reSeArcH Article
markers (Fig. 4h). Moreover, the enrichment of EPCAM+SCA1+ cells
in the mCherry+ niche of different metastatic cell types was confirmed
by FACS analysis (Fig. 4i and Extended Data Fig. 9a–c). Similarly, the
presence of epithelial cells expressing another lung progenitor marker,
integrin β4 (also known as CD104)^29 , was increased in the mCherry+-
niche and in ex vivo co-cultures (Extended Data Fig. 9d–i).
In summary, we describe a parenchymal response to lung metastasis
involving de-differentiated pools of epithelial cells in the niche, which
we define as cancer-associated parenchymal cells (CAPs).
CAPs are activated AT2 cells
To functionally characterize CAPs, we tested their lineage differentiation
potential ex vivo using a 3D Matrigel-based organoid co-culture system^27
(Fig. 5a). Unlabelled resident lung EPCAM+ cells are predominantly
alveolar^27 , and formed mainly alveolar organoids when co-cultured
with CD31+ cells (Fig. 5b–d). mCherry+-niche EPCAM+ cells favoured
the bronchiolar lineage and showed a remarkable capacity to generate
multi-lineage bronchioalveolar organoids (Fig. 5d). Despite the bias in
organoid formation towards the bronchial lineage, we did not detect
mCherry-labelled cells expressing bronchial markers in vivo (Extended
Data Fig. 10a). CAPs also retained high self-renewal capacity over
multiple passages (Fig. 5e).
Next, we tested whether tumour cells could directly induce the
CAP phenotype. When EPCAM+ cells from unlabelled distal micro-
metastatic lungs or naive lungs were co-cultured with metastatic cells,
they generated a higher proportion of bronchiolar and bronchioalveolar
organoids (Fig. 5f–h and Extended Data Fig. 10b, c). Similar altera-
tions were induced by cancer cells when the assay was performed using
mouse lung fibroblasts (MLg cells) instead of CD31+ cells (Extended
Data Fig. 10b, c).
Although lung EPCAM+ cells are predominantly alveolar, they
also contain epithelial progenitors that could be enriched by cancer
cells to generate increased plasticity^27 ,^30. Therefore, we performed
organoid cultures using lineage-labelled AT2 (Sftpc-lineage) cells.
Sftpc-lineage cells, which show no plasticity in co-culture with
CD31+ cells, generated multi-lineage bronchioalveolar organoids
when exposed to cancer cells, supporting the idea of a reprogram-
ming activity driven by cancer-cell-derived factors ex vivo (Fig. 5i, j).
Despite the potential of cancer cells to modulate the organoid forma-
tion ability of lineage-labelled club cells (Scgb1a1 lineage), only rare
single Scgb1a1-lineage cells were found in proximity to lung metas-
tases (Extended Data Fig. 10d–f). Conversely, metastases growing
in Sftpc-lineage lungs demonstrated the alveolar (AT2) origin of the
CAPs (Fig. 5k).
Recently, a rare population of AT2 cells expressing Axin2 with stem
cell and repair activity (AT2 stem cells), was described in the lung
alveoli^31. Whereas a small proportion of Axin2-expressing cells was
found in the unlabelled epithelial cluster, Axin2 was undetectable in
the mCherry+-niche EPCAM clusters (data not shown). Therefore,
even if cancer cell seeding could trigger lung injury, this phenomenon200-20–40 –20 0 20t-SNE 2t-SNE 1Cdh1–
Cdh1+Epcam+Unlabelled lungEpcam+Stromal
cellsCdh1+
Cdh1–mCherry nicheEpcam+
Cdh1+
Cdh1–Cdh1+ Cdh1–Gene signature niche Epcam+ cellsAlveolarProgenitor markerSelected genes from
Epcam+ cell signatureRT–qPCR
unlabelled Epcam+ cellsCancer cells +
EPCAM+ cellsCancer cellsabcdefghiCancer cells
Cancer + EPCAM+ cells0.00.51.01.52.0P = 0.044P = 0.045P = 0.036Day 2Day 4Day 6Normalized integrated densitySftpbSftpcAbca3PdpnAgerAqp5Krt6VimSnailTwistCdh2Cdh1–30–15–9–6–3036(^159)
30
45
P = 0.0005 P = 0.0002 P = 0.0026 P = 0.0490 P = 0.0012 P = 0.0005 P = 0.0028
Fold change EPCAM niche vs lung
0
10
20
30
40
50
Ki67
- cells (% of EPCAM
+)
P = 0.0004 GFP+
cancer cells
EPCAM+
lung cells
20
10
0
–20
–30
–10
–20–30 –10 20100 30
t-SNE2
t-SNE 1
Cdh1–Cdh1+Cdh1–Cdh1+
Tm4sf1
Ly6a Ly6a
Nkx2-1Nkx2-1 Nkx2-1
Sftpb Sftpb Sftpb
Sftpc Sftpc Sftpc
Lcn2 Lcn2 Lcn2
Mid1ip1Mid1ip1 Mid1ip1
Fabp5 Fabp5 Fabp5
Cd36 Cd36 Cd36
Slc34a2 Slc34a2 Slc34a2
Cxcl15Cxcl15 Cxcl15
Lyz1 Lyz1 Lyz1
Sftpa1Sftpa1 Sftpa1
Glrx
Lamp3Lamp3 Lamp3
Etv5 Etv5 Etv5
Scd1 Scd1 Scd1
S100gS100g
Ppp1r14cPpp1r14c
Fasn Fasn
Egfl6*
Epcam+ cells Epcam+ cells
Lung Niche
0
10
20
30
40
50
EPCAM
+Sca1 - cells (% of Lin
- ) P = 0.0007
mCherrymCherry+mCherry–mCherry+–20–6–4246Fig. 4 | Lung epithelial cells in the metastatic niche display a progenitor
phenotype. a, Ki67 staining in FACS-sorted mCherry− and mCherry+
EPCAM+ cells (n = 7 independent sorts). b–d, GFP+ MMTV–PyMT
cancer cell growth in 3D co-culture with MACS-sorted EPCAM+ cells.
b, The co-culture scheme. c, Representative images from 4 independent
experiments (day 6; scale bar, 400 μm). d, Quantification of GFP
signal (n = 4, each with 3 technical replicates, statistical analysis of
biological replicates). Data are normalized to cancer cell growth.
e–g, scRNA-seq analysis; t-SNE plots of CD45− cells from the mCherry+
niche (e; n = 1,473) or distal lung (f; n = 1,996). g, Right, heat map of
mCherry+-niche EPCAM+ cells (ordered genes in rows and hierarchically
clustered cells in columns); left, table shows established lineage markers
(bold); asterisks indicate putative alveolar markers^25. h, RT–qPCR analysis
of EPCAM+ FACS-sorted cells (Sftpc, Aqp5, n = 9; Sftpb, Abca3, Pdpn,
Ager, Vim, Cdh1, n = 8; Krt6, Cdh2, n = 7; Snai1, n = 4; Twist, n = 3).
Data represented as fold change relative to mCherry− lung EPCAM+ cells
(statistical analysis on the ΔCt values). i, EPCAM+SCA1+ cell frequency
among Lin− (CD45−CD31−Ter119−) cells, determined by FACS (n = 13
mice). Statistical analysis by paired two-tailed t-test (a, h, i), one-sample
two-tailed t-test (d). Data represented as mean ± s.e.m.606 | NAtUre | VOl 572 | 29 AUGUSt 2019