690 | Nature | Vol 577 | 30 January 2020
Article
Glioblastoma is devoid of lymphangiogenic signals
Previous studies have shown that tumour-intrinsic overexpression of
VEGF-C in mice or humans results in poorer prognosis for malignancies
that occur in areas outside of the CNS^8 ,^9. We compared the transcriptomes
of brain tissue from healthy control individuals and patients with glioblas-
toma using data from the Genotype-Tissue Expression project (GTEX) and
The Cancer Genome Atlas (TCGA), respectively. Tumour tissue showed
higher expression of VEGFA and CD31 (also known as PECAM1) than control
tissue, as previously shown^10 , but a decrease in the expression of VEGFC
(Fig. 2a–c, Extended Data Fig. 2a, b)—which was also seen in mice that
bear glioblastoma tumours (Extended Data Fig. 2c). When we stratified
the patients with glioblastoma into two groups on the basis of their VEGFC
expression levels (high versus low), we observed no differences in survival
(Extended Data Fig. 2d, e), as the expression levels of VEGFC in patients
with glioblastoma were lower than those in healthy brain tissue even in
the group with high VEGFC expression. However, in a recently published
dataset^11 of patients with glioblastoma who were treated with neoadjuvant
anti-PD-1, the expression of VEGFC was highly correlated with an increase
in the infiltration of T cells (as measured by the expression of the T cell
marker genes CD3E, CD4 and CD8B) after treatment (Fig. 2d–f). Collec-
tively, these results suggest that the glioblastoma microenvironment is
deprived of lymphangiogenic signals both in patients with glioblastoma
and in experimental mice, and that ectopic expression of VEGF-C confers
protection against glioblastoma in a mouse model.
VEGF-C potentiates immune checkpoint blockade
Although treating mice with AAV-VEGF-C led to survival benefits with
no long-term side effects (Extended Data Fig. 1d, e, g), this vector
is—owing to the immunogenicity of AAV^12 —less effective when given
subsequently to the same host. By contrast, an mRNA delivery vector
does not elicit an immune response against the vector^13. Therefore, we
designed an mRNA construct to express VEGFC, with stabilizing base
substitutions (Extended Data Fig. 3a, b). Transfection of HEK293T cells
with the VEGFC mRNA construct produced both the full-length and the
processed form of VEGF-C^14 (Extended Data Fig. 3c). Administration of
the VEGFC mRNA construct into the cisterna magna of mice resulted
in high levels of VEGF-C in the CSF (Extended Data Fig. 3d). We used
Cy5-labelled mRNA to confirm the wide distribution of VEGFC mRNA
within the dura mater (Extended Data Fig. 3f ), and saw its uptake by
immune cells (CD45+), endothelial cells (CD45−CD31+) and other cells
(CD45−CD31−) in the brain, meninges and deep cervical lymph nodes
(Extended Data Fig. 3g, h). Furthermore, we observed an increase in
the levels of secreted VEGF-C protein that was specific to the CSF and
meninges, with no changes in the brain or serum (Extended Data Fig. 3i).
Using AKT phosphorylation to measure vascular endothelial growth
factor receptor (VEGFR) signalling in endothelial cells after treatment
with VEGF-C, we observed an increase in VEGFR signalling that was
specific to the lymphatic endothelial cell (LEC) population; by contrast,
no changes were observed in the blood endothelial cells (BECs) in the
meninges or deep cervical lymph nodes (Extended Data Fig. 4a–c).
This increase in VEGFR signalling was not accompanied by structural
deformities in either the LECs or the BECs, even within an angiogenic
tumour environment (Extended Data Fig. 4b–e). These data indicate
that whereas VEGFC mRNA is taken up by various cells of the brain,
meninges and deep cervical lymph nodes, VEGF-C protein is mostly
confined to the CSF and meninges, and activates LECs.
In addition to the difference in immunogenicity, the expression kinet-
ics of the VEGFC mRNA construct and the AAV-VEGF-C vector differ; thea
Lyve1AAV-CTRL AAV-VEGF-CAAV-CTRL AAV-VEGF-C0123Relative areaCon uence of sinuses
***020406080100DaysSurvival (%)5,000 GL261-Luc cellsAAV-VEGF-C
AAV-CTRL
Naive***Con uence of sinusescb0 20 80 100 120Mandibular/super cial
lymph nodecervicalcervical Deep
lymph node–20 0 20 80 1202040LN ligation (–7 d)60
***de80100
**DaysAAV-VEGF-C
AAV-VEGF-C
LN ligationAAV-CTRL
AAV-CTRL
LN ligation020406080100Days5,000 GL261 cells
****GL261-Luc0510 15AAV-VEGF-C
AAV-VEGF-C +
anti-CD8AAV-VEGF-C +
anti-CD4AAV-CTRL050100150DaysTu mour volume (mm3 )f GL261-LucNaive miceMice that survived
>100 d40 60Survival (%)40 60 100Survival (%)
NS NSNS0 20 40 60 80 100120Fig. 1 | VEGF-C-mediated protection against glioblastoma depends on the
draining lymph nodes and on T cells. a, b, C57BL/6 mice were injected with
A AV-CTRL or A AV-VEGF-C intracisternally through the cisterna magna. Six to
eight weeks later, mice were euthanized and the dura was collected to image
the lymphatic vasculature (LYVE1+) in the conf luence of sinuses (A AV-CTRL,
n = 7; A AV-VEGF-C, n = 8). P = 0.0007. c, C57BL/6 mice that had been injected
with A AV-CTRL or A AV-VEGF-C two months previously were implanted with
5,000 GL261-Luc cells in the striatum and monitored for survival (naive, n = 3;
A AV- C T R L , n = 4; A AV-VEGF-C, n = 8). P = 0.0004 for naive versus A AV-
VEGF-C; P = 0.33 (NS, not significant) for naive versus A AV-CTRL. d, Six-to-eight
weeks after injection of A AV vectors, the deep cervical lymph nodes of mice
were ligated using a cauterizer. Seven days later, mice were challenged with
50,000 GL261-Luc cells in the striatum and monitored for survival (A AV-CTRL,
n = 4; A AV-CTRL with lymph node (LN) ligation, n = 4; A AV-VEGF-C, n = 4;
A AV-VEGF-C with lymph node ligation, n = 10). P = 0.007 for A AV-CTRL
versus A AV-VEGF-C; ***P < 0.0001 for A AV-CTRL versus A AV-VEGF-C LN ligation;
P = 0.15 (NS) for A AV-CTRL versus A AV-CTRL LN ligation. e, Similar to c, but
mice that had previously been injected with A AV-VEGF-C were treated with
anti-CD4 or anti-CD8 antibodies (AAV-CTRL, n = 6; A AV-VEGF-C, n = 5;
A AV-VEGF-C + anti-CD8, n = 5; A AV-VEGF-C + anti-CD4, n = 5) * *P = 0.0014 for
A AV-CTRL versus A AV-VEGFC; P = 0.34 (NS) for A AV-CTRL versus A AV-VEGF-C
with anti-CD8; **P = 0.0014 for A AV-CTRL versus A AV-VEGF-C with anti-CD4.
f, Mice injected with A AV-CTRL or A AV-VEGF-C that survived over 100 days
after challenge with 5,000 GL261-Luc cells were rechallenged with 500,000
GL261-Luc cells in the f lank. Tumours were measured after f lank rechallenge at
days 7 and 15 (n = 3 per group, all 3 mice from the GL261 100 d group had no
measurable tumours). Data are pooled from two independent experiments
(b–f) and are mean ± s.d. P values were calculated by two-tailed unpaired
Student’s t-test or two-sided log-rank Mantel–Cox test.