The Axl receptor tyrosine kinase is an adverse prognostic factor and a therapeutic target in esophageal adenocarcinoma

Abstract

Esophageal adenocarcinoma (EAC) arises in the backdrop of reflux-induced metaplastic phenomenon known as Barrett esophagus. The prognosis of advanced EAC is dismal, and there is an urgent need for identifying molecular targets for therapy. Serial Analysis of Gene Expression (SAGE) was performed on metachronous mucosal biopsies from a patient who underwent progression to EAC during endoscopic surveillance. SAGE confirmed significant upregulation of Axl “tags” during the multistep progression of Barrett esophagus to EAC. In a cohort of 92 surgically resected EACs, Axl overexpression was associated with shortened median survival on both univariate (p < 0.004) and multivariate (p < 0.036) analysis. Genetic knockdown of Axl receptor tyrosine kinase (RTK) function was enabled in two EAC lines (OE33 and JH-EsoAd1) using lentiviral short hairpin RNA (shRNA). Genetic knockdown of Axl in EAC cell lines inhibited invasion, migration and in vivo engraftment, which was accompanied by downregulation in the activity of the Ral GTPase proteins (RalA and RalB). Restoration of Ral activation rescued the transformed phenotype of EAC cell lines, suggesting a novel effector mechanism for Axl in cancer cells. Pharmacological inhibition of Axl was enabled using a small molecule antagonist, R428 (Rigel Pharmaceuticals). Pharmacological inhibition of Axl with R428 in EAC cell lines significantly reduced anchorageindependent growth, invasion and migration. Blockade of Axl function abrogated phosphorylation of ERBB2 (Her-2/neu) at the Tyr877 residue, indicative of receptor crosstalk. Axl RTK is an adverse prognostic factor in EAC. The availability of small molecule inhibitors of Axl function provides a tractable strategy for molecular therapy of established EAC.

Introduction

The incidence of esophageal adenocarcinoma (EAC) is increasing at an alarming pace in the United States (>600% increase since 1975).¹ Established epidemiological risk factors for EAC include male gender, Caucasian ethnicity, abdominal obesity, pre-existing gastro-esophageal reflux disease (GERD) and most importantly, Barrett esophagus, a metaplastic precursor lesion to EAC. In the multistep carcinogenesis model of EAC, meta-plastic Barrett epithelium progresses through the intermediate stages of low-grade and high-grade dysplasia, culminating in invasive cancer. Unfortunately, the overwhelming majority of patients (>90%) with EAC present de novo, i.e., without a pre-existing diagnosis of Barrett esophagus.² Patients with de novo EAC are almost always diagnosed at an advanced stage of disease, which likely accounts for their dismal overall five-year survival (∼13%). The treatment of choice for advanced esophageal cancer is esophagectomy, usually preceded by neo-adjuvant chemo-radiation therapy.³ However, despite the advances in treatment and an increase in the proportion of cases diagnosed at an early and hence potentially curable stage, the overall five-year survival of EAC has changed only minimally in the last three decades, underscoring the dire need for developing more potent therapies for this malignancy.

Global transcriptomic profiling of epithelial cancers and their precursor lesions provide an unbiased opportunity for the identification of candidate diagnostic and prognostic biomarkers and can elucidate novel therapeutic targets. Herein, we generated “Long” Serial Analysis of Gene Expression (L-SAGE) libraries,⁴ using a panel of mucosal biopsies obtained from a Barrett esophagus patient who developed EAC during the course of endoscopic surveillance and identified “tags” corresponding to the Axl receptor tyrosine kinase (RTK) as being significantly and progressively upregulated during multistep esophageal carcinogenesis. In the experiments presented below, we establish that Axl expression is an independent adverse prognostic factor in surgically resected EAC and that this RTK regulates multiple components of the neoplastic phenotype (such as invasion, migration and in vivo engraftment) in EAC cell lines. Further, we characterize the major intracellular effectors of Axl, including the identification of Ral GTPase proteins as novel effectors of the Axl signaling cascade, particularly in the context of regulating cell motility. The recent availability of small molecule pharmacological inhibitors of Axl function provides a unique opportunity for developing targeted therapies for advanced EAC.

Results

Progressive upregulation of Axl and its cognate ligand Gas6 during multistep progression of barrett esophagus to EAC.

In order to identify transcriptomic abnormalities during Barrett esophagus progression, four independent “long” SAGE (L-SAGE) libraries⁴ were generated using RNA obtained from histologically validated mucosal biopsies of normal squamous epithelium (NSE), low-grade dysplasia (LGD), high-grade-dysplasia (HGD) and EAC, all from a single individual on endoscopic surveillance (see Sup. Results, Sup. Tables 1–3 and Sup. Fig. 1 for details of L-SAGE analysis).

From the list of upregulated L-SAGE transcripts in Table 1, we selected the Axl RTK for further validation in tissue samples. This was based on the emerging association between Axl and human cancer,^5–8 and the assumption that RTKs, in general, have emerged as the most promising therapeutic candidates in recent years.⁹ Immunohistochemical analysis for Axl protein expression was performed on an independent set of archival tissue microarray samples, representing the entire histological spectrum from metaplastic Barrett esophagus to adenocarcinoma and comprised of 292 independent tissue “cores” from 92 patients with surgically resected EAC (Fig. 1A). Minimal Axl expression was observed in esophageal squamous epithelium, gastric cardiac mucosa, non-dysplastic Barrett mucosa and in LGD, with significant upregulation of expression observed in samples of HGD, EAC and associated lymph node metastases (p < 0.05, Fig. 1B). When the EACs per se were stratified by level of Axl staining, 56 of 92 (61%) cases demonstrated high levels of expression. Amongst a list of clinicopathological variables, Axl overexpression was significantly associated with depth of invasion (p = 0.007) and with T stage in the TNM classification (p = 0.027) (Sup. Table 4). In this cohort of 92 EACs, Axl overexpression was significantly associated with reduced median survival (p = 0.004) on univariate analysis, along with T stage and presence of metastases (Fig. 1C and Table 2); of note, Axl expression was also an independent predictor of median survival on multivariate analysis (p = 0.036, Table 3). These findings in a relatively large cohort of EACs, and associated precursor lesions, underscore a putative role for Axl in the progression from Barrett mucosa to adenocarcinoma, as well as in imparting an adverse prognosis to the subset of cases with aberrant expression.

Table 1.

Trend of significantly upregulated SAGE tags in esophageal L-SAGE libraries

17bpTAG	Symbol	Gene name	Unigene	Entrez_ID	Band	Fold
ACC AAG AAC CCA AAA GC		Homo sapiens, clone IMAGE:3138608, mRNA	Hs.542984	-	Unknown	175
GTG CAC TGA GCT GTA AC	HLA-A	Major histocompatibility complex, class I, A	Hs.181244	3105	6p21.3	49
GGA GGC TGA GGC AGG AG	ATF7IP	Activating transcription factor 7 interacting protein	Hs.714407	55729	12p13.1	32
TTG GCC AGG CTG GTC TA	AXL	AXL receptor tyrosine kinase	Hs.590970	558	19q13.1	30
TTG GCC AGG CTG GTC CC	EGFL11	EGF-like-domain, multiple 11	Hs.453441	346007	6q12	29
TCT AAA ACA CCA AAA GC		Similar to hypothetical protein (L1H 3 region) - human, mRNA (cDNA clone MGC:46671 IMAGE:5563071)	Hs.617329	201853	Unknown	22
TCC TTT GCC CAC TTT TT						16
AGC CAC CGT GCC TGG CC	CBX5	Chromobox homolog 5 (HP1 alpha homolog, Drosophila)	Hs.349283	23468	12q13.13	16
CAC CTG TAA TCC CAG CT		MRNA; cDNA DKFZp434L042 (from clone DKFZp434L042)	Hs.543963	-	Unknown	12
CCT GTA ATC CCA GCT AC	SLC31A1	Solute carrier family 31 (copper transporters), member 1	Hs.532315	1317	9q31–q32	10
CCT GTA GTC CCA GCT AC	MAFF	V-maf musculoaponeurotic fibrosarcoma oncogene homolog F (avian)	Hs.517617	23764	22q13.1	10
GTT GTG GTT AAT CTG GT	B2M	Beta-2-microglobulin	Hs.709313	567	15q21–q22.2	9
TGC CTG TAG TCC CAG CT		MRNA; cDNA DKFZp586G021 (from clone DKFZp586G021)	Hs.598965	-	Unknown	9
TGC CTG TAA TCC CAG CT	GTSE1	G-2 and S-phase expressed 1	Hs.386189	51512	22q13.2–q13.3	8
CAC CTG TAG TCC CAG CT	C14orf64	Chromosome 14 open reading frame 64	Hs.651477	388011	14q32.2	6
CCT GTA ATC CCA GCA CT	HLA-E	Major histocompatibility complex, class I, E	Hs.650174	3133	6p21.3	6
GCG AAA CCC CGT CTC TA	C10orf58	Chromosome 10 open reading frame 58	Hs.500333	84293	10q23.1	5
GTT CAC ATT AGA ATA AA	CD74	CD74 molecule, major histocompatibility complex, class II invariant chain	Hs.436568	972	5q32	5
GAG AAA CCC TGT CTC TA	VDR	Vitamin D (1,25-dihydroxyvitamin D3) receptor	Hs.524368	7421	12q13.11	5
GTG AAA CCC CAT CTC CA	IL17RA	Interleukin 17 receptor A	Hs.709911	23765	22q11.1	4
GTG AAA CCC TGT CTC TA	CCBE1	Collagen and calcium binding EGF domains 1	Hs.34333	147372	18q21.32	4
CCA CTG CAC TCT AGC CT	DIXDC1	DIX domain containing 1	Hs.655626	85458	Unknown	4

Tags over 3-fold upregulation are shown in this table.

Figure 1.

Axl overexpression is an adverse prognostic factor in EAC. (A) Immunohistochemical analysis of Axl expression was performed in multistage Barrett progression using a panel of archival tissue microarrays. Axl expression restricted to the basal (regenerative) layer of normal squamous epithelium is observed in a minor subset of cases. Minimal to largely absent Axl expression is observed in non-dysplastic Barrett epithelium and in low-grade dysplasia. Robust Axl expression is observed in approximately a quarter of high-grade dysplasia. Axl is overexpressed in ∼60% of primary EACs and in a comparable percentage of lymph node metastases. (B) The histogram summarizes the frequencies of tissue “cores” in each histopathologic category with high levels of Axl expression. (C) Kaplan-Meier survival analysis of 92 Barrett-associated EAC demonstrates that cancers with high AXL expression have a significantly worst median survival than cancers with absent/low AXL expression (Log rank test p = 0.036).

Table 2.

Univariate survival analysis of EAC

Factor	Characteristics	Number	Median survival (months)	p value
Sex	Male	80	21.3	0.72
	Female	12	20.4
Age	<55	15	26.5
	55–74	66	16.3	0.07
	>75	11	9.4
Location	Esophageal	65	23.9
	Esophageal-GE Junction	6	11	0.97
	GE junction	21	17.9
Differentiation grade	Well	3	108.2	0.16
	Moderately	33	30.6
	Poor	53	16.6
T classification	Tis	1	-	0.002*
	T1	21	108.2
	T2	16	30.2
	T3	54	13
Lymph node metastasis	Present	48	15.3	0.25
	Absent	37	30.2
	Unknown	7	-
Metastasis	Present	6	7	0.001*
	Absent	86	23.9
AXL expression	Absent/low	36	30.2	0.004*
	High	56	15.3

^*Significant at the significance level <0.05. (Log rank Test).

Table 3.

Multi-variate analysis of survival in EAC

Variables	p value	Relative risk	95% confidence interval
T classification	0.013*	3.60	1.51–8.66
Metastasis	0.026*	3.12	1.14–8.53
AXL expression	0.036*	1.91	1.04–3.49

^*Significant at the significance level <0.05 (Cox Regression).

In order to confirm whether Axl receptor upregulation is accompanied by concomitant overexpression of the activating ligand, we performed immunohistochemistry for Gas6 in serial sections of the identical TMAs used above. The pattern of Gas6 expression was essentially comparable to that of Axl, with maximal upregulation occurring at the stage of invasive cancer and lymph node metastases (Sup. Fig. 2A demonstrates Gas6 labeling in a single patient along the multistep progression of Barrett esophagus to EAC; B is the summary for all 292 independent cores). In contrast to Axl however, we found no impact of Gas6 expression levels in EACs upon median survival (Sup. Fig. 2C). Since Gas6 upregulation has previously been implicated as a negative prognostic influence in solid cancers, we postulated whether expression of its cognate receptor in tissues might be a limiting factor. The 92 EACs were categorized into four sub-cohorts, based on the independent quantification of Gas6 ligand and Axl receptor expression, respectively (Sup. Fig. 2D). Kaplan-Meier survival estimates demonstrated that, even in cancers with high Gas6 ligand expression, Axl remained a limiting factor in influencing survival, with a persistent significant difference in outcome between cases with “high” and “low” receptor levels.

Axl regulates multiple facets of the transformed phenotype in EAC cell lines.

We then assessed the level of expression of Axl and its cognate ligand, Gas6,^10,11 in a panel of non-neoplastic esophageal keratinocytes (NEK2 and HEEPIC) and EAC cell lines (OE33 and JH-EsoAd1), as a prelude to functional studies. Axl protein was overexpressed in both EAC lines compared to the esophageal keratinocytes (Sup. Fig. 3A). In light of the striking Axl overexpression in JH-EsoAd1 cells, genome-wide copy number analysis was performed using 244 K microarrays (Agilent), which revealed a chromosome 19q13.1 amplification encompassing the AXL gene locus (Sup. Fig. 4); copy number gain in JH-EsoAd1 cells was confirmed by genomic qPCR and by fluorescence in situ hybridization. Pronounced Gas6 expression was also seen in the JH-EsoAD1 cells, suggesting the existence of an autocrine and/or paracrine ligand-receptor feedback loop in these cells. We next investigated whether knockdown of Axl function in EAC lines would impact upon their transformed phenotype. JH-EsoAD1 and OE33 cells were stably infected with lentiviral vectors expressing AXL short hairpin RNA (shRNA) and nearcomplete knockdown of transcript and protein was confirmed in both cell lines (Sup. Fig. 3B and C). No inhibition of in vitro cell viability was observed in either cell line upon Axl blockade (data not shown); in contrast, modified Boyden chamber assays demonstrated significant impairment of invasion and migration in both JH-EsoAd1 and OE33 clones expressing AXL shRNA, compared to the respective scrambled shRNA controls (Figs. 2A–D, respectively). The deleterious impact of Axl knockdown on the transformed phenotype of EAC was further underscored by the significant inhibition of anchorage independent growth in OE33 cells (Fig. 2E) and the complete failure of tumor engraftment in immunocompromised mice using JH-EsoAd1 cells (Fig. 2F).

Figure 2.

Axl regulates multiple facets of the transformed phenotype in EAC cell lines. (A and C) Loss of endogenous Axl function significantly inhibits migration at 24 hours in JH-EsoAd1 and OE33 cells, as assessed by a modified Boyden chamber assay. Error bars = ± S.E.M. (N = 3); p < 0.05, compared with scrambled. (B and D) Loss of endogenous Axl function significantly inhibits invasion at 48 hours in JH-EsoAd1 and OE33 cells, as assessed by a modified Boyden chamber assay. Error bars = ± S.E.M. (N = 3); p < 0.05, compared with scrambled. (E) Loss of endogenous Axl function significantly reduces the property of anchorage independent growth (assessed by colony formation in soft agar) in OE33 cells. Higher magnification of soft agar assay is shown on top. Error bars = ± S.E.M. (N = 3); p < 0.05, compared with scrambled. (F) Loss of endogenous Axl function abrogates the ability of JH-EsoAd1 cells to engraft in NOD/SCID mice. In three animals, xenografts are visible in all flanks implanted with scrambled infected JH-EsoAd1 cells (white arrow), while no xenografts have formed on the opposite flank, corresponding to cells expressing AXL shRNA. Histological analysis of explanted scrambled expressing JH-EsoAd1 xenograft confirms adenocarcinoma histology (bottom, right).

Axl regulates EAC cell motility through Ral-dependent mechanisms.

Previously, multiple reports in cancer cells, as well as in non-neoplastic settings, have implicated the phosphatidylinositol-3-kinase (PI-3-kinase)/Akt and p42/p44 mitogen activated protein kinase (MAPK) pathways as effectors of Axl function.^7,12,13 We examined the status of activation of these two pathways in JH-EsoAd1 and OE33 cells versus the respective clones stably expressing an AXL shRNA; in each case, the pathways were assessed in either the presence or absence of the cognate Axl ligand, Gas6. Not unexpectedly, Gas6 stimulation of the Axl receptor was accompanied by increased phosphorylation of Akt and Erk1/2 (as a measure of MAPK pathway activation) in both cell lines. The effects of AXL shRNA expression were, however, divergent between the two models. Thus, in the JH-EsoAd1 cells, Axl knockdown completely eliminated Gas6-induced phosphorylation of Akt at the Ser⁴⁷³ residue and modestly decreased Erk1/2 phosphorylation (Fig. 3A). In contrast, the Gas6-induced activation of either pathway was, for the most part, unaffected in OE33 clones expressing AXL shRNA, likely due to the existence of other concurrent RTK-dependent upstream signals (Fig. 3B). Nonetheless, in light of the significant phenotypic effects of Axl knockdown on OE33, comparable to that observed in JH-EsoAd1 cells (see above), we postulated the existence of an effector mechanism regulated by Axl that is independent of either PI-3-kinase/Akt or MAPK pathways.

Figure 3.

Sustained Axl function is required for maintaining EGF-dependent activation of Ral GTPase proteins. (A) Gas6 ligand induces tyrosine phosphorylation of Axl receptor in JH-EsoAd1 cells, which is accompanied by phosphorylation of Akt at Ser⁴⁷³ and p42/p44 MAPK (Lanes 1 and 3). Knockdown of endogenous Axl completely abrogates Gas6-induced Akt phosphorylation and modestly inhibits Gas6-induced p42/p44 MAPK phosphorylation (Lane 4). Phosphorylation of Axl receptor is assessed by immunoprecipitation with PY99 anti-phosphotyrosine antibody, followed by western blot with anti-Axl antibody. (B) Gas6 ligand induces tyrosine phosphorylation of Axl receptor in OE33 cells, which is accompanied by phosphorylation of Akt at Ser⁴⁷³ and p42/p44 MAPK (Lanes 1 and 3). In contrast to JH-EsoAd1 cells, knockdown of endogenous Axl has minimal effects on Gas6-induced phosphorylation of either Akt or p42/p44 MAPK. (C) In both JH-EsoAd1 and in OE33 cell lines, knockdown of Axl profoundly decreases EGF-dependent activation of Ral GTPase isoforms (RalA and RalB). Activation of Ral proteins was assessed by immunoprecipitation for GTP-bound moieties in scrambled and AXL shRNA expressing clones, followed by western blot analysis for the respective Ral isoform. (D) In both JH-EsoAd1 and in OE33 cell lines, loss of Ral GTPase activity is also accompanied by decreased levels of GTP-bound (active) Cdc42, a Rho GTPase family member that is a credentialed Ral effector protein. (E) Expression of a constitutively active form (Rlf-CAAX) of the RalGEF, Rgl2, in OE33 cells with Axl knockdown restitutes the levels of GTP-bound RalA protein to that observed in control cells in the presence of EGF ligand. Lane 1: Mock vector expressing OE33 cells with stable Rlf-CAAX expression; Lane 2: Scrambled shRNA expressing OE33 cells; Lane 3: OE cells with stable AXL shRNA expression; Lane 4: OE cells with stable co-expression of AXL shRNA and Rlf-CAAX. (F) Restitution of RalA activity is associated with partial, but significant, rescue of cell motility in OE33 cells, including migration (left) and invasion (right) phenotypes. *significant downregulation of cell motility in AXL shRNA expressing cells compared to scrambled (p < 0.05); **significant rescue of cell motility in AXL shRNA and Rlf-CAAX co-expressing cells compared to cells with Axl knockdown alone.

In recent years, the Ral GTPase proteins have emerged as a key mediator of oncogenic effects downstream of growth factor-stimulated receptor activation.¹⁴ In particular contexts (for example, in the setting on oncogenic Ras), the two Ral GTPase isoforms—RalA and RalB—have been demonstrated to be the principal determinants of both cellular transformation and motility.^15,16 Therefore, we investigated the effects of Axl knockdown on the activity of the Ral GTPase proteins. Exogenous Gas6 ligand failed to enhance the “baseline” levels of GTP-bound (i.e., active) Ral proteins, while addition of epidermal growth factor (EGF) markedly increased these levels in both JH-EsoAd1 and OE33 cell lines (data not shown). This result is to be anticipated, as Ral proteins are well established EGF effectors.^17,18 Axl knockdown strikingly inhibited this EGF-dependent activation of RalA and RalB (Fig. 3C). One of the effector targets of Ral proteins is RalBP1, which demonstrates GTPase activating protein (GAP) activity for the Rho family GTPase, Cdc42.^19,20 In both JH-EsoAd1 and OE33 cells with Axl knockdown, reduction in Ral activity in the presence of EGF was also accompanied by a profound decrease in GTP-bound Cdc42 protein (Fig. 3D). In order to further corroborate that downregulation of Ral signaling is indeed the underlying cause of the observed phenotypic differences between EAC cells with and without endogenous Axl function, we generated OE33 AXL shRNA clones with stable co-expression of a constitutively active form (Rlf-CAAX) of the Ral guanine exchange factor (Ral GEF), Rgl2.²¹ Rlf-CAAX expression restored RalA-GTP levels in OE33 AXL shRNA clones to that observed in scrambled shRNA infected OE33 clones (Fig. 3E). This restitution in Ral function was accompanied by partial, but statistically significant (p < 0.05), “rescue” of both migration and invasion phenotypes when compared to OE33 clones with Axl knockdown (Fig. 3F). To the best of our knowledge, this is the first demonstration of a requirement for sustained Axl function in maintaining EGF-dependent Ral GTPase activation in human cancer and provides a mechanistic link between Axl and the regulation of cell motility.

A novel small molecule antagonist of Axl function impedes invasion and anchorage independent growth of OE33 cells and sensitizes them to lapatinib.

R428 is a recently described selective small molecule inhibitor of Axl tyrosine kinase activity that can block systemic metastases and improve survival of orthotopic xenograft models of breast cancer.²² In OE33 cells, R428 demonstrated dose-dependent and significant reduction in both anchorage independent growth in soft agar and in vitro invasion (Fig. 4A), comparable to results obtained with genetic (shRNA-mediated knockdown) (Fig. 2). Multiple reports have suggested that upregulation of Axl expression might represent a resistance mechanism to targeted therapies in solid tumors.^23–25 Particularly compelling is the recent data demonstrating that Axl activation is observed in breast cancer cells with resistance to ERBB2/Her-2neu targeted therapies like lapatinib and that concurrent inhibition of Axl activity renders these cells sensitive to lapatinib. The Tyr⁸⁷⁷ residue of ERBB2 correlates with, and provides a convenient readout of, the biological activity of the receptor.²⁶ We found dose-dependent reduction in ERBB2 Tyr⁸⁷⁷ phosphorylation upon exposure to R428 (Fig. 4B), which was mirrored by genetic abrogation of Axl function using shRNA (Fig. 4C). These results suggested the intriguing possibility of crosstalk between Axl and ERBB2 receptors. As an extrapolation of these results, we found that co-treatment of OE33 cells with R428 sensitized them to lapatinib in vitro (the IC₅₀ of single agent lapatinib was 5.5 vs 0.4 µM upon combination) (Fig. 4D) and that comparable sensitization was observed using a genetic strategy for Axl knockdown as well (Fig. 4E).

Figure 4.

Pharmacological abrogation of Axl function inhibits invasion and anchorage independent growth and sensitizes OE33 cells to the tyrosine kinase inhibitor lapatinib. (A) The Axl small molecule inhibitor R428 significantly downregulates in vitro invasion (modified Boyden chamber assay) at 48 hours in OE33 cells (left) and significantly blocks colony formation in soft agar at two weeks (right) compared to vehicle treated cells. Two doses (2 and 4 µM) were used in these experiments. (B) Phosphorylation of the Tyr⁸⁷⁷ residue of ERBB2 is inhibited in a dose-dependent manner in OE33 cell line upon pharmacological blockade of Axl with R428. GAPDH is used as loading control. (C) Phosphorylation of the Tyr⁸⁷⁷ residue of ERBB2 is essentially abrogated in OE33 cells with genetic (shRNA-mediated) knockdown of Axl, without affecting total ERBB2 levels. GAPDH is used as loading control. (D) MTS assays in OE33 cells confirm that addition of R428 can sensitize this cell line to lapatinib. The IC₅₀ for lapatinib is reduced from 5.5-0.4 µM, when co-treated with R428 versus the single agent. All MTS assays were performed in triplicate and cell viability is normalized to vehicle treated cells at 48 hours. (E) MTS assays in isogenic OE33 cells confirm that genetic knockdown of AXL with RNA interference sensitizes these cells to lapatinib, reducing the IC₅₀ for from 4.8-1.8 µM. All MTS assays were performed in triplicate and cell viability is normalized to vehicle treated cells at 48 hours.

Discussion

The primary objective of this study was to perform transcriptomic profiling of Barrett esophagus-associated EAC, in order to identify candidate biomarkers and putative therapeutic targets for this malignancy. Ours is not the first report of generating SAGE libraries in either Barrett esophagus or EAC;^27–29 nevertheless, there are several unique facets to our study design that deserve comment. To the best of our knowledge, this is the first study to create SAGE profiles from non-invasive dysplastic lesions (LGD and HGD) that precede invasive EAC. The availability of these preneoplastic libraries in the public domain (http://cgap.nci.nih.gov/SAGE) should greatly facilitate future biomarker discovery efforts in Barrett esophagus. Second, our global profiling experiments were performed on endoscopic mucosal biopsies obtained from a single patient who progressed to cancer during surveillance, rather than in mismatched tissues from disparate individuals. Using material from one patient provides a unique perspective on multistep Barrett esophagus progression, devoid of confounding influences due to inter-individual variations. This is also the first SAGE analysis in this disease model that utilizes a 17 bp Long-SAGE (L-SAGE) strategy, rather than the “conventional” 10 bp approach (Short-SAGE). L-SAGE provides a higher specificity and fewer false positive results in tag-to-gene mapping compared to Short-SAGE.³⁰ Finally, our study validates the use of minute endoscopic biopsies (∼2–3 mm in greatest diameter) as a feasible substrate for library construction and large scale nucleic acid profiling. With the ever decreasing costs of next generation sequencing,³¹ it is likely that this approach will be extended to larger numbers of biopsy samples, in the context of genomic, epigenetic or transcriptomic profiling.

Bioinformatics analyses of the SAGE libraries identified the L-SAGE tag corresponding to the AXL gene transcript as being progressively upregulated during the transition from NSE through dysplasia to EAC. Axl is a member of the Tyro-3, Axl and Mer (TAM) family of RTKs, which has been implicated in a diverse array of physiological functions such as cell adhesion, cell motility, proliferation and regulation of inflammation.⁸ Axl was originally isolated as a transforming gene from human leukemia cells.³² Since then, aberrant expression of Axl has been reported in a number of solid tumors, including pancreatic and breast cancers and gliomas, amongst others.^5,7,22,33,34 Based on this known association with human cancer and the general propensity for RTKs to emerge as tangible therapeutic targets in oncology,⁹ we selected Axl for further validation in EAC. In a relatively large cohort of 92 surgically resected EAC patients, Axl expression in the neoplastic cells was significantly associated with depth of invasion, T stage and decreased median survival; Axl also retained its independent adverse prognostic impact on survival on multivariate analysis, underscoring the biological relevance of cancer-specific overexpression. In the multistep progression of Barrett esophagus, significant Axl upregulation was observed essentially at the stage of HGD and beyond, with minimal to no expression in either non-dysplastic Barrett epithelium or in LGD. This raises the speculation that surface Axl expression on HGD or EAC could be targeted as a diagnostic biomarker (for example, using a fluorescent anti-Axl antibody or peptide),³⁵ or for the localized delivery of ablative therapies without incidental adverse effects in the surrounding epithelium. In EAC cell lines, genetic knockdown of endogenous Axl mitigated multiple aspects of the transformed phenotype, such as cell motility (migration and invasion), anchorage independent growth and engraftment in immunocompromised mice, reiterating the importance of sustained Axl function for tumor maintenance.

In previous studies, the PI-3-kinase/Akt and p42/p44 MAPK pathways have been implicated as principal intracellular effectors of Axl signaling in cancer cells, as well as in non-cancer settings.^7,12,13 In our current experiments, while Gas6-dependent enhanced phosphorylation of Akt Ser⁴⁷³ and p42/p44 was observed in both Axl-expressing EAC cell lines, knockdown of Axl failed to abrogate the activation of these effector pathways in OE33 cells, despite the obvious effects on phenotype. This led us to postulate the existence of additional effector(s) of Axl signaling in cancer cells. Although much of the scientific literature on effectors of RTK signal transduction has focused on PI-3-kinase/Akt and MAPK, the Ral GTPase proteins (RalA and RalB) have emerged as critical mediators of growth factor signaling in recent years.¹⁴ For example, epidermal growth factor (EGF) ligand stimulates lamellipodia formation and migration through Ras-dependent recruitment of RalGEFs to the plasma membrane, with resulting activation of Ral GTPase proteins.^17,18 Previously, Counter and colleagues have shown that genetic knockdown of the two Ral GTPase isoforms (RalA and RalB) in KRAS2-mutant pancreatic cancer cells can abrogate both tumor engraftment and experimental (tail vein) metastases in vivo.^15,16 One of the principal functions of Ral proteins is regulation of the multi-protein complex known as the exocyst, which, in turn, regulates diverse biological functions, such as maintenance of epithelial cell polarity, cell motility and cytokinesis.³⁶ As is the case for other GTP-binding proteins in the Ras superfamily, the level of Ral activation corresponds to the level of GTP occupancy of the Ral proteins, which is proximally controlled by the action of guanine nucleotide exchange factors (RalGEFs, which load GTP) and GTPase activating proteins (RalGAPs, which promote GTP hydrolysis to GDP). Genetic knockdown of Axl leads to profound diminishing of GTP-bound (active) Ral proteins in both OE33 and JH-EsoAd1 cells. Further, loss of Ral activity is paralleled by reduction in GTP-bound Cdc42, a Rho GTPase that is one of the best characterized effectors of Ral proteins, especially in the context of cell motility. Finally, a compelling evidence supporting a role for the Ral pathway as a mediator of Axl function comes from the “rescue” experiments performed using a constitutively active form (Rlf-CAAX) of the Ral GEF, Rgl2. Our studies present the first evidence for Ral GTPase as an effector of Axl signal transduction and provide a facile kinase target for therapeutic inhibition of Ral function in cancers.

In this study, we also confirm that R428, a newly described orally bioavailable and selective small molecule antagonist of Axl function,²² blocks invasion and anchorage independent growth of EAC cells in vitro. R428 inhibits in vivo metastases in multiple breast cancer preclinical models, which engenders the possibility that this agent might be beneficial in advanced EACs, especially in those adenocarcinomas that demonstrate evidence of Axl overexpression. Axl has also recently been implicated as a mechanism of chemoresistance in solid tumors, particularly to ERBB2/Her-2neu targeted therapies, such as lapatinib.²⁵ We have demonstrated that either pharmacological or genetic inhibition of Axl function abrogates phosphorylation of ERBB2 at the critical Tyr⁸⁷⁷ residue and sensitizes OE33 cells to lapatinib in vitro. Our data raises an intriguing possibility of crosstalk between Axl and the ERBB/HER families, which might contribute to sustained receptor activation and resistance to targeted agents like lapatinib. Comparable cross-talk has been reported between the epidermal growth factor receptor (EGFR) and insulin-like growth factor-1 receptor (IGF-1R) pathways, rendering cancer cells resistant to EGFR-antagonist therapies.³⁷ We speculate that combinatorial therapy using R428 or similar Axl inhibitors might provide additional efficacy in preclinical models of EAC in vivo and are currently pursuing this subject of investigation.

Materials and Methods

Sample collection for serial analysis of gene expression.

All studies with human tissues described herein were performed using a Hopkins Institutional Review Board approved protocol. SAGE libraries were constructed on endoscopic mucosal biopsies obtained from a 69 year old man with a diagnosis of Barrett esophagus, who progressed to EAC on surveillance. The histopathology on each of the four biopsies was confirmed on cryostatembedded sections by two expert gastrointestinal pathologists (EAM and AM) using established criteria,³⁸ and corresponded to normal squamous epithelium (NSE), low grade dysplasia in Barrett esophagus (LGD), high grade dysplasia in Barrett esophagus (HGD) and esophageal adenocarcinoma (EAC) (Sup. Fig. 1). Total RNA was extracted using the MirVana RNA Isolation kit (Ambion, Austin, TX), as per the manufacturer's instructions. RNA integrity was confirmed on the Agilent 2100 Bioanalyzer, following which RNA quantification was performed on the Voctor2 (Perkin Elmer, Waltham, MA) coupled with RiboGreen dye (Invitrogen, Carlsbad, CA).

Long SAGE library construction and data analysis.

Four Long SAGE (L-SAGE) libraries were generated using 10 µg of input RNA, with NlaIII as the anchoring enzyme and MmeI as the tagging enzyme, as previously described.⁴ The SAGE 2000 v4.5 software (www.sagenet.org) was used to extract SAGE tags, remove duplicate ditags and tabulate tag counts. Linker sequences used in library construction, 1 bp sequence variations and tag sequences occurring only once were removed from analysis. The SAGE tags were normalized to 200,000 tags and statistical pair-wise comparison of tag frequencies in a binomial approach was performed as described by Romualdi et al. The esophagus L-SAGE library information and tag counts are posted at the Cancer Genome Anatomy Project's SAGE Genie web site (cgap.nci.nih.gov/SAGE). Unsupervised cluster analysis of genes and samples was performed using a subset of the most differentially expressed tags, using the TM4 software.

Conflicts of Interest

Authors with the last names Holland, Yu, Heckrodt, Zhang, Ding, Goff and Sigh are employees of Rigel Pharmaceuticals. The small molecule inhibitor R428 is owned by Rigel. Authors with the last names Alvarez, Montgomery, Canto, Dunbar, Wang, Feldmann, Hong, Waghray, Haffner, Meeker, Roa, Marimuthu, Riggins, Eshleman, Nelkin, Pandey and Maitra have no conflicts to declare.

Author Contributions

H.A., G.F., S.M.H., M.W., M.C.H., A.K.M., J.C.R. and A.M. performed the experimental procedures described in this study; E.A.M., M.C., K.B.D., J.S.W., S.J.H., J.Y., T.J.H., J.Z., P.D., D.G., R.S., T.R.E., B.D.N. and A.P. provided critical reagents and technologies essential towards completion of this project; H.A., E.A.M., G.J.R. and A.M. conceived the project and wrote the manuscript.

Deposition of L-SAGE Data in a Public Database

The four L-SAGE libraries discussed in this study have been deposited on the CGAP SAGE Genie website (cgap.nci.nih.gov/SAGE/) and can be retrieved using the following four IDs: LSAGE_Esophagus_Normal_B_CN01; LSAGE_Esophagus_Dysplasia_B_LGD01; LSAGE_Esophagus_Dysplasia_B_HGD02; LSAGE_Esophagus_Adenocarcinoma_B_CA1D.