P.P.), and Intramural Research Program of the National Institute of Allergy and Infectious Diseases (grant number ZIA AI000904-16; M.O.). This work was also supported in part by the United Negro College Fund–Merck Science Initiative Postdoctoral Fellowship (D.J.), Burroughs Wellcome Fund Postdoctoral Enrichment Program Award (D.J.), NIH National Cancer Institute (F32CA183465; D.J.), and Swiss National Science Foundation (C.B.). Author contributions: D.J. and T.P.P. conceived and designed the study and analyzed data. D.J., E.F.J.M., E.R.P., E.M.B., K.J., and S.M.C. designed and performed experiments and analyzed data. C.B. and B.J.V. conducted imaging analyses and contributed imaging tools. P.H. contributed αSMAP-DsRed/C57BL/6 mice. S.L. and L.L.M. contributed to the development of imaging tools. M.O. contributed MRSA strains and data interpretation. D.J. and T.P.P. wrote the manuscript, which all coauthors commented on. Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Translational Medicine

Navigate This Article