Figure 1:

(1)

Figure 1: growth kine0cs of established breast PDXs. The average dura0on of the 3 ﬁrst passage was calculated for each PDX and used as an es0mate of their growth rate. Time is presented in days. Each bar represents a PDX. Color of the bar indicates the molecular subtype to which the PDX was assigned. As shown in Figure 2 red=Bas-‐L, light blue=Lum-‐B, pink=Lum-‐C, orange=m-‐Apo.

0 50 100 150 200 250 300 350

Average'duration'of'3'fisrt'passages' '(days)'

(2)

Figure 2: Molecular classiﬁca0on of PDX and tumors of origin are highly concordant. Iden0cal assignment were found in 87 % of the cases with at least 3/5 iden0cal classiﬁca0ons for each PDX. Divergences were restricted to the luminal subtypes. Transcriptome PDX and

corresponding tumor of origin were classiﬁed using the Guedj, Hu, PAM 50, Sorlie and Neve breast cancer classiﬁers and assignments were indicated by a color code. Red: Basal or Bas-‐L, Orange: M-‐Apo (ER-‐/HER2+), Pink: Lum-‐C (ER+/PR-‐), light blue: Lum-‐B (ER+/PR+/HER2-‐), dark blue: Luminal A. ER, PR status determined by IHC is presented in red for ER-‐, in green for ER+. HER2 status was noted nega0ve in green when scored <2+, posi0ve in red score = 3+. TP53 muta0on status determined by DNA sequencing, mutated in red, wt in green.

TP53%status CIT Hu PAM50 Sorlie Neve ER%status PR%status HER2%status Mut.%

B2164 Primary basL Basal Basal Basal BasalA ER< PR< 0 no

B2164_P2_353 Xenograft basL Basal Basal Basal BasalA ER< PR< 0 no

B3467 Primary basL Basal Basal Basal BasalA ER< PR< 0 yes

B3467_P2_548 Xenograft basL Basal Basal Basal BasalA ER< PR< 0 yes

B4122 Primary basL Basal Basal Basal BasalA ER< PR< 0 no

B0444 Primary basL Basal Basal Basal BasalA ER< PR< 1+ yes

B0444_P2_916 Xenograft basL Basal Basal Basal BasalA ER< PR< 1+ yes

B1240_P2_1036 Xenograft basL HER2 Basal Basal BasalA ER< PR< 0 yes

B2412 Primary basL Basal Basal Basal BasalB ER< PR< 0 yes

B2412_P2_339 Xenograft basL Basal Basal LuminalB BasalB ER< PR< 0 yes

B3160 Primary basL Basal Normal Basal BasalA ER< PR< 0 no

B3804 Primary basL HER2 LuminalB HER2 BasalB ER< PR< 0 yes

B3804_P3_1177 Xenograft basL HER2 LuminalB Basal BasalB ER< PR< 0 yes

B3499 Primary mApo HER2 HER2 HER2 Luminal ER< PR< 1+ yes

B3499_P2_606 Xenograft mApo HER2 HER2 LuminalB Luminal ER< PR< 1+ yes

B3925 Primary mApo HER2 LuminalB LuminalB Luminal ER< PR< 3+ no

B3925_P2_1295 Xenograft mApo HER2 LuminalB LuminalB Luminal ER< PR< 3+ no

B3921 Primary lumC LuminalA LuminalB LuminalB Luminal ER+ PR< 3+ yes

B3921_P2_1369 Xenograft mApo HER2 LuminalB LuminalB Luminal ER+ PR< 3+ yes

B3828 Primary lumB LuminalA LuminalB LuminalB Luminal ER+ PR+ 0 yes

B3828_P2_777 Xenograft lumB LuminalA LuminalB LuminalB Luminal ER+ PR+ 0 yes

B0627 Primary lumB LuminalA LuminalB LuminalB Luminal ER+ PR+ 2+ no

B0627_P2_1033 Xenograft lumB LuminalA LuminalB LuminalB Luminal ER+ PR+ 2+ no

B2115 Primary lumC LuminalA LuminalB LuminalB Luminal ER+ PR< 3+ no

B2115_P2_703 Xenograft lumB LuminalB LuminalB LuminalB Luminal ER+ PR< 3+ no

B3034 Primary lumC LuminalA LuminalA LuminalB Luminal ER+ PR< 0 no

B3034_P2_814 Xenograft lumB LuminalA LuminalB LuminalB Luminal ER+ PR< 0 no

Molecular%subgroups% Steroid%receptor%status

(3)

Figure 3: Breast tumors of origin and corresponding PDXs show elevated similarity at the genomic level. CNC profiles of all PDXs (* indicates corrected profiles) and breast tumors of origin were determined by array-‐CGH and convergence tested by clustering analysis. Breast tumors and their cognate PDXs were systema0cally clustered together (pairs, triplet or quadriplet according to profile availability).

B3804 B3804-‐P3-‐1177* B3034B3034-‐P2-‐814* B3925-‐P2-‐1295* B3925 B3029 B3029-‐P3-‐1116* B3029-‐P3-‐1078* B2164 B2412-‐P2-‐337* B2412 B1240-‐P2-‐1036* B4122 B3160-‐P2-‐746*B2164-‐P3-‐430* B1240 B4122-‐P2-‐1048* B3467-‐P3-‐563*B0444-‐P2-‐916*B3160 B0444 B3467 B3977-‐P3-‐566* B3977 B0460-‐P3-‐1255* B0460-‐P3-‐1263* B0460 B0460-‐P1-‐1119* B0523 B1915 B1915-‐P2-‐887* B3828 B3828-‐P2-‐777* B2115 B2115-‐P2-‐703* B1995 B1995-‐P3-‐896* B3921 B3921-‐P1-‐1240* B3921-‐P0-‐924* B3921-‐P2-‐1369* B3499-‐P2-‐606* B3499 B0627 B0627-‐P2-‐1033*

(4)

Figure 4:

CNC proﬁles of diﬀerent graes from the same primary tumor remain remarkably stable . A: Grae tree indica0ng the posi0on of the analyzed PDXs which have been highlighted by a color code.

B: Histograms showing the frac0on of overlapping events in the tumor of origin and the corresponding PDX which are iden0ﬁed by color codes indicated in the grae tree. C: whole genome CNC proﬁles, gains are shown in blue, losses in red,

B0460 B3921

P0 P1 P2 P3 P4

P0 P1 P2 P3

B0460

B0460-‐P1 B0460-‐P3 B0460-‐P3

B3921

B3921-‐P0 B3921-‐P1 B3921-‐P2

A

B

C

(5)

A B

Figure 5 :

quan0fica0on of CNC specifically found in primary tumors (A) or specific of the PDX (B). Events specific to the tumors give an insight on tumor heteroclonality, whereas those found only in the PDX represent events acquired de novo during propaga0on in the mice. A: Each bar represents the percentage of the genome interrogated by the array involved in primary tumor specific CNC. B: the frac0on of the genome corresponding to CNC occurring de novo in the PDX. CGH profiles of the PDXs and cognate primary tumors were analyzed aeer correc0on for the contamina0on by normal stromal cells.

(6)

Figure 6:

Expression differences between primary tumors that give rise to PDX (Take) and those that did not (No Take). We analyzed 3 expression signatures with known prognos0c impact and two further signatures (VEGF, IL8). The tumor set was stra0fied according to molecular subgroups to verify whether these differences were not due to differences in composi0on of the Take and the No Take groups. GGI iden0fies the molecular grade signature (ref), Wound healing (ref) and

prolifera0on signature (ref).

Whole

tumor set Stra0ﬁed tumor set

prolifera0on

p=9 10^-‐10

GGI

p= 1.3 10^-‐13

VEGF

p= 5.1 10^-‐7

wound healing

p= 1.7 10^-‐14

IL8

p= 0.02

p= 0.05 p= 0.06

NS

p= 0.0004 NS

p= 0.05 NS

NS

p= 0.03 NS

Bas-‐L Lum

T

No T No T T No Take

Take