VEGF-A from Granuloma Macrophages Regulates Granulomatous Inflammation by a Non-angiogenic Pathway during Mycobacterial Infection

VEGF-A from Granuloma Macrophages Regulates Granulomatous Inflammation by a Non-angiogenic Pathway during Mycobacterial InfectionGraphical Abstract
A-producing macrophages
d VEGF-A recruits macrophages to the granuloma via a non-
angiogenic pathway
limited effect on protection
resistant to Mtb infection
Harding et al., 2019, Cell Reports 27, 2119–2131 May 14, 2019 ª 2019 The Authors. https://doi.org/10.1016/j.celrep.2019.04.072
Authors
Yuli Chen, ..., Zsuzsanna Fabry,
Andras Nagy, Matyas Sandor
inhibitors reduce granuloma-associated
pathologies, including mycobacterial-
heterogeneous inflammatory cell
death. VEGF-A potentiates excessive
inflammation by recruiting macrophages
inhibitors exist and could help limit
tuberculosis-associated inflammation.
Iacovos P. Michael,5 Zsuzsanna Fabry,1,2 Andras Nagy,5,6 and Matyas Sandor1,2,7,* 1Department of Pathology and Laboratory Medicine, School of Medicine and Public Health, University of Wisconsin-Madison, Madison, WI
53706, USA 2Cellular and Molecular Pathology Training Program, University of Wisconsin-Madison, Madison, WI 53706, USA 3Neuroscience Training Program, University of Wisconsin-Madison, Madison, WI 53706, USA 4Department of Anesthesiology, Irving Medical Center, Columbia University, New York, NY 10032, USA 5Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, ON M5T 3H7, Canada 6Department of Obstetrics and Gynecology, and Institute of Medical Science, University of Toronto, Toronto, ON M5S 1A8, Canada 7Lead Contact
*Correspondence: [email protected]
https://doi.org/10.1016/j.celrep.2019.04.072
SUMMARY
Many autoimmune and infectious diseases are char- acterized by the formation of granulomas which are inflammatory lesions that consist of spatially orga- nized immune cells. These sites protect the host and control pathogens like Mycobacterium tubercu- losis (Mtb), but are highly inflammatory and cause pathology. Using bacille Calmette-Guerin (BCG) and Mtb infection in mice that induce sarcoid or caseating granulomas, we show that a subpopula- tion of granuloma macrophages produces vascular endothelial growth factor (VEGF-A), which recruits immune cells to the granuloma by a non-angiogenic pathway. Selective blockade of VEGF-A in myeloid cells, combined with granuloma transplantation, shows that granuloma VEGF-A regulates granuloma- tous inflammation. The severity of granuloma-related inflammation can be ameliorated by pharmaceutical or genetic inhibition of VEGF-A, which improves sur- vival of mice infected with virulent Mtb without altering host protection. These data show that VEGF-A inhibitors could be used as a host-directed therapy against granulomatous diseases like tuber- culosis and sarcoidosis, thereby expanding the value of already existing and approved anti-VEGF-A drugs.
INTRODUCTION
fectious and non-infectious diseases including tuberculosis
(Davis and Ramakrishnan, 2008), leprosy (Turk and Bryceson,
1971), Crohn’s disease (Chambers andMorson, 1979), sarcoid-
osis (Chen and Moller, 2011), histoplasmosis (Kauffman, 2007),
and schistosomiasis (Pearce and MacDonald, 2002), among
Ce This is an open access article under the CC BY-N
many others. The granuloma is a collection of highly organized
innate and adaptive immune cells, and during tuberculosis it is
the site where Mycobacterium tuberculosis (Mtb) bacilli are
contained (Davis and Ramakrishnan, 2008). Granulomas are
highly dynamic sites and require complex regulation by host
factors to recruit and retain effector cells, and they can form
in almost any tissue of the body. Granuloma cells have finite life-
spans and so dying cells must be continuously replaced to sus-
tain Mtb-constraining inflammation (Schreiber et al., 2011a,
2011b).
In humans high levels of vascular endothelial growth factor
(VEGF-A) have been found in almost every granulomatous dis-
ease where it has been measured, including pulmonary tubercu-
losis (Abe et al., 2001; Alatas et al., 2004;Matsuyamaet al., 2000),
tuberculosis meningitis (Husain et al., 2008; van der Flier et al.,
2004), schistosomiasis (Shariati et al., 2011), Crohn’s disease
(Bousvaros et al., 1999; Griga et al., 1999, 2002; Kanazawa
et al., 2001), and sarcoidosis (Sekiya et al., 2003; Tolnay et al.,
1998), and is also present in human tubercle granulomas (Kang
et al., 2014). In patients with active tuberculosis, serum VEGF-A
levels decrease after antibiotic treatment (Alatas et al., 2004).
Additionally, specific VEGFR single nucleotide polymorphisms
correlate with susceptibility to sarcoidosis (Morohashi et al.,
2003; Pabst et al., 2010). VEGF-A can induce angiogenesis
(Leung et al., 1989; Plouet et al., 1989), release cells from the
bone marrow (Galiano et al., 2004; Hattori et al., 2001), increase
vascular permeability and cell extravasation through VEGFR2
(Dvorak et al., 1999; Senger et al., 1983), and attract monocytes
by chemotaxis via VEGFR1 (Barleon et al., 1996; Clauss et al.,
1996). VEGF-A is a powerful regulator over angiogenesis, and
there have been several recent reports about the effects of
VEGF-A-mediated angiogenesis during mycobacterial infec-
tions. In these reports, VEGF-A inhibition was associated with
decreased mycobacterial proliferation and survival in low vascu-
larized areas (Oehlers et al., 2015), normalization of aberrant
angiogenesis and increase in antibiotic access to infected tissue
(Datta et al., 2015), and decreased mycobacterial dissemination
resulting from decreased angiogenesis (Polena et al., 2016).
ll Reports 27, 2119–2131, May 14, 2019 ª 2019 The Authors. 2119 C-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Figure 1. VEGF-A Is Upregulated in CD11b+ Mycobacterial Granuloma Cells
(A) Granulomatous inflammation in the liver (i.p. injection, upper panel) or lung (intranasal, lower panels) after infection with the bacillus Calmette-Guerin (BCG)
isolate of Mycobacterium bovis. Micrographs taken 21 (i.p.) or 28 (intranasal) days after infection. White arrows show bacilli (Ziehl-Neelsen stain). Fluorescent
CD11b+-stained cells shown in red (upper panels) or green (lower panels). White dotted lines outline granulomas. Slice of entire left lung is shown.
(B) Upregulation of VEGF-A protein in supernatants of granuloma leukocytes, but not splenic, isolated and cultured ex vivo 21 days after i.p. BCG infection. n = 5
mice, two replicate experiments. ANOVA with Tukey’s post-hoc test; ***p < 0.001; error bars, mean ± SEM.
(C) VEGF-A and F4/80 staining on acute BCG liver granulomas showing variability of granuloma VEGF-A+ cell frequency among different granulomas.
(D) Granuloma size versus frequency of VEGF-A-producing cells (high, >10%; medium, 0.5%–10%; low <0.5%) in acute, BCG-infected liver. 107 randomly-
selected granulomas were sampled and analyzed with imageJ. Representative microphages from each group shown in (C). ANOVA with Tukey’s post hoc test,
***p < 0.001, error bars median with interquartile range
(legend continued on next page)
2120 Cell Reports 27, 2119–2131, May 14, 2019
Dissemination ofMtb bacilli is a well-described clinical feature of
tuberculosis disease.
VEGF-A during mycobacterial infection. Granuloma macro-
phages that differentiate from blood-derived monocytes prevent
unchecked proliferation, but they also provide a biological niche
for bacteria to persist, A subset of these macrophages acquires
a VEGF-A-producing phenotype. We show that macrophage-
produced VEGF-A recruits granuloma-forming cells to bacillus
Calmette-Guerin (BCG) and Mtb-infected tissue in a non-redun-
dant way that cannot be compensated for when VEGF-A is
blocked. We also show that this effect is independent of the vas-
culogenic and angiogenic effects of VEGF-A.
While granulomas are necessary to restrain certain patho-
gens, they are highly inflammatory foci by nature that can lead
to tissue injury as well as the pathology associated with infec-
tious and non-infectious granulomatous diseases. The eroding
lung function in patients with active pulmonary tuberculosis
stems from the accumulation of activated leukocytes in myco-
bacterium-containing granulomas. We also show that inhibition
of VEGF-A reduces the severity of granulomatous inflammation
without compromising host protection against BCG or Mtb.
These results support an expansion of existing anti-VEGF-A
therapies, which are already developed and tested in humans,
to treat granulomatous diseases like tuberculosis. This kind of
immunomodulatory therapy could decrease pathological inflam-
mation during active tuberculosis, give clinicians more time to
find effective antibiotics in patients with drug-resistant Mtb,
and may apply broadly to granuloma-associated diseases like
sarcoidosis.
RESULTS
VEGF-A Is Upregulated in Mycobacterium bovis BCG Granulomas In mice, infection with BCG, an attenuated isolate of Mycobac-
terium bovis, results in acute human-like sarcoid granulomas
21–28 days after infection (Figure 1A). Intraperitoneal (i.p.) injec-
tion results in granulomas mostly in the liver (Figure 1A, upper
panels), while granulomas forming after intranasal infection
are mostly restricted to the lungs (Figure 1A, lower panels).
Despite differences in size, these BCG-containing lesions are
both populated by CD11b+ cells, and the massive influx of
cells into infected areas encompass significant portions of
the total tissue area (Figure 1A, lower panels). Granuloma
leukocytes, but not splenic, isolated from BCG-infected mice
produced high levels of VEGF-A protein (Figure 1B), indicating
that increased VEGF-A is produced by local granulomas. To
confirm the presence of VEGF-A in BCG liver granulomas,
we co-stained liver sections with antibodies against VEGF-A
(E) Distribution of VEGF-A-producing granulomas from (D).
(F) VEGF-A locus activity in granuloma cells from BCG-infected mice. Represen
HypoVEGF (b-galactosidase fused to VEGF-A protein) mice stained with FDG an
(G) Quantification of data from (F) (far right row flow plots) showing the proporti
expression, as measured by FDG positivity (n = 3 mice). Two-tailed unpaired t te
(H) Normalized mean fluorescent intensity (MFI) of VEGF-A expression (measure
BCG granulomas. n = 3 mice. Two-tailed unpaired t test; *p < 0.05; **p < 0.01; e
and macrophage F4/80 (Figure 1C). These stainings verified
that granulomas are the primary source of VEGF-A and also
showed the variability in VEGF-A expression among different
granulomas (Figures 1D and 1E). In general, larger granulomas
contained a higher proportion of VEGF-A+ cells compared to
smaller ones, suggesting that VEGF-A production is coupled
to factors that are also highest in larger granulomas. Next, to
quantify the cellular source of granuloma VEGF-A, we analyzed
granuloma leukocytes by flow cytometry in HypoVEGF mice, a
transgenic VEGF-A reporter strain in which VEGF-A is trans-
lated as a fusion protein together with b-galactosidase. VEGF-A
expressing cells can be identified with the b-galactosidase-
specific fluorescent substrate FDG. CD11b+ cells in liver and
lung BCG granulomas expressed VEGF-A with the highest fre-
quency of any cell type, and there were 16-fold (liver) or 4-fold
(lung) more CD11b+ VEGF-A+ cells than CD11bVEGF-A+ cells
(Figure 1F, left two columns). Among the CD11b+VEGF-A+ cells,
a significantly higher proportion were Ly6Clow compared to
Ly6Chigh (Figures 1F, right column, and 1G). Furthermore,
CD11b+ cells had a significantly higher intensity of VEGF-A
expression compared to CD11b cells (Figure 1H). Together,
these data show that a subset of CD11b+ cells are the most
abundant source of VEGF-A protein in BCG liver and lung gran-
ulomas. Since monocytes lose Ly6C expression as they exit the
blood and differentiate, these data also suggest that this subset
acquires their VEGF-A+ phenotype after exposure to the granu-
loma compartment.
Blockade of VEGF-A Reduces Granulomatous Inflammation without Compromising Bacterial Containment We then tested if VEGF-A was important in the in vivo regulation
of granulomatous inflammation using a murine model of acute
BCG infection coupled with genetic or pharmaceutical inhibi-
tion of VEGF-A. We correlated reductions in VEGF-A activity
to changes in granuloma and leukocyte accumulation in the
liver and asked if the resulting phenotype suppressed the host’s
control over infection as measured by bacterial burden. For
each VEGF-A-inhibited and control group, several thousand
granulomas were counted among several regions of tissue (Fig-
ures 2A, 2B, and S1). We first tested the HypoVEGFmice strain,
which express a hypomorphic VEGF-A allele with reduced
VEGF-A binding to the VEGF receptor (Damert et al., 2002).
Mousemodels with total genetic ablation of VEGF-A are unavai-
lable due to the lethal developmental phenotype, but mice
hemizygous for the HypoVEGF allele are healthy and viable.
BCG-infected HypoVEGF mice had a significant reduction in
the number of accumulated granulomas and granuloma leuko-
cytes (Figures 2B and 2C), yet no increase in bacterial burden in
infected tissue compared to littermate controls (wild type [WT]
tative flow cytometry plots of liver and lung granuloma leukocytes from WT or
d antibodies against CD11b (left two columns) and Ly6C (right column).
on of Ly6Chigh and Ly6Clow cells among the CD11b+ population with VEGF-A
st, *p < 0.05, error bars mean ± SEM.
d by FDG positivity) among CD11b+ and CD11b populations in liver and lung
rror bars, mean ± SEM.
Cell Reports 27, 2119–2131, May 14, 2019 2121
Figure 2. Blockade of VEGF-A Reduces Granulomatous Inflammation without Compromising Mycobacterial Containment
Mice with either transgenic or drug-induced inhibition of VEGF-A activity were i.p. infected with BCG and harvested 21 days later to quantify the number of
granulomas and, granuloma leukocytes, as well as bacterial burden, in the liver. n = 4–6 mice per group. Representative H&E stainings showing granuloma-
containing liver from selected VEGF-A-inhibited groups shown in Figure S1.
(A) Representative area from DAPI-stained liver section of WT VEGF-A and HypoVEGF mice. BCG-containing granulomas outlined with white dotted lines.
(B) Number of liver granulomas per 403 field. Granulomas were counted among 3–5 micrographs at 403magnification for each mouse (each section containing
200–400 granulomas). 2,000–5,000 granulomas were counted for each experimental and control group. HypoVEGF mice have hypomorphic VEGF-A allele (see
Figure S2A). SU5416 is a small molecule RTK inhibitor of VEGFR1 and VEGFR2. VEGF-A Trap is a fusion protein made of a human IgG Fc linked to the VEGFR1
and VEGF R2 binding domains. Vegfa fl+/+ LysMcre+ mice have monocyte-specific inactivation of VEGF-A protein. Data normalized to control groups. Two-tailed
unpaired t test; *p < 0.05; **p < 0.01; ***p < 0.001; error bars, mean ± SEM.
(C) Number of granuloma leukocytes, isolated from the liver using a specialized density separation protocol, as described in methods. Light gray bars show all
leukocytes, dark gray bars show the fraction of granuloma leukocytes that are CD11b+ as measured by flow cytometry. Data normalized to control group.
Two-tailed unpaired t test; *p < 0.05; **p < 0.01; ***p < 0.001; error bars, mean ± SEM.
(D) Bacterial burden as measured by colony forming units (CFU) assay.
VEGF) (Figure 2D). We then asked if a similar phenotype re-
sulted after inhibition of the VEGF receptors. We treated
BCG-infected mice with SU5416, a small molecule receptor
tyrosine kinase (RTK) inhibitor of VEGFR1 and VEGFR2 (Sun
et al., 1998) that inhibits VEGFR-mediated tumor growth in
mice (Vajkoczy et al., 1999) and has been tested in a phase III
clinical trial against colorectal cancer. Similar to HypoVEGF
mice, VEGF receptor-inhibited mice had 35% fewer liver gran-
ulomas and 50% fewer granuloma leukocytes (Figures 2B, 2C,
and S1), but also no changes in bacterial burden (Figure 2D).
Since SU5416 is an RTK inhibitor with possible off-target
effects, we next inhibited VEGF-A protein using a targeted bio-
logic—a variant of VEGF Trap, which is a chimeric protein made
of the binding domains from VEGFR1 and VEGFR2 fused to a
human IgG Fc (Holash et al., 2002). Variant VEGF Trap-treated
mice had 25% fewer liver granulomas and 50% fewer liver leu-
kocytes (Figures 2B and 2C), but also no changes in bacterial
burden (Figure 2D).
After showing that VEGF-A regulates granulomatous inflamma-
tion, we asked about the relative importance ofmyeloid-produced
VEGF-A and derived a mouse strain with myeloid-specific inacti-
vation of VEGF-A protein. We crossed LysMcre mice (Cre recom-
binase expressed under the monocyte lysozyme M locus)
(Clausen et al., 1999), with a Vegfa floxed (fl+/+) strain that has crit-
ical exons of the Vegfa gene flanked with loxP sequences (Gerber
et al., 1999). Myeloid-restricted Cre expression leads to recombi-
nation at the loxP sites, which inactivates VEGF-A protein in these
cells. Granuloma leukocytes isolated from BCG-infected, Vegfa
fl+/+ LysMcre+ mice produced 3-fold less VEGF-A protein
ex vivo compared to Vegfa fl+/+ LysMcre littermate controls (Fig-
ure S2A). BCG-infected Vegfa fl+/+ LysMcre+ mice had roughly
half as many liver granulomas and half as many leukocytes as
Vegfa fl+/+ LysMcre littermate controls (Figures 2B, 2C, and
S1), demonstrating the importance of myeloid-produced VEGF-A
in granulomatous inflammation. Similar to our other experiments,
these mice had no changes in bacterial burden (Figure 2D).
These in vivo results demonstrate that VEGF-A regulates gran-
ulomatous inflammation during mycobacterial infection. Neutro-
phils also express the LysMCre promoter, but they are only
present at a low frequency in the granuloma. The data also
show that granulomatous inflammation can be reduced during
mycobacterial infection by inhibiting VEGF-A or VEGF-A recep-
tors without increasing bacterial burden. Despite decreased
inflammation, the cellular composition and proportion of cell
types within the granuloma remained remarkably similar to con-
trol groups (Figure S2B), with only modest or no changes to the
proportion of CD11b+, CD4+, B220+, and Ly6Chigh cells. The acti-
vation of T cells, measured by LFA-1 expression and production
of interferon (IFN)-g, was also unchanged despite VEGF-A inhibi-
tion (Figures S3A and S3B). On the other hand, granuloma leuko-
cytes from SU5416-treated mice had altered expression of
several key cytokines including less CCL2, CCL4, CCL7,
CCL12, CCL22, CXCL6, XCL1, interleukin (IL)-1g, IL-2, IL-6,
and IL-10, while tumor necrosis factor alpha (TNF-a) and IL-1a
were somewhat elevated. The full list of factors that were
measured is included in Figure S4.
VEGF-A Regulates Granulomatous Inflammation through Monocyte Recruitment and Not through Vasculogenic Effects We next asked if vascular changes were responsible for the
decreasedgranulomatous inflammation observed duringVEGF-A
ability via VEGFR2, which could support monocyte infiltration
independently from its role as a chemokine via VEGFR1. We
also observed that BCG infection of the liver increased the density
of sinusoidal blood vessels in the organ, but this increase was ab-
lated when the mice were treated with SU5416 (Figures S5A and
S5B), which we previously linked to decreased granulomatous
inflammation. To isolate the possible effects of VEGFR2-medi-
ated angiogenesis, we treated BCG-infected mice with the
VEGFR2-blocking antibody DC101 (Bruns et al., 2002; Prewett
et al., 1999), which does not bind to VEGFR1 and therefore
does not block the chemotactic activity of VEGF-A. While
DC101 treatment reduced the sinusoidal blood vessel density
associated with infection (Figures 3A and 3B), there were no
changes in the number of liver granulomas (Figure 3C) or liver
granuloma leukocytes (Figure 3D), and no changes in the bacterial
burden (Figure 3E). These results show that while increased
VEGF-A supports angiogenesis during mycobacterial infection,
the increased vascular density is not required for granuloma for-
mation, possibly due to the existing and extensive vascular
network in these organs.
cells induce monocyte chemotaxis directly. The migration of
granuloma-forming cells in response to chemokine gradients
is critical to granuloma formation (Algood et al., 2003), and
VEGF-A is a chemokine for monocytes via VEGFR1 (Barleon
et al., 1996; Clauss et al., 1996). First, we used an in vitro
approach in which isolated lung leukocytes or splenic cells
(attractant cells) from BCG-infected mice were loaded into the
bottom compartment of a microporous membrane-containing
chamber (Figure 3F, upper panels). Then, purified blood leuko-
cytes from CX3CR1eGFP mice, (a monocyte reporter strain
(Jung et al., 2000)), were loaded on top of the membrane, and
CX3CR1+ cell migration across the barrier was measured 24 h
later (Figure 3F, lower panels). There was a 3-fold increase in
the proportion of blood CX3CR1+ cells that migrated across
the barrier when the bottom compartment was loaded with
lung leukocytes cells compared to splenic cells from the same
infected mouse, but the increase was blocked when CX3CR1+
blood leukocytes were pre-treated SU5416. When the bottom
compartment was loaded with granuloma cells from Vegfa fl+/+
LysMcre+ mice, migration of CX3CR1 blood leukocytes was
also reduced.
previously been used to study the kinetics of granuloma cell
repopulation (Schreiber et al., 2011a). In this model, granu-
loma-containing liver from BCG-infected mice is transplanted
underneath the kidney capsule of recipient syngeneic mice (Har-
ding et al., 2011; Schreiber et al., 2011a, 2011b). First, we trans-
planted granulomatous liver from WT mice into uninfected GFP+
recipients (Figure 3Gi) and measured the influx of recipient GFP+
monocytes into transplanted granulomas up to 7 days later (Fig-
ure 3Gii). By that time, 40%of CD11b+ cells found in transplanted
granulomaswere from the recipient (GFP+), showing the dynamic
cell turnover in mycobacterial granulomas. We next measured
recruitment of monocytes into transplanted granulomatous tis-
sue when the donor, but not recipient, had decreased VEGF-A
activity. For this, we transplanted BCG-containing granulomas
from HypoVEGF or non-transgenic WT VEGF littermate controls
into CX3CR1-eGFP recipients (Figure 3Hi) and measured the
influx of CX3CR1+ cells into transplanted tissue 7 days later (Fig-
ure 3Hii). CX3CR1+ cell immigration was measured as the pro-
portion of transplanted tissue area (in pixels) containing eGFP+
signal. Approximately 30% fewer CX3CR1+ cells…