Background
Current standard of care for late stage and metastatic colorectal cancer (CRC) includes surgical resection followed by adjuvant 5-fluorouracil (5-FU) plus leucovorin, and oxaliplatin (Ox) or irinotecan (IT) chemotherapy until progression [
1]. However, the development of drug resistance and relapse with refractory disease limits the 5-year survival for metastatic CRC to less than 10 % [
2].
Progression is driven in-part by the G-protein-coupled chemokine receptor CXCR4, with high expression of CXCR4 in CRC being associated with a greater risk of recurrence and poor survival [
3‐
5]. CXCR4 enables migration towards its ligand CXCL12 (stromal cell-derived factor-1; SDF-1) present in common sites of metastasis such as the liver, lungs, lymph nodes and bone [
6‐
8]. In addition to assisting in the dissemination of CRC, CXCR4 has also been shown to support growth of metastases through its co-expression with glycoprotein CD133, a marker of cancer stem cells [
9].
It is now recognized that most tumors arise from a minority of cells capable of tumorigenesis [
10]. These ‘cancer stem cells’ or ‘cancer initiating cells’ (CICs) possess the capacity to both self-renew and differentiate to re-populate the full phenotypic heterogeneity of the tumor [
11]. Surface antigens used to identify CICs within CRC include CD133 [
12] and the cell adhesion molecule CD44 [
11], however their roles in metastatic disease remain controversial [
13]. With the identification of CICs, the possibility that selective action of drugs may not eliminate the cell populations responsible for recurrence has received attention. Recently, a subpopulation of uniquely metastatic CICs has been reported to co-express CD26 [
14].
CD26 is a multifunctional cell-surface protein that is variably expressed between different cancers but plays a role in regulating cancer progression and spread [
15]. Its overexpression is linked to reduced invasiveness of ovarian cancer [
16] and it is down-regulated during carcinogenesis of melanoma [
17]. However, CD26 expression is up-regulated in renal cell carcinoma [
18] as well as papillary and follicular thyroid carcinomas [
19,
20]. High levels of CD26 expression is associated with worse survival in CRC [
21].
That the consequence of CD26 expression by tumors remains equivocal is likely due to its diverse roles. In addition to directly facilitating adhesion to the extracellular matrix [
22,
23], CD26 has two associated enzymatic activities: an intrinsic dipeptidyl peptidase IV (DPPIV; EC 3.4.14.5) activity and a hydrolase activity due to anchored ecto-adenosine deaminase (ADA; EC 3.5.4.4) for which CD26 is the major binding protein [
15]. Clearance of adenosine by ADA can enhance immune surveillance, a process that is inhibited by the accumulation of adenosine in the tumor [
24‐
26]. Through its peptidase activity, CD26 controls CXCL12 concentrations [
27] and subsequent homing of hematopoietic progenitor cells to the bone marrow [
28] along with the metastatic behavior of endometrial carcinoma [
29] and Sézary cutaneous T cell lymphoma [
30]. The contrasting effects of CD26 in different tumors may arise in part from the consequences of different levels of DPPIV activity, leading to a biphasic effect on the CXCL12 axis. Regional degradation of CXCL12 by DPPIV activity may serve to hone the CXCL12 gradient in a similar fashion to its second receptor CXCR7, preventing ligand-mediated receptor desensitization of CXCR4 [
31], but higher levels of CD26 expression may result in gradient ablation.
Given the important roles of CXCR4 and CD26 in the tumorigenesis and metastatic outgrowth of CRC, we sought to observe whether their expression and function might change following exposure to anticancer agents. Our data show that the migratory phenotype of CRC cells is suppressed immediately following exposure to chemotherapeutic agents due to loss of CXCR4+ cells and elevation of CD26 peptidase; and this is associated with enrichment of a CD44+/CD133− cell subset.
Methods
Cell culture
HT-29, T84, HRT-18, SW480 and SW620 human CRC cells and the additional cell lines mentioned were obtained from the American Type Culture Collection. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, without antibiotics) supplemented with 5 % (HT-29, T84, and HRT-18) or 10 % (SW480 and SW620) v/v heat-inactivated newborn calf serum (NCS) and maintained as stocks in 75-cm2 flasks at 37 °C in a humidified atmosphere of 90 % air/10 % CO2. Cells for use in binding assays or for measurements of DPPIV enzyme activity were seeded into 48-well plates at 50,000 cells/well and allowed to adapt to culture for 48 h. The cells were then cultured in medium containing 1 % NCS for a further 48 h. For flow cytometry, cells were seeded into 6-well plates and allowed to adapt for 48 h prior to treatment. Cells for migration assays were cultured and treated in 10-cm dishes. Anticancer agents used were: 5-fluorouracil (5-FU), irinotecan (IT), cisplatin (Cis), vinblastine (Vin), and methotrexate (MTX) from Mayne Pharma Canada; oxaliplatin (Ox) from Sanofi Canada; and SN-38 from Toronto Research Chemicals. Where single drug concentrations are used, these were defined as optimal (typically, just maximal) based upon the response of the cells at that passage level and the lot(s) of drug as obtained from the supplier(s).
Assay for cell-surface CD26 and CXCR4 on cell monolayers
Cellular CD26 or CXCR4 protein levels were determined on cell monolayers at 4 °C as previously described [
32]. Our assay for CD26 and CXCR4 measures native protein expressed at the surface of viable cells, rather than total cellular protein as with e,g, a western blot. Briefly, plates were placed on ice, and the cultures were washed with binding assay buffer (BAB; 137 mM NaCl, 5 mM KCl, 24.8 mM Tris, 0.7 mM Na
2PO
4, 0.5 mM MgSO
4, 1 mM CaCl
2, pH 7.4), containing 0.2 %
w/
v bovine serum albumin (BSA) followed by a 60-minute incubation with 125 μL BAB containing 1 % BSA and 1 μg/mL antibody or isotype-matched control. Primary mouse anti-human monoclonal antibody against CXCR4 (clone 12G5) and mouse IgG
2a isotype-matched control antibody (clone G155-178) were from BD Pharmingen, mouse anti-human monoclonal antibody against CD26 (clone M-A261) and mouse IgG1 (clone W3/25) isotype controls were from Cedarlane Laboratories, and secondary
125I-labeled goat anti-mouse IgG, F(ab’)
2 fragment was obtained from PerkinElmer Life Sciences. After two washes with BAB containing 0.2 % BSA, the cells were incubated for 60 minutes with 125 μL BAB containing 1 % BSA and 1 μCi/mL
125I-labeled goat anti-mouse IgG, F(ab’)
2 fragment. After three more washes the cells were solubilized in 0.5 M NaOH and radioactivity was counted. To determine antigen-specific radioactivity, the nonspecific binding in the presence of the isotype control antibody was subtracted from that obtained with the target antibody. Cell counts were performed with a Coulter® model ZM30383 particle counter (Beckman Coulter). Data are expressed as antigen-specific radioactivity (cpm) per 100,000 viable cells.
Orthotopic tumor model
HT-29 cells (5 × 106 in 100 μL serum-free DMEM) were injected s.c. into the flanks of six week old female CD-1 nu/nu mice (Charles River) and tumors were allowed to grow for 18–20 d until approximately 7 mm in diameter. The tumor tissue donors were euthanized under ketamine/xylazine anesthesia, tumors were harvested aseptically, and all non-tumor tissue was dissected away. The tissues were washed in ice-cold DMEM and cut into ~1 mm3 pieces for tumor transplantation. Recipient immunodeficient mice were anesthetized with 70 mg/kg ketamine and 14 mg/kg xylazine i.p. and treated proactively with 0.3 mg/kg buprenorphrine i.p. for post-surgical analgesia. A 1-cm abdominal incision was made to the right of midline and the distal small intestine was exteriorized to locate the ileocecal junction. The proximal end of the ascending colon was identified and abraded gently with the wooden end of a cotton-tipped applicator. Three 1-mm3 tissue pieces were sutured onto the muscularis of the proximal ascending colon, taking care not to pierce the colon wall. The intestine was interiorized and the incision was sutured. Twenty-six and 28 days following surgery, mice were weighed and injected i.p. with drugs or vehicle control (saline). Two days after the second dose, they were euthanized. The treatment and analysis period of days 26–30 represented the best time window between formation of an anatomically well-integrated tumour (by day 24) and a risk of occlusion of the intestinal lumen by the expanding tumour (from day 32) in the case of HT-29 cells. Tumors were harvested and tissues were weighed and snap-frozen in liquid nitrogen or fixed in 4 % formaldehyde for later analysis. All procedures were approved by the Carleton Animal Care Facility University Committee on Laboratory Animals at Dalhousie University.
Immunolocalization of CD26 and CXCR4 in tumours
For visualisation of CD26, tumors were frozen in OCT® and sectioned at a thickness of 8 μm with a Leica CM 3050S cryostat (Leica Microsystems). Sections were mounted on slides and maintained at −20 °C. For immunohistochemistry, all steps were carried out at 4 °C, unless otherwise described. Sections were thawed briefly, rinsed with phosphate-buffered saline (PBS) containing 1 mg/mL BSA and 0.1 % Tween 20 (PBS/BSA/Tween), blocked with 3 % goat serum in PBS/BSA/Tween for 30 min, then incubated with 25 μL of PBS/BSA/Tween containing 5 μg/mL mouse anti-human CD26 primary antibody for 2 h in a humidified chamber. Sections were washed three times with PBS/BSA/Tween, and then incubated with 25 μL of PBS/BSA/Tween containing 2 μg/mL of an Alexa Fluor® 488-conjugated goat anti-mouse IgG secondary antibody for 2 h in a humidified chamber in the dark. Slides were washed a further three times, post-fixed with PBS containing 10 % formaldehyde for 10 min at room temperature, and rinsed with distilled water. Coverslips were mounted on sections using low-fade Gel/mount® and fluorescence was observed using a Leica DM 2000 fluorescence microscope (Leica Microsystems).
To observe CXCR4, formalin-fixed and paraffin wax-embedded tissue was sectioned and processed for immunoperoxidase procedures. Deparaffinised sections were subjected to antigen retrieval using 10 mM citrate buffer, pH 6.0 at 95 °C in a microwave. Rinsed sections were then stained for CXCR4 using the same procedure as for CD26, except that the bound primary antibody was identified using a Vectastain ABC kit (Vector laboratories, Burlingame, CA). Quantitation was performed in the absence of counterstain; the distribution of CXCR4 was visualised with a Harris’ hematoxylin counterstain.
Levels of CD26 or CXCR4 were analyzed with QCapture Pro® software. The average staining intensity of the tumor was measured in a randomly-selected area of constant dimension, determined by a blinded observer.
Flow cytometry
HT-29 cells were released from 6-well plates by TrypLE™ Express. Cells were washed with chilled flow buffer (PBS, 25 mM HEPES, 1 mM EDTA, 1 % BSA) and resuspended in 2 μg/mL CXCR4-APC (clone 12G5; BD Pharmingen) and CD26-FITC (clone M-A261; Serotec), combined CD26-FITC (clone M-A261; Serotec), CD44-APC (clone G44-26; BD Pharmingen), CD133-PE (clone AC133; Miltenyi Biotec), or fluorophore-tagged isotype controls (Miltenyi Biotec) for 45 min at 4 °C. Cells were then washed twice with flow buffer and resuspended in BSA-free flow buffer for analysis. Flow cytometry analysis was carried out with a BD FACSCalibur™ flow cytometer (BD Biosciences). Cell debris and aggregates were excluded based on scatter signals and 10,000 events were captured per sample. Data were analyzed using Flowing Software version 2.5.0 (University of Turku, Turku, Finland).
Following labeling with CD26-PE (clone M-A261; Serotec) cells were stained with annexin-V-FITC and propidium iodide (PI) according to manufacturer’s protocol (Roche Diagnostics) for detection of necrotic, apoptotic, and live cells. Analysis was carried using a Guava® easyCyte™ 8HT flow cytometer and associated InCyte software (Millipore).
DPPIV activity and ADA-binding capacity assays
DPPIV enzyme activity was measured spectrophotometrically using Gly-L-Pro
p-nitroanilide (Gly-Pro-
pNA; Sigma-Aldrich) as the DPPIV substrate [
32]. To measure the cellular capacity for ecto-ADA binding, HT-29 cells in 48-well plates were treated with 10 μg/mL calf spleen ADA1 (Worthington Biochemical) in medium for 60 min at 37 °C and then assayed for bound ADA using 1 μg/mL rabbit anti-bovine ADA antibody (Alpha Diagnostic International) and 0.5 μCi/mL
125I-labeled donkey anti-rabbit secondary antibody F(abʹ)
2 fragment (Amersham Biosciences), using the procedures previously described [
32].
Migration assays
Transwell® 8 μm pore size polycarbonate membrane inserts (Corning) were coated overnight at 37 °C with 1 μg/mL type V collagen. Drug- and vehicle-treated cells were released from culture by brief exposure to trypsin and resuspended at 0.5 - 2.5 × 106 cells/mL in serum-free DMEM containing 1 mg/ml BSA. One hundred microlitres of cell suspension were added to the upper chamber, and 600 μL of DMEM containing 1 mg/mL BSA and 100 ng/mL CXCL12 or vehicle control were added to the bottom chamber. Chambers were incubated for 18 h at 37 °C, and filters were fixed and stained with Mayer’s hematoxylin. Cells remaining on the upper surface of the membrane were removed using a cotton-tipped applicator, and the filter was mounted using Cytoseal 60®. Cells that had migrated to the lower surface of the membrane were counted microscopically by a blinded independent observer.
Statistical methods
Statistical significance of differences between data was determined using ANOVA or t-test models as indicated. Linear regression analysis was performed to identify parameters affecting the percent change in cells within a population after being subjected to a specific drug. Regression analysis and Tukey’s post hoc tests were completed using R version 3.0.2 (Vienna, Austria).
Discussion
Drivers in the recurrence of CRC include the chemokine receptor CXCR4, its ligand CXCL12, and the ectoenzyme CD26 for which CXCL12 is a substrate [
35]. We find that this regulatory framework is altered in the population of surviving cells following incomplete eradication by chemotherapeutic agents.
A diverse range of established cytotoxic agents share the ability to down-regulate CXCR4 but up-regulate CD26, in this case on five different CRC cell lines (T84, HT-29, HRT-18, SW480, SW620). Furthermore, suppression of the CXCL12-CXCR4 axis due to loss of CXCR4 is potentially accentuated by this concomitant up-regulation of CD26, because its DPPIV activity degrades CXCL12 and there is therefore a parallel reduction in the bioactivity of the chemokine ligand as well as the decline in its receptor. These effects on CXCR4 and CD26 were independent of the cytotoxic mechanisms of action of these drugs, which acted through DNA cross-linking (Cis, Ox), inhibition of DNA synthesis (5-FU, MTX), microtubule disruption (Vin) or topoisomerase inhibition (IT; SN-38). The ability of agents (e.g. MTX, VB) that are not components of current cytotoxic regimens may allow them to be added in late-stage disease to manipulate CXCR4/CD26 levels without contributing to existing dose-limiting toxicities.
Loss of the average CXCR4 expression of the cell population was due to selective elimination of CXCR4+ cells. The same diverse collection of agents caused an increase in the abundance of CD26 at the CRC cell surface after treatment. In this case the change in CD26 did not result from selection toward a greater proportion of CD26+ cells, but an overall increase in CD26 net cell-surface expression. This was found to be (i) an authentic increase on viable cells, (ii) an increase in fully-functional CD26 (able to bind ecto-ADA and with DPPIV activity), (iii) independent of ambient levels of CXCL12, and (iv) regulated separately from DPPIV activity (as was the decrease in CXCR4). (This phenomenon of chemotherapy-induced CD26 elevation was also reproducible in four other, non-CRC carcinoma cell lines (LNCaP and PC-3; prostate, A549; lung, and T-47D; breast) and a non-epithelial cancer cell line (SH-SY5Y; neuroblastoma); data not shown).
The mechanism of chemotherapeutic drug-induced elevation of CD26 clearly differs from the simple selective process of CXCR4 and remains under investigation. For most drugs the EC
50 for CD26 elevation did not differ significantly from that for reduction of CXCR4, although for cisplatin the concentration required to elevate CD26 was 10-fold higher than that required to reduce CXCR4
+ cells, indicating that the mechanisms are not linked in a consistent manner. Furthermore, the maximum decrease in CXCR4 was essentially the same for all drugs at around ~ 80 % (Additional file
5: Table S1), consistent with a cell selection process in which the outcome is defined by the survival of CXCR4+ cells. However, the maximum achievable elevation for CD26 varied between 22 % and 72 %, revealing differences in how this outcome is signalled. These two extremes in fact corresponded to the two agents – Cis and Ox – that are the most similar in their cytotoxic mode of action of all the major drugs studied, suggesting a distinction from pathways that lead to cell death. It was also noted that IT was more able to elevate CD26 than would be expected if compared with its active metabolite SN-38. The data overall point to a substantially different pathway by which these agents up-regulate CD26, albeit one that is initiated by many cytotoxic agents. It is worth noting that general approaches to kill CRC cells by nonspecific means (exposure to pH 8.6, 300 mM NaCl or 0.01 % w/v deoxycholate) neither elevated CD26 nor selected for an altered CXCR4+ subpopulation (data not shown).
HT-29 cells that had survived treatment with chemotherapeutic agents at modestly cytotoxic concentrations were fully proficient in terms of their capacity for movement at 48 h after treatment - basal migration did not differ from that of untreated cells. However, the directed migration toward CXCL12 was completely ablated. We attribute this loss of migration towards CXCL12 to the decline in CXCR4 and concomitant increase in DPPIV activity through gain of CD26. Comparable treatment with these drugs had no effect on the (low) levels of the alternate CXCL12 receptor CXCR7 expressed at the surface of human CRC cell lines (data not shown).
Outgrowth of residual disease has been attributed to a subpopulation of drug-resistant CICs capable of differentiating into the different cellular hierarchies that make up the overall tumor bulk [
11,
42]. Chronic treatment of CRC cell lines with 5-FU or Ox enriches for CIC markers CD44 and CD133 [
43], and irinotecan-treated xenografts show a greater frequency of CD44
+ cells [
44]. This implicates both CD44 and CD133 as markers for the putative CRC CIC subset. Although classically recognized as a marker of differentiation [
40,
45], CD26 expression has recently been reported to define a metastatic subpopulation of CICs within CRC and associated with development of metastasis in CRC patients [
14,
21]. The paradox of CD26 being both a differentiation marker and correlate of cancer aggression likely reflects its multiple roles [
15] and the consequence that its impact on cell function is context dependent.
Using primary colon cancer cells 5-FU has been reported to increase the CD26
+/CD133
+ subpopulation [
14]. This is consistent with our finding that CD26 is increased across all cell subpopulations after 5-FU treatment. We do however extend the observation to show that many agents increase CD26 levels on CRC cells, and report for the first time that treatment with the agents currently used to treat CRC favors the emergence of a CD26
+/CD44
+ cell population. This is an intriguing finding in that there is increased co-expression of two markers both interact with the extracellular matrix and might be able to cooperate in the process of metastasis associated with CD26 identified by Pang and colleagues [
14]. CD26 is known to associate with both fibronectin and collagen [
46,
47] while CD44 is a receptor for hyaluronic acid (HA) and functions in cell adhesion and tumorigenicity [
48,
49].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MJC performed flow cytometry and migration experiments and drafted the manuscript. ELL and CLR performed binding, migration, and xenograft experiments. DMH performed statistical modelling. PAS participated in experimental design. JB coordinated the research, participated in experimental design and edited the manuscript. All authors read and approved the final manuscript.