Gene expression fingerprinting for human hereditary hemorrhagic telangiectasia

Abstract

Hereditary hemorrhagic telangiectasia (HHT) or Osler–Weber–Rendu syndrome is an autosomal dominant vascular disorder characterized by telangiectases and internal arteriovenous malformations. It is caused by mutations in elements of the transforming growth factor-β (TGF-β) receptor complex: endoglin , a co-receptor, responsible for HHT1, or ALK1 (activin receptor-like kinase 1), a type I receptor leading to HHT2. Recently, we have established cultures of HHT endothelial cells, primary targets of the disease. These cells showed deficient TGF-β signaling and angiogenesis, representing a useful human model to study the molecular mechanism of this disease. To understand the pathogenic mechanism underlying HHT, we have used total RNA probes to compare HHT versus non-HHT cells by expression microarrays. This work represents a systematic study to identify target genes affected in HHT cells. Given the similarity of symptoms in HHT1 and HHT2, special interest has been put on the identification of common targets for both HHT types. As a result, 277 downregulated and 63 upregulated genes were identified in HHT versus control cells. These genes are involved in biological processes relevant to the HHT pathology, such as angiogenesis, cytoskeleton, cell migration, proliferation and NO synthesis. The type of misregulated genes found in HHT endothelial cells lead us to propose a model of HHT pathogenesis, opening new perspectives to understand this disorder. Moreover, as the disease is originated by mutations in proteins of the TGF-β receptor complex, these results may be useful to find out targets of the TGF-β pathway in endothelium.

INTRODUCTION

Hereditary hemorrhagic telangiectasia (HHT) or Rendu–Osler–Weber syndrome is an autosomal dominant vascular disorder, characterized by recurrent nose bleeding, telangiectases in mucosa and arteriovenous malformations in lung, brain and liver ( 1 , 2 ). HHT patients have dilated blood vessels with thin walls and show anomalous arterio-venous shunts. There are two main types of HHT: type 1 is due to mutations in endoglin , an accessory endothelial transforming growth factor-β (TGF-β) receptor ( 3 ), and type 2 caused by mutations in ALK1 ( ACVRL1 ; activin receptor-like kinase 1) gene ( 4 ). ALK1 is a TGF-β type I receptor expressed in endothelial cells. ALK1 binds TGF-β ( 5 , 6 ), and possibly other ligands (BMPs, bone morphogenetic proteins). In both, HHT1 and HHT2, the TGF-β signaling pathway of endothelial cells is affected ( 7 ). Because of the predominant expression of endoglin and ALK1 in endothelial cells, these are primary cellular targets of the disease, although other targets such as smooth muscle and endothelial precursor cells may be also important initiating the lesion secondary to an endothelial cell defect ( 8 , 9 ).

TGF-β belongs to a superfamily of soluble factors, including activins and BMPs. It is involved in different biological processes such as cell proliferation, migration, differentiation, survival, cell-to-cell and cell-matrix interactions and in oncogenesis ( 10 ). It is essential for the recruitment of pericytes and smooth muscle cells which lead to vascular maturation and stabilization ( 11 ). TGF-β members bind to heteromeric complexes consisting of type I and type II transmembrane serine/threonine kinases. After binding, type II receptors recruit type I receptors, which in turn phosphorylate and activate the Smad family of downstream coactivators ( 12 ) that bind and stimulate different target genes. In cultured endothelial cells, two types of TGF-β type I receptors coexist, the ubiquitously expressed ALK5, which signals through Smad2/Smad3, and ALK1 transmitting the signal through Smad1/Smad5 ( 13 ). Interestingly, a differential non-overlapping expression of ALK1 and ALK5 may occur during vascular development in mice ( 9 ). According to this report, ALK1 would be endothelial-specific, while ALK5 would be expressed in the adjacent smooth muscle cells surrounding the endothelial layer. It has been recently shown that endoglin and ALK1 cooperate in endothelial cells to mediate Smad1/Smad5 TGF-β signaling ( 14 , 15 ). This functional cooperation ALK1/endoglin may explain why both types of HHT have similar clinical symptoms. On the other hand, knock out mice for endoglin and ALK1 are lethal due to vascular defects ( 6 , 8 ), while a percentage of heterozygous endoglin and ALK1 mice develop symptoms similar to HHT patients ( 8 ). The lethality of homozygous conditions of endoglin and ALK1 and the HHT clinical symptoms in humans due to mutations in heterozygous condition, point to essential roles of endoglin and ALK1 in angiogenesis, but the precise molecular mechanisms are largely unknown.

Recently, we have developed primary cultures of endothelial cells from HHT1 to HHT2 patients and shown that these cells have decreased TGF-β signaling, abnormal tube formation and anomalous F-actin cytoskeleton ( 16 ). To deepen into the molecular mechanisms disrupted in HHT, we have studied the differential gene expression of HHT endothelial cells using microarrays. To the best of our knowledge, there are no previous reports describing differentially expressed genes in HHT cells. The closest studies related to this field tried to unravel the role of ALK1 in endothelial cells by infection of primary human dermal microvascular and umbilical vein endothelial cells and the human microvascular endothelial cell line HMEC-1 with a constitutively active ALK1 adenovirus ( 17–19 ). However, these systems have the drawback of the artificial situation generated by the overexpression of ALK1 and the viral infection. Our microarray model using RNA from HHT endothelial cells derived from clinically diagnosed patients, is approaching closer the physiological situation of the HHT.

The result of the study provides the first list of target genes whose expression is significantly affected in HHT cells. In total, 277 genes were identified as downregulated and 63 upregulated in HHT versus control cells. Those genes are involved in biological processes relevant to the HHT pathology: angiogenesis, cell adhesion, migration, transmigration, cell cycle, cytoskeleton organization and vascular physiology. The validation of some of the differentially expressed genes and the analysis of their functions has been assessed. At the same time, as the HHT is caused by mutations in components of the TGF-β signaling pathways, this study provide also a clue for the knowledge of downstream genes affected by the TGF-β cascade in the endothelial cell.

RESULTS

Endothelial cells from HHT patients unravel a differential pattern of expression compared to healthy endothelial cells

We have made an attempt to unravel a common gene signature affected in HHT cells by comparing overall expression patterns of HHT endothelial cell cultures with healthy cells. For this purpose, four types of primary endothelial cell cultures, previously well characterized ( 16 ), were used as RNA probes in a whole genome expression microarray hybridization: healthy donor (Controls), HHT1 (family no. 1), HHT2 nonsense mutant (HHT2n, family no. 10) and HHT2 missense mutant (HHT2m, family no. 2).

After statistical analysis, the genes that were differentially expressed in at least one of the HHT types (HHT1, HHT2n or HHT2m) versus control are shown in the Supplementary Material, Table S1 (on line). Only those genes significantly upregulated or downregulated, with P < 0.05 in the three HHT sets analysed, were chosen. Tables  1 and 2 show the annotated genes identified as downregulated (277 genes, Table  1 ) or upregulated (63 genes, Table  2 ), by the comparison of healthy blood outgrowth endothelial cells (BOECs) with respect to HHT BOECs (irrespective of their type 1 or type 2 origin).

Table 1.

Downregulated genes in HHT versus control BOECs*

Gene	Access GenBank	log 2 (HHT1/C)	log 2 (HHT2n/C)	log 2 (HHT2m/C)	ID
ABLIM1	NM_002313	−2.51	−3.22	−3.67	3983
ACVRL1	AK056725	−2.45	−2.96	−4.77	94
AF15Q14	AA662240	−2.27	−0.92	−1.76	57082
AHNAK	M80899	−2.67	−1.67	−2.04	195
AKR1C3	NM_003739	−2.46	−2.87	−1.21	8644
ALPK3	AL833276	−2.48	−2.75	−4.67	57538
AMD1	NM_001634	−4.23	−3.88	−1.58	262
ANGPT2	NM_001147	−1.42	−2.05	−7.49	285
ARHGAP10	NM_024605	−3.99	−2.92	−1.61	79658
ARHGEF15	NM_173728	−1.76	−2.37	−2.98	22899
ASB9	NM_024087	−3.51	−2.17	−2.92	140462
ASK	NM_006716	−2.57	−2.51	−1.56	10926
B4GALT5	NM_004776	−1.45	−0.54	−2.35	9334
BAT3	NM_004639	−1.67	−2.22	−1.18	7917
BNIP1	NM_001205	−3.23	−2.68	−2.03	662
BNIP2	AK125533	−3.36	−3.50	−2.33	663
BTBD6	NM_033271	−3.15	−2.56	−1.38	90135
C10orf3	NM_018131	−2.55	−1.98	−2.01	55165
C14orf109	BU739864	−2.47	−2.42	−1.99	26175
C21orf63	NM_058187	−2.57	−1.68	−2.27	59271
C6orf11	NM_005452	−3.01	−2.19	−1.87	9277
C9orf112	NM_138778	−3.85	−3.54	−2.30	92715
C9orf58	NM_031426	−1.62	−1.41	−3.71	83543
C9orf59	NM_033387	−1.24	−1.54	−3.02	286336
CALCRL	NM_005795	−1.99	−3.60	−6.25	10203
CAV2	R01546	−3.42	−2.80	−2.03	858
CBX2	BG354579	−0.93	−1.17	−2.10	876
CCNA2	NM_001237	−1.61	−1.55	−1.78	890
CCNB2	NM_004701	−1.63	−0.53	−1.81	9133
CDC25B	NM_212530	−1.47	−0.64	−1.25	994
CDYL	NM_004824	−1.32	−1.61	−1.89	9425
CDYL2	NM_152342	−0.96	−1.41	−2.42	124359
CECR6	NM_031890	−1.15	−1.54	−2.43	27439
CENPF	NM_016343	−2.30	−1.48	−1.99	1063
CFLAR	NM_003879	−2.40	−1.56	−2.45	8837
CGI-119	NM_016056	−3.09	−3.59	−1.74	51643
CHAF1A	NM_005483	−3.61	−2.91	−2.00	10036
CHCHD3	NM_017812	−1.84	−1.92	−2.02	54927
CHEK1	NM_001274	−3.37	−2.75	−1.82	1111
ChGn	NM_018371	−1.15	−1.97	−3.10	55790
CHMP1.5	NM_020412	−3.71	−3.40	−2.30	57132
CIDEB	NM_014430	−2.92	−2.93	−2.69	27141
CKS1B	NM_001826	−2.79	−1.57	−1.35	1163
COPS3	NM_003653	−4.25	−3.73	−1.70	8533
CPSF1	NM_013291	−1.75	−2.85	−2.22	29894
CPSF3	NM_016207	−2.91	−2.60	−2.31	51692
CSE1L	NM_001316	−1.45	−2.58	−1.38	1434
CSRP1	NM_004078	−3.75	−3.77	−2.69	1465
CYYR1	NM_052954	−1.82	−1.32	−5.26	116159
DEK	NM_003472	−2.84	−2.16	−2.02	7913
DKFZP547E1010	R43702	−3.00	−3.02	−1.56	26097
DKFZp762E1312	NM_018410	−1.81	−0.74	−1.81	55355
DLG7	NM_014750	−1.51	−1.01	−2.63	9787
DNAJA1	NM_001539	−1.82	−2.18	−1.45	3301
DNASE1L3	NM_004944	−3.39	−4.97	−4.98	1776
DRB1	NM_152945	−3.44	−3.79	−1.93	129831
DSTN	NM_006870	−2.47	−2.44	−0.96	11034
DTYMK	NM_012145	−3.08	−2.12	−1.21	1841
DVL3	NM_004423	−3.51	−3.70	−1.38	1857
DYSF	NM_003494	−3.13	−3.54	−6.37	8291
EDG1	NM_001400	−2.36	−3.06	−4.49	1901
EIF2S2	NM_003908	−3.86	−3.27	−1.57	8894
EIF3S5	NM_003754	−1.91	−2.24	−1.16	8665
EIF4A1	NM_001416	−5.77	−4.91	−2.07	1973
EIF4A2	R89137	−2.34	−2.67	−2.19	1974
ERCC3	NM_000122	−3.82	−3.63	−1.92	2071
ESAM	NM_138961	−0.73	−0.79	−7.47	90952
ETFA	NM_000126	−3.11	−3.26	−1.53	2108
F11R	NM_016946	−2.76	−3.33	−3.39	50848
FABP4	NM_001442	−1.07	−1.24	−7.29	2167
FABP5	NM_001444	−3.42	−3.59	−2.89	2171
FAHD1	NM_031208	−3.96	−3.90	−2.52	81889
FEN1	NM_004111	−4.16	−3.30	−1.92	2237
FGD5	NM_152536	−2.71	−3.45	−5.64	152273
FLJ10213	NM_018029	−2.77	−3.31	−2.12	55096
FLJ10853	NM_018246	−3.49	−3.57	−1.29	55246
FLJ11724	AK021786	−1.82	−1.61	−3.41	92340
FLJ11795	NM_024669	−1.09	−2.16	−4.37	79722
FLJ12604	NM_024621	−2.19	−1.81	−2.43	79674
FLJ12649	BC040518	−2.31	−1.63	−2.24	79649
FLJ13912	NM_022770	−1.77	−0.87	−2.80	64785
FLJ20641	NM_017915	−3.34	−0.99	−1.36	55010
FLJ21103	NM_024556	−2.67	−3.38	−1.58	79607
FLJ21438	AK024488	−0.92	−5.20	−4.54	64926
FLJ21616	NM_024567	−2.05	−1.64	−2.75	79618
FLJ22635	NM_025092	−0.96	−1.81	−1.67	80162
FLJ22794	BC071759	−2.92	−1.60	−2.15	63901
FLJ36031	AK098422	−0.78	−1.02	−1.35	168455
FLJ40504	NM_173624	−2.30	−2.03	−4.83	284085
FN5	NM_020179	−2.48	−2.96	−1.93	56935
FPGS	NM_004957	−2.52	−2.59	−1.64	2356
FTS	NM_022476	−1.54	−1.65	−1.88	64400
FYN	NM_153047	−4.24	−4.33	−1.46	2534
GABARAPL1	NM_031412	−1.31	−2.49	−2.37	23710
GABPB2	NM_005254	−2.75	−3.07	−2.81	2553
GAL	NM_015973	−3.18	−1.36	−2.17	51083
GALNT1	U41514	−2.43	−1.96	−4.17	2589
GALNT6	AA419568	−0.96	−0.70	−2.03	11226
GBP3	NM_018284	−0.95	−3.15	−3.25	2635
GDF3	NM_020634	−4.00	−3.60	−3.66	9573
GDI2	NM_001494	−2.81	−3.17	−1.57	2665
GIMAP1	NM_130759	−0.77	−0.93	−6.77	170575
GIMAP4	NM_018326	−1.27	−1.81	−8.52	55303
GIMAP5	NM_018384	−1.42	−1.69	−2.87	55340
GIMAP7	NM_153236	−0.75	−0.80	−5.59	168537
GLCE	BQ889128	−5.62	−4.55	−3.87	26035
GNAS	NM_016592	−2.24	−2.28	−1.23	2778
GOT2	NM_002080	−0.86	−1.21	−1.79	2806
GPS1	NM_212492	−1.99	−2.90	−1.91	2873
GTF2I	NM_032999	−3.28	−3.69	−2.23	2969
HAN11	NM_005828	−1.21	−1.65	−1.24	10238
HAT1	NM_003642	−3.24	−2.16	−1.70	8520
HCAP-G	AF331796	−1.80	−0.64	−2.35	64151
HMG20B	NM_006339	−2.35	−3.01	−1.70	10362
HMGB1	NM_002128	−2.59	−2.01	−1.72	3146
HMGB2	NM_002129	−1.24	−0.53	−1.48	3148
HMMR	NM_012484	−1.74	−1.20	−1.41	3161
HNRPA2B1	NM_002137	−1.61	−1.36	−1.02	3181
HNRPK	NM_002140	−1.61	−2.02	−1.40	3190
HSPA1A	NM_005345	−0.83	−1.06	−1.80	3303
HSPA9B	NM_004134	−2.68	−3.09	−1.22	3313
HSPC132	NM_016399	−2.10	−1.09	−1.58	51499
HSPC152	NM_016404	−3.07	−2.32	−1.11	51504
HSPCB	NM_007355	−0.75	−0.78	−1.40	3326
HSPG2	NM_005529	−1.36	−2.47	−2.35	3339
HT036	NM_031207	−2.61	−2.48	−3.41	81888
ILF2	NM_004515	−0.81	−1.66	−1.25	3608
ITGA6	NM_000210	−2.03	−2.05	−3.95	3655
ITGB3BP	NM_014288	−2.46	−1.57	−2.01	23421
KIAA0102	NM_014752	−2.86	−3.19	−1.23	9789
KIAA0367	AL119769	−2.91	−3.00	−3.87	23273
KIAA0540	BU429796	−2.94	−2.93	−3.29	23218
KIAA0804	BX647915	−1.48	−1.53	−1.79	23355
KIAA1272	AY007156	−0.53	−1.08	−2.92	57186
KIF11	NM_004523	−1.96	−1.38	−2.18	3832
KIF2C	NM_006845	−4.03	−3.48	−2.39	11004
KPNA2	NM_002266	−0.79	−0.71	−1.81	3838
KRTCAP2	NM_173852	−2.99	−3.16	−1.10	200185
KTN1	NM_182926	−0.77	−0.57	−1.52	3895
LDHB	NM_002300	−3.19	−2.93	−1.70	3945
LOC130617	NM_138802	−3.34	−3.42	−2.07	130617
LOC133619	NM_130809	−2.34	−3.09	−1.26	133619
LOC348645	AI628657	−0.94	−1.46	−5.39	348645
LOC91120	NM_033196	−1.13	−1.42	−3.18	91120
LOC92162	NM_203411	−3.00	−1.85	−6.48	92162
LRIG3	NM_153377	−4.07	−4.01	−3.44	121227
LRPAP1	NM_002337	−1.81	−2.45	−1.34	4043
LRRC28	NM_144598	−1.70	−1.29	−1.34	123355
LTA4H	NM_000895	−0.59	−1.75	−1.58	4048
MAT2B	NM_182796	−2.59	−2.75	−3.00	27430
MBNL1	AW296451	−2.11	−1.72	−2.83	4154
MCTP1	AK058012	−1.39	−1.76	−3.65	79772
MED28	BM711190	−1.80	−2.42	−2.09	80306
MEF2C	NM_002397	−2.03	−2.14	−3.33	4208
MELK	NM_014791	−2.62	−2.08	−2.15	9833
METAP1	NM_015143	−3.85	−3.64	−2.32	23173
MGC13204	NM_031465	−2.45	−3.28	−2.57	83695
MGC2714	NM_032299	−2.53	−2.63	−1.40	84259
MGC3265	NM_024028	−1.78	−2.45	−1.86	78991
MGC3794	NM_152902	−3.05	−3.00	−1.42	261726
MGC49942	NM_174893	−3.08	−2.98	−1.36	124944
MGC51082	NM_182498	−3.91	−2.80	−2.80	126299
MGC57827	NM_207418	−1.59	−0.77	−1.49	389835
MIS12	NM_024039	−3.31	−3.32	−1.87	79003
MLZE	NM_031415	−3.44	−3.85	−3.36	56169
MMRN1	NM_007351	−0.54	−1.32	−4.12	22915
MRPL13	NM_014078	−3.05	−3.08	−1.14	28998
MRPL35	NM_145644	−1.52	−1.02	−1.69	51318
MRPS7	NM_015971	−3.33	−3.26	−1.64	51081
MTA1	NM_004689	−3.58	−3.20	−1.51	9112
MTB	T99312	−6.21	−4.07	−2.34	54892
MTUS1	NM_001001931	−0.49	−1.06	−2.90	57509
MTUS1	CA429135	−2.04	−2.51	−4.46	57509
MX1	NM_002462	−2.09	−3.04	−2.79	4599
MYCT1	NM_025107	−1.15	−1.53	−3.03	80177
NAGPA	NM_016256	−3.04	−3.68	−1.91	51172
NDUFV2	BI908152	−3.46	−2.12	−2.41	4729
NEU1	NM_000434	−2.76	−2.12	−2.69	4758
NFATC2	H80534	−1.79	−2.44	−3.47	4773
NFIB	NM_005596	−0.60	−1.21	−2.96	4781
NK4	NM_004221	−2.74	−3.67	−3.30	9235
NOP5/NOP58	NM_015934	−3.56	−2.58	−1.32	51602
NOS3	NM_000603	−1.63	−0.99	−3.18	4846
NT5C3	NM_016489	−3.98	−3.00	−2.07	51251
OIP5	NM_007280	−4.93	−4.43	−3.88	11339
OXA1L	AI829834	−3.43	−3.16	−2.00	5018
OXCT1	NM_000436	−3.80	−3.51	−1.84	5019
PAI-RBP1	NM_015640	−1.43	−0.92	−1.45	26135
PAPLN	NM_173462	−1.04	−1.52	−2.80	89932
PBEF1	AK023341	−2.02	−1.79	−2.27	10135
PCCB	NM_000532	−3.57	−2.73	−1.23	5096
PCDH12	NM_016580	−1.36	−2.36	−3.94	51294
PCNP	NM_020357	−1.18	−1.96	−1.87	57092
PDCD6IP	W77963	−3.54	−4.02	−1.85	10015
PDE2A	NM_002599	−0.78	−0.99	−5.64	5138
PDF	NM_022341	−3.97	−3.45	−2.06	64146
PECAM1	NM_000442	−1.82	−2.35	−1.90	5175
PGBD1	NM_032507	−1.06	−2.05	−1.85	84547
PGM2	NM_018290	−3.48	−4.31	−3.06	55276
PIK3C2B	NM_002646	−0.95	−0.73	−5.37	5287
PIM3	NM_001001852	−0.67	−1.27	−2.26	415116
PITX1	NM_002653	−3.39	−4.31	−4.44	5307
PLAT	NM_000930	−1.19	−0.79	−1.31	5327
PLCB1	AW444937	−3.74	−2.67	−2.55	23236
PLCG2	NM_002661	−1.25	−2.26	−4.39	5336
PLSCR4	NM_020353	−2.01	−2.42	−3.69	57088
PMF1	NM_007221	−1.69	−1.79	−1.34	11243
PPGB	NM_000308	−1.97	−3.08	−1.48	5476
PPT1	NM_000310	−1.70	−1.95	−2.40	5538
PRKCH	NM_006255	−0.93	−1.46	−4.28	5583
PRPF8	NM_006445	−1.22	−2.19	−2.32	10594
PRPSAP2	NM_002767	−3.71	−4.16	−3.21	5636
PSIP1	NM_033222	−1.32	−1.70	−2.34	11168
PSMA4	NM_002789	−4.18	−3.85	−2.06	5685
PSMB9	NM_002800	−1.71	−2.86	−1.60	5698
PTPRB	AL080103	−1.94	−2.67	−3.25	5787
PXK	NM_017771	−1.15	−1.69	−1.21	54899
RAB8B	NM_016530	−3.34	−2.68	−1.60	51762
RAD1	NM_002853	−4.35	−3.35	−1.85	5810
RASIP1	NM_017805	−2.42	−2.23	−6.29	54922
RBMS3	BX649052	−4.07	−3.66	−2.98	27303
REA	NM_007273	−2.94	−2.73	−1.12	11331
RHOBTB1	NM_014836	−2.85	−3.33	−2.31	9886
RHOJ	AA187037	−2.01	−2.74	−3.30	57381
RNF149	NM_173647	−3.62	−2.83	−1.44	284996
RRM1	NM_001033	−3.25	−2.66	−2.13	6240
SEC15L1	NM_019053	−3.81	−3.80	−4.81	54536
SERPINB1	NM_030666	−0.64	−1.26	−1.19	1992
SF3B14	NM_016047	−3.37	−3.28	−1.11	51639
SIAT1	NM_003032	−1.37	−2.31	−2.99	6480
SIAT9	NM_003896	−1.99	−2.31	−2.99	8869
SLC12A2	U30246	−0.66	−1.96	−2.63	6558
SLC25A29	BU197217	−3.04	−3.18	−1.29	123096
SLC27A3	NM_024330	−0.85	−1.17	−2.07	11000
SLC35A5	NM_017945	−3.14	−3.37	−2.41	55032
SLC40A1	NM_014585	−3.67	−3.09	−3.93	30061
SNRK	NM_017719	−3.13	−3.08	−2.78	54861
SPATA7	NM_018418	−4.87	−4.92	−2.86	55812
Spc25	NM_020675	−2.15	−0.56	−1.43	57405
SPTBN1	NM_003128	−0.77	−2.18	−2.73	6711
SRA1	AK054960	−1.07	−0.69	−4.10	10011
SRPRB	NM_021203	−2.85	−2.69	−1.54	58477
STMN1	NM_203401	−2.88	−3.34	−2.54	3925
STMN1	NM_203401	−0.85	−0.69	−1.39	3925
STRBP	NM_018387	−2.08	−2.21	−2.21	55342
SULT1B1	NM_014465	−3.01	−0.93	−2.46	27284
SYNCRIP	NM_006372	−4.26	−3.12	−2.08	10492
SYNPO	AB028952	−3.10	−3.18	−3.62	11346
TACC3	NM_006342	−1.63	−0.70	−1.11	10460
TAF7	NM_005642	−2.46	−2.44	−1.87	6879
TAX1BP3	NM_014604	−2.45	−3.50	−2.19	30851
TGFBRAP1	BI857114	−1.15	−1.24	−1.59	9392
TIMM13	NM_012458	−3.64	−2.97	−1.32	26517
TM4SF14	NM_030927	−0.70	−1.21	−2.00	81619
TM7SF3	NM_016551	−2.40	−2.76	−1.95	51768
TMED9	BM920666	−1.86	−2.12	−1.85	54732
TMEM37	BC046362	−1.07	−0.98	−4.22	140738
TncRNA	W37893	−1.49	−0.83	−1.60	283131
TOMM22	NM_020243	−1.00	−1.23	−1.50	56993
TOP2A	NM_001067	−1.42	−1.13	−2.37	7153
TPX2	NM_012112	−1.45	−0.93	−3.08	22974
TREX1	NM_016381	−3.03	−2.95	−1.47	11277
UBE2R2	NM_017811	−3.20	−3.09	−2.18	54926
UPP1	NM_003364	−1.59	−2.53	−2.73	7378
UREB1	NM_031407	−3.41	−3.50	−1.54	10075
USP1	NM_003368	−3.46	−2.60	−2.84	7398
VPS11	NM_021729	−3.12	−3.44	−1.52	55823
WARS	NM_004184	−1.17	−1.97	−1.98	7453
WASL	D88460	−3.71	−2.30	−3.15	8976
WDFY1	NM_020830	−1.96	−2.34	−1.60	57590
XLKD1	NM_006691	−1.16	−0.94	−4.37	10894
YWHAB	NM_003404	−0.68	−0.97	−1.52	7529
YWHAQ	NM_006826	−2.98	−3.36	−1.70	10971
ZA20D3	NM_019006	−1.95	−2.85	−2.15	54469
ZCSL3	NM_181706	−2.29	−2.09	−1.52	120526
ZCWCC1	NM_014941	−4.28	−4.01	−1.60	22880
ZNF9	NM_003418	−2.78	−2.60	−1.61	7555

*Log ₂ ratios were calculated using the median of the corresponding intensity values. Duplicated gene names (MTUS1, STMN1 and SERPINE1) correspond to two different probes.

Table 2.

Upregulated genes in HHT versus control BOECs*

Gene	Access GenBank	log 2 (HHT1/C)	log 2 (HHT2n/C)	log 2 (HHT2m/C)	ID
APOBEC3C	NM_014508	0.76	1.22	1.50	27350
C18orf21	NM_031446	0.66	0.81	1.81	83608
C21orf57	BM829897.1	1.03	1.51	1.95	54059
CaMKIINalpha	NM_018584.4	1.57	1.43	2.20	55450
CAPG	NM_001747.1	1.46	1.81	2.79	822
CCR7	NM_001242	1.99	1.66	1.80	1236
CLCN5	BX098173.1	2.81	2.46	2.50	1184
COX7A2L	AA888530.1	5.45	5.58	5.38	9167
CYP1B1	NM_000104.2	4.89	5.58	5.90	1545
DCTD	BC108278	1.25	1.15	1.44	1635
DLGAP1	NM_001003809	0.75	0.89	1.51	9229
DYRK4	NM_003845.1	4.10	4.02	4.12	8798
FAM14A	NM_032036.1	0.89	0.96	1.85	83982
FAM20C	NM_020223.1	1.08	0.98	2.68	56975
FBN2	NM_001999.2	4.01	6.41	5.02	2201
FHOD3	NM_198900	1.28	0.90	1.08	80206
FLJ10357	NM_018071	2.09	1.75	1.61	55701
FLJ20582	NM_014106	1.71	1.53	1.50	54989
FLJ23588	NM_198856	3.50	3.22	3.01	64800
FLJ25351	AI919317.1	2.51	2.26	2.11	151254
FTH1	BG939615.1	0.70	0.62	1.04	2495
GALP	NM_033106.1	0.59	0.68	1.55	85569
GAS6	NM_000820.1	2.45	1.23	2.99	2621
GLIPR1	NM_006851.1	2.16	2.29	2.28	11010
GLS2	NM_013267	1.52	0.91	1.16	27165
GRIP2	AF052177.1	2.31	2.15	1.80	80852
HLA–DMB	NM_002118.3	2.97	2.42	2.38	3109
HRH1	NM_000861.2	0.93	0.89	1.36	3269
INHBA	NM_002192.1	1.28	1.42	1.80	3624
KIAA1641	AK024934.1	1.91	1.60	1.48	57730
KIF1A	NM_022113	2.74	2.25	2.36	547
KITLG	NM_000899	0.88	1.26	1.35	4254
KRT19	NM_002276.3	1.11	3.26	1.70	3880
LRRN6A	NM_032808	2.38	2.25	2.11	84894
MT1E	NM_175617	0.63	1.32	2.43	4493
MYL9	NM_006097.3	1.44	1.11	3.13	10398
MYO15A	BQ447410.1	3.15	2.82	2.83	51168
NDN	NM_002487.2	2.46	4.22	3.90	4692
NDUFA8	NC_000009	3.58	3.84	2.72	4702
NFAT5	NM_173215	2.08	2.00	1.64	10725
PCBP4	NM_033010	0.61	0.72	1.25	57060
PHF12	NM_020889.1	2.65	2.41	2.32	57649
PHLDA2	NM_003311.2	1.11	1.35	1.99	7262
PPP1R12A	NC_000012	1.41	0.92	1.21	4659
PRSS12	NW_922162	1.37	0.99	2.18	8492
REC14	NM_025234	2.32	2.26	1.99	80349
RNASET2	NM_003730.3	1.76	1.28	1.62	8635
SEMA3B	NM_004636.1	2.11	2.04	2.06	7869
SERPINE1	NM_000602.1	1.22	1.16	1.02	5054
SERPINE1	AU158706.1	1.01	1.08	2.50	5054
SETDB1	NM_153759	1.92	3.00	5.41	9869
SLC35B2	NM_178148	1.24	0.93	1.37	347734
SLC35D3	NM_001008783	1.71	1.44	1.38	340146
SPARC	AA634615.1	1.03	0.77	1.20	6678
TNFRSF12A	NM_016639.1	1.41	1.07	1.35	51330
VPS16	NM_007259	3.43	3.19	3.17	64601
WFDC3	BF002489.1	1.87	1.81	1.63	140686
ZNF169	NM_194320	0.88	0.77	1.08	169841

*Log ₂ ratios were calculated using the median of the corresponding intensity values. Duplicated gene names (MTUS1, STMN1 and SERPINE1) correspond to two different probes.

As expected, ALK1 was not only downregulated in HHT2, but also in HHT1 (Table  1 ), in agreement with previous reports ( 16 ). Unfortunately, the stringent selection procedure did not reveal the decrease of endoglin levels. Out of the three different endoglin probes used, two of them gave very low hybridization signals and the third one cross-hybridized with Cynl2 , according to the CodeLink information. Therefore, endoglin data were excluded from Table  1 . Nevertheless, the endoglin downregulation was evidenced by quantitative PCR as described below.

The most representative genes affected in HHT cells, and hence putative targets of the TGF-β endothelial pathway, were: genes involved in cell adhesion/cytoskeleton/migration (15–20%), genes affecting DNA and RNA regulation (15%), genes controlling cell cycle and survival (10–15%), genes involved in angiogenesis (5–10%), genes affecting signal transduction (10–12%), in addition to genes involved in general metabolism (20%) (Fig.  1 A and B).

Figure 1.

Diagrams representing the distribution in percentages of the different cellular functions played by the downregulated ( A ) and upregulated ( B ) genes in HHT endothelial cells, compared to controls. The percentage of the genes represented in each sector, followed by the number of genes (in parenthesis), is indicated.

Experimental validation at the mRNA and protein levels of differentially expressed genes in HHT endothelial cells

Due to their interest in the HHT pathology, we decided to focus our attention in some genes related to angiogenesis, migration/wound-healing, cell-to-cell adhesion, extracellular matrix (ECM) or NO synthesis. We chose two genes affecting cell cycle division since 15% of differentially expressed genes are involved in this function. At the same time, the validation of molecular targets involved in the disease may be useful as putative therapeutic targets. Selected genes are shown in Table  3 (italicized in Tables  1 and 2 and Supplementary Material, Table S1), grouped according to their function. Eight genes were chosen: seven of them, WASL (Wiskott–Aldrich syndrome protein), NOS-3 (endothelial nitric oxide synthase, eNOS), CCNB2 (cyclin B2), CDC25B (cell division cycle 25B) , PCDH12 (vascular endothelial cadherin 2, VE-cad 2), PECAM 1 (platelet endothelial cell adhesion molecule) and ANGPT-2 (angiopoietin-2, Ang-2) were downregulated, while Serpine 1 (plasminogen activator inhibitor 1, PAI-1) was upregulated. The validation of these results was carried out at the RNA level by quantitative PCR and at protein level by Western blot or ELISA using different samples of RNA and proteins from healthy individuals and HHT patients. In addition, heterozygous eng^+/− mice versus their normal siblings were used to support the findings in humans at protein level.

Table 3.

Selected genes differentially expressed in HHT cells

Adhesion/Migration	Cell cycle	Cytoskeleton	Vascular physiology	Angiogenesis	ECM remodeling
PECAM 1 ↓PCDH12 ↓	CCNB2 ↓CDC25B ↓	WASL ↓	NOS-3 ↓	ANGPT-2 ↓	Serpine1 ↑

Quantitative PCR was performed for each of the selected genes. As shown in Figure  2 A, PCR confirmed the results of mRNA downregulation for the seven genes selected: WASL, NOS-3 , CCNB2, CDC25B, PCDH12 (VE-cad 2), PECAM 1 and ANGPT-2 (Ang-2 ) and the upregulation of Serpine 1 ( PAI-1) . The mRNA levels of endoglin and ALK1 were also measured and found to be downregulated in HHT1 and HHT2 cells (Fig.  2 B), in agreement with our previous reports ( 16 , 20 , 21 ). The experimental validation by quantitative RT–PCR of all the genes selected supports the validity of our microarray analysis. As an internal control of the quantitative RT–PCRs, Figure  2 A shows the upregulation of COX-2 (cyclooxigenase 2) in HHT versus control BOECs. This is in agreement with the reported increase of COX-2 in HHT1 mice as an adaptative mechanism to the decreased expression of eNOS ( 22 ). Taken together, the quantitative RT–PCR of all the genes selected supports the validity of our microarray analysis.

Figure 2.

Quantitative RT–PCR of genes differentially expressed in HHT cells. ( A ) Selected genes differentially expressed in HHT BOECs. RT was made with RNAs from different control ( n = 4), HHT1 ( n = 5), HHT2n ( n = 3) and HHT2m ( n = 3) independent BOECs. Quantitative PCR was carried out using the system based in the universal probe library from Roche. The oligonucleotides used for the amplification of each probe are described in Table  4 . In all cases, the RNA levels are shown as fold downregulation or upregulation relative to control taken arbitrarily as 1. Results were made in triplicates, and repeated at least twice. Representative experiments are shown. Differences are in all cases statistically significant with respect to control ( P < 0.001, except in those bars marked with * P < 0.05). ( B ) Endoglin and ALK1 in HHT cells. Quantitative RT–PCR was made as in panel A, using endoglin and ALK1 specific oligonucleotides. In all cases, the RNA levels are shown as fold downregulation relative to control taken arbitrarily as 1. Results were made in triplicates, and repeated at least twice. Differences with respect to control BOECs are in all cases statistically significant (* P < 0.05; ** P < 0.01; *** P < 0.001).

Table 4.

Oligonucleotide sequences and probe numbers used for the quantitative RT–PCR

Gene	Primers	Probe §
ANGPT-2	Left TGCAAATGTTCACAAATGCTAA	75
	Right AAGTTGGAAGGACCACATGC
PECAM 1	Left AGAAAACCACTGCAGAGTACCAG	31
	Right GGCCTCTTTCTTGTCCAGTGT
CDC25B	Left AATGAACTTCACGCCTCAGC	28
	Right CGTCGAGACAGGGATAGGTG
CCNB2	Left TGGAAAAGTTGGCTCCAAAG	7
	Right CTTCCTTCATGGAGACATCCTC
NOS-3	Left GACCCTCACCGCTACAACAT	5
	Right CCGGGTATCCAGGTCCAT
Serpine 1	Left TCCAGCAGCTGAATTCCTG	15
	Right GCTGGAGACATCTGCATCCT
PCDH12	Left GGAGGAGCTCACTGTGGATT	14
	Right CTGATGCAGCAAGGACAGC
WASL	Left CCAGATACGACAGGGTATCCA	5
	Right GCAGGTGTTGGTGGTGTAGA
COX-2	Left TCACGCATCAGTTTTTCAAGA	23
	Right TCACCGTAAATATGATTTAAGTCCAC
ENG	Left AGCCACATCGCTCAGACAC	41
	Right GCCAATACGACCAAATCC
ACVRL1	Left ATCTGAGCAGGGCGACAC	79
	Right ACTCCCTGTGGTGCAGTCA
GAPDH *	Left AGCCTCAGCCCCACAAGT	60
	Right GTCACCTCGTCCCTCTCG
18S *	Left CTCAACACGGGAAACCTCAC	77
	Right CGCTCCACCAACTAAGAACG

*These genes were used as endogenous controls. ^§ Numbers correspond to the Universal probe Library from Roche.

Figure  3 A shows western blots for proteins encoded by downregulated and upregulated genes in HHT BOECs. In all cases, a clear downregulation of the protein level for the six first products and an upregulation of PAI-1 were obtained. Most of these proteins (eNOS, cyclin B2, CDC25B, PECAM 1 and PAI-1) were also analyzed by western blot using lung extracts from heterozygous eng^+/− and eng^+/+ mice (Fig.  3 B) corroborating the data obtained with human cells.

Figure 3.

Western blots of selected proteins whose genes are differentially expressed in HHT cells. ( A ) HHT BOECs. Around 30 µg of protein extracts were loaded in 7.5% SDS–PAGE gels. After electroblotting, nitrocellulose membranes were incubated with specific primary antibodies and the corresponding secondary antibodies. The experiments were repeated at least three times using different control ( n = 4), HHT1 ( n = 5), HHT2n ( n = 3) and HHT2m ( n = 3) independent BOECs. The results of the western blots shown are representative. Bands corresponding to each protein were quantified by densitometry with respect to actin (loading control). The average ± SD was calculated and represented in histograms on the right hand side (* P < 0.05; ** P < 0.01; *** P < 0.001). ( B ) Lung lysates of eng^+/+ and eng^+/− mice. Western blots were essentially made as in A, but using lung extracts from homozygous wild type endoglin mice eng^+/+ , and their corresponding heterozygous siblings. The results were normalized against tubulin. The experiments were repeated at least three times, and the bands shown are representative. An endoglin western blot was included as a control to show that heterozygous mice are expressing about a half of the levels of endoglin, due to the haploinsufficiency. Blots for WASL and VE-cadherin were not included, since the antibody used for human cells in A, was not working in mice. Bands corresponding to each protein were quantified by densitometry with respect to tubulin (loading control). The average ± SD was calculated and represented in histograms on the right hand side (* P < 0.05; ** P < 0.01; *** P < 0.001). RU, Relative Units.

Ang-2 is a soluble factor present in plasma with a potential functional relevance in HHT ( 23 ). Thus, protein levels of Ang-2 were also validated in human and murine systems. Decreased levels of Ang-2 were confirmed in plasma samples from 10 healthy adult donors and 26 HHT patients (Fig.  4 A), as well as in plasma from eng^+/+ and eng^+/− mice (Fig.  4 B). In both systems, Ang-2 levels in HHT are about half the levels of healthy donors.

Figure 4.

Ang-2 levels in plasma from HHT and non-HHT patients and from eng^+/+ and eng^+/− mice. Ang-2 average concentration (with standard deviation) was determined by an ELISA test from plasma. ( A ) HHT Patients. Samples from ten control donors and 26 HHT patients including both types, HHT1 and HHT2 were measured. Ang-2 levels were determined by interpolation in the standard curve obtained with the R&D kit. ( B ) eng^+/− mice. Four homozygous wild type mice and four endoglin heterozygous siblings were used. The values are expressed as relative concentration compared to wild type (control mice). Results were highly significant (*** P < 0.001 in both cases).

Adhesion, migration and transendothelial leukocyte infiltration are decreased in HHT endothelial cells

The downregulation in HHT BOECs of genes such as PECAM 1 ( 24 ) and VE-cad 2 ( 25 ), directly involved in cell-to-cell and cell-to-substrate adhesion, and others like WASL ( 26 ), participating in the cytoskeleton structure, prompted us to study the adhesion properties of HHT versus control BOECs. Overall, HHT cells displayed less adhesion than control cells, but the most marked difference was shown by HHT1, with less than half of the adhesion shown by control BOECs (Fig.  5 A). Although the degree of adhesion varied among the different HHT types, the general conclusion is that adhesion is decreased in HHT versus control cells. These results agree with a previous report showing in vitro that cancer prostate cells have less adhesion to plates when they express less endoglin ( 27 ).

Figure 5.

Adhesion and Migration (wound-healing experiments) in HHT BOECs. ( A ) Adhesion of HHT BOECs. Cell adhesion experiments were performed in P-96 plates using different control ( n = 3), HHT1 ( n = 5), HHT2n ( n = 3) and HHT2m ( n = 3) BOECs. Results were obtained after normalization using a crystal violet colorimetric method. The difference in adhesion of HHT-BOECs relative to non-HHT (control) cells is significant. Each experiment was made by triplicates and repeated twice (* P < 0.05; ** P < 0.01; *** P < 0.001). ( B ) Wound-healing experiments in HHT BOECs. Confluent monolayers from different control ( n = 4), HHT1 ( n = 5), HHT2n ( n = 3), and HHT2m ( n = 3) BOECs were scratched with the tip of a pipet. At different time intervals, 12, 18 and 24 h, the degree of proliferation and migration of cells to cover the monolayer discontinuity was recorded by photography. Each type of cell was studied in triplicates, and the experiment was repeated twice. ( C ) The surface area of the wound healing experiments (B) was quantified and the results are shown relative to the initial denuded area, taken arbitrarily as 100%. This figure represents the mean and standard deviation of the experiments performed.

In addition to adhesion, cell motility is another property affected by the downregulation of genes involved in cytoskeleton organization, signal transduction from cell receptors to actin cytoskeleton and genes coding for components of the ECM. Thus, the migration ability of HHT versus control BOECs was visualized and measured by time course wound-healing (Fig.  5 B and C). While control cells close the cell layer discontinuity 18 h post wound, HHT1 and HHT2n close after 24 h. HHT2m migrate but do not heal the wound in a uniform way, and this is not closed after 24 h, not even after 48 h (data not shown). These results are in agreement with migration deficiencies shown by murine eng^+/− endothelial cells and angiogenesis deficiencies in eng^+/− mice ( 28 ).

The downregulation of genes such as PECAM 1, PLCG2, F11R/JAM-1 , and ESAM (Table  1 ), directly involved in leukocyte transmigration, could result in a decreased interaction with leukocytes during the steps of firm adhesion, rolling and endothelial transmigration. A decreased leukocyte migration from the blood stream to the inflammatory foci in tissues would lead to an incomplete inflammatory response. This function was tested in HHT1 heterozygous mice eng^+/− versus their wild type siblings treated with carrageenin as inflammatory agent to stimulate the transmigration to peritoneal fluid. The transmigration of total white cells was estimated counting the number of leukocytes in blood and peritoneal fluid (Fig.  6 A and B). As predicted, a significant decrease in leukocyte endothelial transmigration in carrageenin-treated eng^+/− mice compared with their wild type siblings was observed, as the number of leukocyte in peritoneal fluid was lower in eng^+/− than in eng^+/+ after stimulation with carrageenin. Furthermore, leukocyte count in blood decreased after i.p. carrageenin injection less in eng^+/− than in eng^+/+ , thus revealing a lower transmigration rate to the tissues (Fig.  6 A).

Figure 6.

Endothelial leukocyte transmigration in HHT1 eng^+/− heterozygous mice versus their eng^+/+ siblings. Total leukocyte endothelial transmigration was monitored in carrageenin-treated and -untreated eng^+/− mice and their eng^+/+ siblings by counting the number of leukocytes in blood ( A ) and peritoneal fluid ( B ). Data are the average of three experiments with four to five mice per experiment and group. * P < 0.01, carrageenin-treated versus controls.

HHT endothelial cells differ in the proliferation rate compared to control endothelial cells

The downregulation in HHT cells of genes controlling the cell cycle, such as CDC25B and cyclin B2 , justified assessing proliferation rates and cell cycle studies in HHT BOECs when compared with control cells. A proliferative analysis showed that the proliferation rate of HHT1 was about half of the control endothelial cells (BOECs or HUVECs), while HHT2n cells divided slightly faster than HHT1, but still significantly slower than non HHT-BOECs. Of note, HHT2m cells that show the highest proliferation rate (Fig.  7 A), also display the lowest migration capacity (Fig.  5 B and C), thus excluding a proliferation-dependent effect in the migration experiments. Support for these proliferative results was obtained by flow cytometry analysis of the different cell cycle stages (Fig.  7 B). While most of the HHT2m cells were in S-phase, most of the HHT1 cells were in G ₀ –G ₁ , and only a very small fraction (tail in the G ₀ –G ₁ peak) corresponded to S-phase in HHT2n. Control-BOECs and HUVECs clearly showed a peak representing G ₂ .

Figure 7.

Studies of cell proliferation and cell cycle in HHT BOECs. ( A ) Proliferative assay. Cells were seeded by triplicates in P-24 well plates and allowed to proliferate for 24 h in an incubator at 37°C and 5% CO ₂ . After 24 h, the number of cells present was measured by a MTT assay, measuring relative absorbance at 542 nm. The experiment was repeated three times using different control ( n = 4), HHT1 ( n = 5), HHT2n ( n = 3) and HHT2m ( n = 3) BOECs, and the results shown are representative. In addition to control cells, HUVEC (Human umbilican vein endothelial cells) were used. Differences of HHT BOECs with respect to control are significant. HUVECs showed a slightly higher and significant proliferation rate than BOECs. (* P < 0.05; *** P < 0.001). ( B ) Cell cycle analysis by flow cytometry. Cells from the different sources described in panel A were seeded and allowed to proliferate for 24 h, fixed and their DNA content was estimated by propidium iodide in a flow cytometry apparatus. Experiments with each cell type were repeated three times, and a representative experiment is shown.

DISCUSSION

The present work represents a systematic and careful attempt to throw light on endothelial downstream HHT target genes. Our experimental model has tried to approach closely the cellular situation of the disease since the probes for the analysis come from primary cultures of HHT endothelial cells, targets of the disease. This study differs from previous attempts using HUVECs since in most cases the newborns do not show clinical symptoms of the disease. Moreover, BOECs have previously being reported as an adequate cellular model to dissect the disease ( 16 ). Previous attempts have used the microarray analysis technique to survey the multiple targets of TGF-β/ALK1 pathway in endothelial cells ( 17–19 , 29 ). These experiments were made by infection with adenovirus encoding a constitutively active mutant of ALK1. Each of these reports identified several genes related to proliferation and tubulogenesis of endothelial cells and angiogenesis, consistent with vascular defects observed in HHT patients and HHT mice models. Unfortunately, distinct and non-overlapping gene expression patterns were observed depending on the endothelial cell type, the infection protocol or the stimuli used (TGF-β or constitutively active ALK1) ( 19 ). Furthermore, a comparative analysis of those studies with the expression pattern described in the present report, revealed no overlapping among the downstream HHT target genes.

We have selected genes differentially expressed in HHT whatever the mutation ( endoglin or ALK1 ) to establish a common pattern of genes altered, i.e. an ‘HHT gene fingerprinting’. The list of differentially expressed genes includes 277 downregulated and 63 upregulated. Proteins encoded by many of these genes are involved in physiological processes such as angiogenesis, cell adhesion, migration, transmigration and matrix remodeling, TGF-β signaling, cell cycle control, cytoskeleton and vascular physiology (Fig.  8 ). Upon vascular injury or inflammation, angiogenesis and leukocyte transmigration are the two main events that occur in vascular physiology. In both situations, endothelial cells accurately respond to angiogenic signals by an appropriate coordination of genes involved in migration/adhesion to ECM, cytoskeleton reorganization, cell cycle control and survival of endothelial cells to keep the vessel homeostasis. In addition, blood pressure changes are buffered under the eNOS/COX-2 system, and the eventual hemorrhages occurring in capillaries are managed by a fibrinogenic/fibrinolytic equilibrium. It is important to stress that these processes are altered in HHT BOECs, as we have demonstrated in results.

Figure 8.

The gene expression fingerprinting of HHT and vascular biology. A schematic diagram shows a vessel with some associated functional processes. The genes listed in the figure are involved in these vascular processes, and are dysregulated in HHT cells. According to the literature, some of these genes may be involved in more than one physiological process. Asterisks indicate genes very important for the angiogenesis process. Downregulated genes are in green boxes, while upregulated genes are in red boxes.

Angiogenic factors are decreased in HHT

Different observations report that HHT patients have alterations in angiogenesis and microvessel density ( 30 ). Moreover, in the murine HHT model, endothelial cells display impaired tube formation ( 28 ) and eng ^+/− mice showed impaired angiogenesis. Furthermore, by ‘ in vitro ’ culture of endothelial cells from HHT patients, Fernandez-L et al . ( 16 ) have shown that HHT BOECs are unable to form proper tubes on matrigel compared with the organized cord structure formed under the same conditions by endothelial cells of healthy donors. A whole set of genes tightly related with angiogenesis are downregulated in HHT cells (Table  1 and genes marked with asterisks in Fig.  8 ). Fyn is a Src-family tyrosine kinase induced by Ang-2 in the process of migration and tube formation of capillaries ( 31 ). The angiogenic switch gene HMGB1 modulates hypoxia-stimulated angiogenesis ( 32 ). EDG-1, ESAM and PLCG2 maintain vascular integrity, vessel formation and permeability. Interestingly, angiogenic and inflammatory processes are in close relationship as illustrated by the fact that Ang-2 sensitizes endothelial cells to TNF-α ( 33 ). On the other hand, SPARC , an anti-angiogenic gene, is upregulated in HHT endothelial cells (Table  2 ). Of note, genes controlling cell cycle, survival and migration, should be kept in mind, as they also contribute to the angiogenic process.

The fact that Ang-2, a soluble factor, is downregulated in HHT, might have an important diagnostic value. Thus, in this work we have demonstrated that plasma of HHT patients and eng^+/− mice have less circulating Ang-2 than that of healthy individuals or wild type littermate mice. This decrease in Ang-2 levels is about half of the normal levels.

Cell cycle progression is affected in HHT

Around 10% of the downregulated genes are involved in cell cycle control. Cyclin B2 ( CCNB2 ) is involved in G ₁ /S cycle arrest by TGF-β ( 34 ). CDC25B and CHEK1 are controlling the S/mitosis transition checkpoint (Fig.  8 and Table  1 ). CCNA2 is a cyclin that binds and activates CDC2 or CDK2 kinases, and thus promotes both G ₁ /S and G ₂ /M cell cycle transitions, while the tyrosine monooxygenase protein (YWHAQ) controls specifically the G2/M checkpoint.

In general, there is a tight relationship between TGF-β and the cell cycle progression by a crosstalk among cytokines in endothelial cells ( 35 ). Thus, HHT cells, with decreased TGF-β signaling ( 14 , 16 ) and decreased expression of cell cycle promoting genes would be affected in cell division. In this context, angiogenesis and blood vessel stabilization, requiring cell division after injury or inflammation, would also be hampered.

Genes affecting endothelial cytoskeleton, cell adhesion, migration and leukocyte transmigration are altered in HHT

We have shown that HHT cells are less adherent than control BOECs and that migration of HHT cells is delayed when closing the layer discontinuity, artificially induced in wound-healing experiments. Moreover, in vivo experiments in HHT1 mice have shown that upon inflammatory stimuli, leukocyte transendothelial migration is also decreased (Fig.  6 ). Altogether, almost 15% of the total HHT downregulated genes (Table  1 ) affect cytoskeleton, migration, cell adhesion and transmigration. Special interest deserves GTPases (RhoJ, RhoB1, Rho GD1) and WASL which associated with GTPase Cdc42 control the cytoskeletal organizing complex and cell adhesion ( 36 ). In this respect, VE-cad 2 and PECAM 1 contribute to cell adhesion, cell migration, transmigration and recognition of cytoskeletal elements ( 24 , 37 , 38 ). As both genes are downregulated in HHT, all these processes would be decreased. Moreover, components of the ECM and receptors, Col3A, HMMR, HSPG2 and ITGA6, downregulated in HHT, would contribute to a delayed cell migration.

Leukocyte transmigration is crucial for the inflammatory response. The downregulation of genes as PECAM 1, PLCG2, F11R/JAM-1 and ESAM would result in an incomplete inflammatory response. This function has been shown to be disturbed in eng ^+/− heterozygous mice (Fig.  6 ), and is compatible with the recent finding that endoglin has a crucial role in blood cell-mediated vascular repair ( 38 ).

Taken together, a decreased capacity for migration, adhesion and leukocyte transmigration would lead to difficulties in wound-healing following inflammation or vascular accidents. This is in agreement with the fragile capillaries and frequent nose and gastrointestinal bleedings present in HHT patients ( 7 ).

Vascular functional alterations in HHT

Here we find that HHT cells have reduced levels of NOS-3 mRNA. This is in agreement with the finding that eNOS levels in eng^+/− mice arteries are about half than in eng^+/+ mice ( 39 ). Also, the eNOS half-life was significantly reduced in HHT1 HUVECs due to a deficiency in eNOS/Hsp90/endoglin coupling ( 40 ). Compatible with this, we have found a downregulation of HSPCB and other chaperones, in addition to a decrease in NOS-3 mRNA, thus supporting the decrease of NO production in HHT1 mice. Moreover, we have demonstrated that HHT BOECs have an increase in COX-2 RNA levels, possibly as an adaptation for eNOS decrease in these cells. This finding agrees with previous results showing specific upregulation of COX-2 in the vascular endothelium of eng^+/− mice as well as in endothelial cells with a decreased endoglin expression ( 38 ).

PAI-1 is upregulated in HHT-BOECs. It affects migration, p53-mediated senescence and endothelial apoptosis, being essential for the plasminogen/plasmin equilibrium in fibrinolysis. As reported previously, endoglin and PAI-1 expression are inversely correlated: cells overexpressing endoglin have reduced levels of PAI-1 ( 41 ). Accordingly, we have found in this paper that HHT BOECs with less endoglin ( 16 ) show an increase in PAI-1 expression. Interestingly, PAI-2 ( Serpine B ) is among the downregulated genes. This proves the strength of our microarray-based model since it has been recently shown that PAI-1 and PAI-2 levels are inversely correlated in human placenta ( 42 ).

Developmental genes affected in HHT

It is worth mentioning that four genes among those upregulated in HHT are involved in development. In particular, the products of the genes FAM14 and FAM20 are expressed during early stages of hematopoietic development ( 43 ). Other upregulated genes in HHT are LRRN6 and FHOD-3 that are involved in axonal guidance and organogenesis, respectively ( 44 , 45 ). The dysregulation of these developmental genes in HHT is compatible with the finding that endoglin and ALK1 null mice embryos die at 10–11.5 days after gestation ( 6 , 8 ).

The present study opens the interesting question of how to relate mutations in endoglin or ALK1 and the observed misregulation of their downstream targets. According to the literature, the genes validated in this study are directly affected by the TGF-β pathway. Angiogenesis is regulated by VEGF, in turn induced by hypoxia and TGF-β in endothelial cells ( 46 ). Moreover, hypoxia, TGF-β and VEGF are the main specific upregulators of Ang-2 in microvasculature ( 47 ). Cell cycle is tightly regulated by TGF-β ( 48 ) and in this regard, a functional association TβRII/cyclin B2 has been described in endothelial cells ( 35 ). PECAM 1, VE-cad 2 and WASL participating in cell adhesion, migration and actin polymerization are regulated by TGF-β signaling ( 49 ). eNOS is localized in caveolae interacting with TGF-β receptors ( 50 ), and at the same time, NOS-3 expression is induced through Smad2 ( 51 ). The eNOS/COX-2 balance is regulated by TGF-β in mesangial cells ( 52 , 53 ). Finally, PAI-1 is also regulated by TGF-β through Smad2 signaling ( 54 ). In summary, TGF-β is, directly or indirectly, controlling the genes differentially expressed in HHT cells.

Conclusions

This work has allowed the identification of downstream target genes affected in HHT cells. Special emphasis was put on the identification of common targets to reveal a ‘gene expression fingerprinting’ for HHT, given the similarity in symptoms in HHT1 and HHT2. We propose that HHT clinical manifestations are consequence of an imbalance in expression of multiple genes. These genes are involved in migration, cytoskeleton, cell-to-cell and cell-to-ECM adhesion, inflammation, transmigration, angiogenesis and vascular physiology. These affected functions allow us to draw an hypothetical sequence of events that could explain the clinical manifestations of HHT starting at prompted sites of recurrent infectious or physical vascular stress. A decrease in endothelial cell migration, leukocyte extravasation, local angiogenesis, slower cell cycle progression and altered cell cytoskeleton would result in improper tissue remodelling at those sites. Moreover, a decrease in NO production and the consequences in blood pressure on the wall vessel would contribute a capillary network breakdown. This hypothesis is compatible with the spatial distribution of the lesions observed in the HHT patients: nose, lung, liver, tissues preferentially exposed to inflammation, infection or mechanical aggression, and provides a suitable framework to investigate new therapeutical approaches to treat HHT.

MATERIALS AND METHODS

Blood outgrowth endothelial cells from HHT patients

BOECs were grown from 50 ml peripheral blood, culturing buffy coat mononuclear cells on collagen I coated wells using EGM-2/EBM-2 medium (Cambrex) ( 16 ). Control BOECs were obtained from four independent non related donors (ages 24, 40, 45 and 35). The nomenclature used to identify the HHT patients and mutations was as described ( 21 ). HHT1 BOECs were from family no. 1 (R171X mutation; patients 01/01/03 and 01/02/03), family no. 11 (P131L mutation; patients 11/01/04, 11/02/04 and 11/03/04) and family no. 30 (K216Q mutation; patient 30/01/05). HHT2n (nonsense) BOECs were from family no. 10 (R145fs mutation; patients 10/01/04 and 10/02/04) and family no. 26 (R47P mutation; patient 26/02/04). HHT2m (missense) BOECs were from three members of family no. 2 (R384W mutation, patients 02/20/05, 02/09/04 and 02/02/03).

Microarray hybridization and signal detection

RNAs were extracted from 10 ⁶ BOECs cells using RNAeasy kit (Qiagen). Total RNA from each sample was labeled, processed and independently hybridized on a CodeLink human whole genome DNA chip (Amersham Biosciences, Uppsala, Sweden), containing 55 000 human gene targets. Briefly, hybridizations were made in triplicates for each probe according to manufacturer's instructions. Slides were scanned with a GenePix Array Scanner and the image from each bioarray was processed using the CodeLink expression analysis software.

Microarray statistical analysis

Raw intensity values were normalized by the quantile method implemented in the Bioconductor package limma ( http://www.bioconductor.org ). Data for each experimental set (control, HHT1, HHT2m and HHT2n) were filtered according to the following steps: (1) only spots with two or three quality spots assigned as G (good) or L (low) were selected for subsequent analysis; (2) for gene spots with only two suitable data, the third one was assigned as the mean of the other two values; and (3) data values less than 0 were assigned a value of 10. After the indicated filtering procedures were performed, data were joined in the following experimental groups: control versus HHT1, control versus HH2m and control versus HHT2n. Gene spots with differences in their median values less than 50 between the two groups compared were not included in the statistical analysis. Statistical analysis for each of the selected groups was carried out by the local pooled error method (package LPE from Bioconductor, http://www.bioconductor.org ). The P -values obtained were adjusted for multiple hypothesis testing using the control of the false discovery rate based procedure developed by Benjamini et al . ( 55 ), and implemented in the Bioconductor package multitest. Those genes with adjusted P -values lower than 0.05 were considered as differentially expressed. To gain insight of the common genes affected in HHT, independently of the genomic defect, and to reduce noise due to inter-individual gene expression, only annotated genes (those with an assigned Gene ID) differentially expressed in the three experimental groups were acquired. All procedures to handle data were implemented by R, Perl, gawk and bash scripts.

Quantitative RT–PCR

Endoglin and ALK1 oligonucleotides were purchased from Sigma and designed according to Roche software for real time PCR. Total RNA from different types of BOECs were extracted using the RNAeasy kit (Qiagen), retrotranscribed using the AMV RT kit from Roche company and amplified in a real time PCR using the Universal Human Probe Roche library and the Real time PCR kit (Roche). The assays were made by triplicates, compared with two different types of endogenous controls ( 18S RNA and GAPDH ) and repeated at least twice. The oligonucleotide sequences for the analysis of each gene are shown in Table  4 .

Western blot analysis

BOECs or mouse lung extracts were obtained by lysis on ice for 30 min in 1% SDS. Lysates were centrifuged at 14 000 × g for 5 min. Aliquots of cleared cell lysates were boiled for 5 min in SDS sample buffer and analyzed in 7.5% SDS–PAGE under reducing conditions. Proteins were electrotransferred to nitrocellulose membranes, followed by immunodetection with the following antibodies from Santa Cruz: CDC25B (sc-5619), eNOS (sc-654), PAI-1 (sc-8979), WASL (sc-20770) and cyclin B2 (sc-2830). Anti-human vascular endothelial cadherin 2 monoclonal (VEcadherin 2) was purchased from Chemicon (MAB1989), and anti-human monoclonal PECAM 1 (clone HC1/6) was produced in our laboratory. Incubations were done following the manufacturer. To normalize the protein load, anti-human actin polyclonal (A-2103), or anti-tubulin (T-5168) antibodies from Sigma Corp. were used diluted 1:10 000. Secondary antibodies were horseradish peroxidase mouse or rabbit conjugates from Dako. Membranes were developed by enchanced chemiluminescence Amersham (ECL, Advanced Western blot kit, GE).

Ang-2 ELISA

A commercial kit from R&D Systems was used for human Ang-2 detection from 200 µl of plasma collected from HHT patients and healthy, non-HHT controls. The same procedure was carried out from 50 µl of eng^+/− ( 8 ) or their wild type homozygous eng^+/+ littermates. The whole protocol was performed following the instructions given by the manufacturer.

Flow cytometry

To analyse the cell cycle in BOECs, a total of 10 ⁵ cells were seeded in a well of a P-6 plate. Cells were collected after 24 h in culture, fixed in PBS:formaldehyde (9:1) and incubated with RNase and propidium iodide. Samples were analyzed on a Coulter Epics XL flow cytometer. Experiments were repeated at least twice.

Adhesion and migration (wound-healing) assays

For adhesion assays, cells (5×10 ⁴ in 100 µl) were added in triplicates to 96-well dishes (highbinding; Costar, Cambridge, MA, USA). Plates were spun for 15 s at 400 rpm to place cells in contact with the ligands, incubated for 2 min at 37°C and unbound cells were removed by three washes with DMEM. Bound cells were quantified using a plate reader after crystal violet staining.

For wound-healing assays, in vitro scratched wounds were created by scraping confluent cell monolayers in P-24 plate wells with a sterile disposable pipet tips. The remaining cells were washed with HBSS buffer (Hanks, Gibco) and incubated with the enriched EBM-2/EGM-2 (10% FCS) medium (Cambrex) up to 48 h. Endothelial cell migration into denuded area was monitored by photography of the plates every 4–6 h.

Transendothelial migration of total leukocytes in eng^+/− and eng^+/+ mice

Haploinsufficient mice for endoglin ( eng^+/− ), kindly provided by Dr Michelle Letarte ( 8 ), and their littermates ( eng^+/+ ) were handled as previously described ( 22 , 28 , 39 ). Animals weighing between 18 and 20 g, were intraperitoneally injected with 0.2 ml of a 1.5 mg/ml carrageenin solution (mixture of lambda and kappa, Sigma Chemical Co, St Louis, MO, USA). Twenty-four hours after the injection of carrageenin, resident peritoneal cells were harvested by washing the pleural cavity with 10 ml of sterile PBS. After a slight massage, 8–9 ml of PBS were withdrawn with a side-down-bevelled syringe. Cells were centrifuged, resuspended in 1 ml of PBS and subjected to analysis using a cytometer (Advia 120, Bayer) counting of white blood cells (WBC). The total number of cells was corrected for the dilutions performed along the experiments. Blood was obtained from the tail tip (200 µl), collected into heparinized tubes and WBC counted using a cytometer (Advia 120, Bayer). Results are expressed as an average ± SD of three experiments, with n = 4–5 mice per experiment.

Cell proliferation assay

Cell survival was estimated using a microculture tetrazolium assay. This assay measures the reduction of the substrate [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide] (MTT) to a dark blue formazan product by mitochondrial dehydrogenases in living cells. Briefly, 50 µl of 2 mg/ml MTT (Sigma) was added to each well. After 3 h at 37°C in darkness, 0.01 N HCl containing 10% SDS was added to dissolve the formazan product. Absorbance at 540 nm was measured in a plate reader (Bio-Tek Instruments, ELX 800). Experiments, in triplicates, were repeated twice, and the results shown are representative.

Statistical general analysis

Results are expressed as mean ± SD of data obtained from three or more experiments performed in triplicates, unless otherwise stated. Statistical significance was determined using the two-tailed Student's t -test. * P < 0.05, ** P < 0.01, *** P < 0.001.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online.

ACKNOWLEDGEMENTS

Authors are indebted to Drs A. Puig-Kröger, E. Sierra-Filardi and A. Corbi for assessments and facilities with the microarray technique, Dr P. Lastres for flow cytometry analysis, the Genomics laboratory of the Universidad Complutense de Madrid , for assistance with the microarrays, Dr. R. Zarrabeitia (Hospital Sierrallana) for providing HHT blood samples, and Drs M. Guzman and A. Carracedo for help with real time PCR experiments. This work was supported by grants Fondo de Investigación Sanitaria (PI020200) to CB and Ministerio de Educación y Ciencia (SAF05–01090 to LMB and SAF2004–01390 to CB, GEN2003–20649-C06–06/NAC to MAV, BFU2004–00285/BFI to JML-N), and HHT Foundation International Inc to CB. LMB, MAV and CB were supported by Genoma España to the development of the microarray experiments. Fernandez-L is a predoctoral fellow of I3P CSIC, MP is a predoctoral fellow of Junta de Castilla y Leon and EMG-M is a predoctoral fellow of Ministerio de Educacion y Ciencia . ARB is the recipient of a fellowship from ‘Ramon y Cajal’ program.

Conflict of Interest statement . None declared.