EURURO Vol. 72 No. 3

361 476

We then used the top 30 most prevalent mutations and
CNAs in TCGA’s whole exome sequencing dataset in
Firehose and cBioportal, respectively, and examined their
prevalence in the UNC basal-like and luminal subtypes
(Supplementary and 2). Included among them were
alterations that were enriched in the breast cancer intrinsic
subtypes (TP53, RB1 , ERBB2, and PIK3CA ), genes that
displayed different mutation frequencies in NMIBCs versus
MIBCs (FGFR3 , KDM6A, and STAG2), and genes that encode
for mRNAs that were enriched in basal or luminal MIBCs
(EGFR, PPARG, GATA3, ELF3, and ERBB3). Consistent with the
overall hypothesis, several of the alterations were signifi-
cantly enriched in either UNC basal-like or luminal cancers
.
We then investigated whether creating further subdivi-
sions of the UNC molecular subtypes caused additional
patterns of enrichment as had been documented previously
[59_TD$DIFF] . Although the mutations and CNAs that were enriched
in the UNC basal-like and luminal MIBCs were also enriched
in the MD Anderson basal and luminal MIBCs, isolating the
p53-like tumors did not further enhance enrichment
. Similarly, no mutations or CNAs were specifically
enriched in the Lund infiltrated tumors as compared with
the other Lund subtypes in this panel ). Therefore, it
appears that the biology of these infiltrated tumors is
dictated less by genetic influences than by other factors,
such as the tumor microenvironment, explaining why their
subtype membership was somewhat unstable [70_TD$DIFF] . On
the other hand, subdividing the UNC basal-like and luminal
tumors into the other Lund subtypes yielded additional and
highly informative patterns of mutation and CNA enrich-
ment. The uroA and uroB tumors were both highly enriched
with activating FGFR3 mutations [71_TD$DIFF] and the uroB
tumors also contained a higher number of CDKN2A (p16)
deletions ). The uroA and uroB tumors were also
characterized by fewer RB1 mutations, and the uroB tumors
could be distinguished from the uroA tumors by their
content of PIK3CA, NFE2L2, ERBB2 , and ERBB3 mutations
. Finally, the Lund subdivision of the UNC luminal
MIBCs into the GU and uroA subtypes yielded additional
informative patterns of mutation and CNA enrichment
[59_TD$DIFF] . The GU tumors could be distinguished from the uroA
tumors by the absence of activating FGFR3 mutations and by
the presence of TP53 and ERCC2 mutations, RB1 deletions,
[(Fig._3)TD$FIG]
ERCC2 NFE2L2
RB1 TP53
0
20
40
60
Basal
% of samples
Basal p53 like Luminal
*
* ELF3
ERBB2ERBB3FGFR3FOXA1GATA3KDM6APIK3CARXRASTAG2
0
10
20
30
40
Luminal
% of samples
Basal p53 like Luminal
* * *
EGFR ERCC2
RB1_DEL TP53_DEL
0
5
10
15
Basal
% of samples
Basal p53 like Luminal CDKN2A_DEL
E2F3
ERBB2 ERBB3 FGFR3 GATA3 PPARG
SOX4
0
10
20
30
40
Luminal
% of samples
Basal p53 like Luminal
* * * *
Mutation (MDA)
CNA (MDA)
Fig. 3 – Enrichment of significantly mutated genes and CNAs in the MD Anderson subtypes. Alterations are grouped according to predicted enrichment
in basal versus luminal tumors, and the results are displayed as percentages of tumors in each subtype that contained the indicated alteration. The
CNAs correspond to chromosomal amplification unless specifically identified as deletions (‘‘del’’). Fisher’s exact test was used to determine differences
between subtypes. CNA = copy number aberration; MDA = MD Anderson. * p < 0.05 was considered significant.
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 3 5 4 – 3 6 5 361