A Unique Nonsaccharide Mimetic of Heparin Hexasaccharide Inhibits Colon Cancer Stem Cells via p38 MAP Kinase Activation

Stem cells show a remarkable ability to self-renew and differentiate and replenish cells of a particular tissue that are lost as part of the natural processes or due to injury (1). Accumulating evidence suggests that colorectal cancers arise as a result of accumulating mutations in tissue-resident stem cells, which then transform into cancer stem cells (CSC; ref. 2). Hence, many of the self-renewal signaling pathways, e.g., Wnt–β-catenin–TCF4 signaling, are shared between normal stem cells (NSC) and CSCs (3). Increasing evidence points to a critical role of CSCs in driving cancer metastasis, resistance to chemotherapy, and relapse following complete tumor resections leading to poor clinical outcomes (4). As a result, intensive efforts are being directed at discovering therapies that selectively target CSCs. Most such therapies have focused on targeting one of the transcriptional signaling that regulates CSC self-renewal, e.g., β-catenin, SHH, etc. Given the redundancy in the regulatory network of CSCs, the approach may have limited success while also raising concerns for toxicity to NSCs due to a shared transcriptional mechanism of self-renewal (5). Hence, there is an urgent need to identify upstream pathways that differentially regulate CSCs and NSC growth. Likewise, agents that modulate these pathways to effect selective targeting of CSCs are also critically needed.

Arguably, microenvironmental components that serve as the stem cell niche have shown key differences with respect to NSCs and CSCs in many instances (6, 7). Among the many cellular and molecular components of the microenvironment, glycosaminoglycans (GAG) have emerged as essential regulators of stemness (8). GAGs, which are part of the cell surface macromolecules called proteoglycans, have been found to induce precise and coordinated modulation of key growth factors, cytokines and morphogens, resulting in selective MAPK and/or another intracellular signaling (9). This in turn is known to regulate self-renewal (10, 11). In fact, we recently demonstrated that a heparan sulfate (HS) sequence that is hexasaccharide (HS06) in length, but neither longer nor shorter than that, inhibits CSC self-renewal by isoform-specific activation of p38 MAPK (11). Interestingly, p38 MAPK was recently shown to promote differentiation of intestinal NSCs (12). These results provide a clue that HS06, or its mimetics, which induce p38 activation could prove to be valuable as clinically viable anti-CSC agents.

For a long time, it has been hypothesized that GAGs would inhibit cancer (13, 14). In fact, polymeric heparin and variants thereof, e.g., 2-O,3-O desulfated heparin (ODSH) as well as oligosaccharide mixture PI-88, have been examined in cancer treatment (15–18). Heparin, PI-88, and other GAGs are highly heterogeneous, which generates major barrier for target selectivity. Problems of selectivity also arise from the primarily electrostatic nature of their interactions with proteins, which disfavor hydrogen bonding and hydrophobic (van der Waals) interactions. We have shown that sulfated nonsaccharide GAG mimetics (NSGM), which are fully synthetic and homogeneous, bind to GAG-binding proteins through electrostatic, hydrogen bonding, and van der Waals forces, thereby exhibiting much higher target selectivity (19–21). NSGMs also offer several other key advantages over GAGs as viable anticancer agents. These include ease and scalability of synthesis, ease of monitoring homogeneity of drug dose, and inexpensive cost of large-scale preparation. In fact, we recently discovered that an NSGM called G2.2, which selectively inhibits CSC self-renewal in vitro (22), is a structural mimetic of HS06 (23). Hence, we hypothesized that G2.2 can inhibit CSC self-renewal, but promote NSC differentiation, in vivo by inducing p38 MPAK activation.

We show here that G2.2 inhibited colon CSCs-induced xenografts that were further enriched in CSCs through prior treatment with oxaliplatin and 5-fluorouracil, the most commonly used colorectal cancer chemotherapy combination, in a dose-dependent fashion. G2.2-induced p38α/β MAPK was required for former's anti-CSC effects in vitro as well as in vivo. Moreover, G2.2 demonstrated no gross toxicities or vital organ damage and had minimal anticoagulant effects. Overall, the studies indicate that activation of p38α/β MAPK might serve as a CSCs-suppressive signaling, and G2.2, an activator of p38α/β MAPK, represents a novel selective anti-CSC therapy with a significant translational potential.

Synthesis of G2.2 (purity 99%) was performed as described earlier (22). SB203580 (Selleckchem; ref. 24), as well as 5-flurouracil, oxaliplatin, Tween 80, dimethyl sulfoxide, and PEG300 (Sigma Aldrich) were purchased from commercial vendors.

Human colorectal (HT29 and HCT116) and pancreatic (Panc-1) cells were obtained from the ATCC. MDA-MB-231 cell line was a gift from Dr. Kolblinski (Virginia Commonwealth University). The cells were passaged as monolayers (11, 22, 24) for no more than six passages. The cells were tested for mycoplasma with the kit (ATCC 30-1012K) and disinfected with MPbio (San Diego) catalog # 093050044 within last 12 months. HT29 cells were transfected using mammalian p38α (Dharmacon # M-003512-02-0005) smart pool as well individual p38α siRNA (Dharmacon # D-003512-15 and D-003512-19; 25 nmol/L final concentration) with dharmafect duo reagent in a 6-well plate using manufacture's transfection protocol and plated for spheroid assay 48 hours later.

All experiments involving animals were approved by the Animal Component of Research Protocol Committee at Richmond VAMC.

Xenografts were generated by injecting 10⁵ CD133 hi/CXCR4 hi (Dual hi) fluorescence-activated cell sorter (FACS) isolated HT-29/HCT 116 cells suspended in 50% reduced growth factor Matrigel (BD Bioscience; in 50 μL sterile PBS) into the right flank of 6-week-old, female NCr nude mice (Taconic Farms) subcutaneously. Once the average tumor volume reached 25 to 50 mm³ (∼day 13), animals were randomly assigned to respective treatments for the defined time.

Each group (N = 16) was injected with 2,000 to 100,000 (N = 4 mice/cell concentration) of CD133⁺/CXCR4⁺ (Dual hi) cells from HCT 116 or HT-29 subcutaneously. The animals were then observed for the presence of palpable tumors >150 mm³.

Groups of animals (N = 8/group) were injected i.p. with 200 μL saline or G2.2 (10→200 mg/kg) × 3 times a week for 5 weeks. Following treatment, mice were sacrificed at day 43 (1 week after the last injection) and xenografts used for ex vivo CSC phenotype studies.

Xenografts excised from mice treated with 100 mg/kg were finely chopped and digested as described below. Single-cell suspension was obtained, and 10⁶ cells were injected s.c. Once palpable tumors were observed, these were randomized (N = 5) to receive i.p. injections of 200 μL saline or G2.2 (100 mg/kg) × 3 times a week for 3 weeks.

Mice were injected with 10⁵ Dual hi cells as described above. Groups of animals (N = 5/group) were injected i.p. with 200 μL saline or G2.2 (175 mg/kg) x3 times a week for 3 weeks. Following treatment, mice were sacrificed an hour after the final injection and xenografts used for ex vivo CSC phenotype studies.

In the first phase, a group of 6-week-old, female NCr nude mice were randomized to vehicle (N = 10) or FUOX (5-FU 25 mg/kg and oxaliplatin 2 mg/kg weekly; n = 17) for 3 weeks followed by the second randomization of FUOX-treated animals (D-21 of initiation of FUOX) to vehicle (n = 7) or G2.2 (200 mg/kg 3 times a week x 5 weeks; n = 7). Mice were euthanized (1) after chemotherapy (day 21), (2) at day 30, and (3) at day 42 after completion of treatment (an hour after the final injection) and ex vivo CSC phenotype studies performed.

Animals were randomly assigned to four groups for mechanistic studies that were injected i.p. (1) Vehicle (Veh), (2) G2.2 (200 mg/kg), (3) SB203580 (SB; 10 mg/kg), and (4) (SG) SB, 3 hours prior to G2.2 (200 mg/kg) 3 times a week for 5 weeks. Following treatment (an hour after the final injection), mice were sacrificed and xenografts used for ex vivo CSC phenotype studies.

Tumor measurements were made 3 times a week with Vernier calipers, and tumor volume was calculated using the formula: V = W² × (L)/2, where V is the volume in mm³, and W and L are the width and length in mm. At the end of the 5 weeks after treatment, appropriate numbers of animals were sacrificed in each group, and the tumor tissues were collected and processed as below. The remainder of animals was monitored till they reached predefined humane end-points.

Animals were sacrificed per Institutional Animal Care and Use Committee–approved methods of euthanasia. The tumor tissue was finely chopped and digested with 400 μg/mL Collagenase Type IV (STEMCELL Technologies). Single-cell suspension was filtered with 70 μm cell strainer and used for one of the studies described below. The vital organs were harvested and fixed in 4% paraformaldehyde and sectioned using microtome. The slides were stained with H&E and examined under light microscope.

Flow cytometric analyses for CSC markers as well as Western-blot analyses for CSC markers, self-renewal factors, and p38 MAPK were performed using methods described in earlier publications (11, 22, 24). The details of antibodies are provided in Supplementary Methods.

Cells derived from xenografts as well as primary cells maintained as monolayer were plated in nontreated, low-adhesion, 96-well plates for 1/2/3/4 spheroids as described earlier (24).

Serum chemistry was performed by the mouse phenotyping, physiology, and metabolism core at the Penn diabetes research center located at the University of Pennsylvania health system.

PT and aPTT were measured in plasma using standard one-stage clotting assays (STA PT-Neoplastin CI, STA PTT- Automate, respectively) on the STA Compact analyzer (Diagnostica Stago) according to the manufacturer's instructions.

Mice intestinal pieces (2–4 cm) were subjected to chelation and dissociation into crypts using previously described methods (25). Approximately 200 to 500 crypts were then resuspended in 100 μL of Matrigel per well of a 96-well plate and overlayed with 100 μL of IntestiCult Organoid Growth Medium (Mouse; Stemcell technologies) after allowing for Matrigel polymerization and cultured in the CO₂ incubator (37°C, 5% CO₂). The intestinal organoids develop after 5 to 7 days of culturing.

Briefly, a single-cell suspension of mononuclear cells from mouse bone marrow was prepared to obtain approximately 2 to 4 × 10⁷ hematopoietic cells using previous described methods (26). The cells were resuspended in 10 mL of Iscove's modified Dulbecco's medium (2% FBS media; Allcells) and mixed with Methyl cellulose (R&D systems), and 1.1 mL of the final cell mixture was added to a nontreated 6-well plate using a 3 mL syringe. Water was added to one of the wells of the 6-well plate to maintain humidity necessary for colony development. Plates were incubated for 8 to 12 days at 37°C and 5% CO₂. Colonies consisting of at least 30 cells were counted.

A high percentage of CD133⁺/CXCR4⁺ (Dual hi) in primary human primary colorectal cancer was associated with a poor prognosis (25). Earlier, we observed significantly increased spheroid formation using FACS Dual hi HT-29 cells compared with Dual lo (CD133⁻/CXCR4⁻) controls (26), suggesting enrichment of CSC population. Furthermore, magnetic-assisted cell sorting (MACS) experiments showed that only Dual hi but not CD133⁺/CXCR4⁻, CD133⁻/CXCR4⁺ HT-29 cells showed robust increase in 1° spheroid formation compared with Dual lo cells (Fig. 1A). Indeed, Dual hi cells showed robust overexpression of not only CD133 and CXCR4, but also DCMKL1, LGR5, EpCAM, LRIG1, and NANOG mRNA (Fig. 1B) and/or protein levels (Fig. 1C), other bona fide makers of colon CSCs, compared with Dual lo controls in two colon cancer models, HT-29 (KRAS wt., P53 mut.) and HCT 116 (KRAS mut., and p53 wt.) that differ in common colorectal cancer genetic variant status. However, an accurate measure of CSCs can only be judged through in vivo limiting dilution assay. Indeed, both HT-29 and HCT 116 cells showed tumor formation (size >150 mm³) in ≥50% of the animals with as few as 2 × 10e3 cells (Fig. 1D). On the other hand, a similar number of Dual lo cells failed to form any tumor with either of the cells for >35 days following injection (Supplementary Fig. S1A). As 10e5 Dual hi cells generated xenografts consistently (100%) and rapidly in both cell line models (Fig. 1E; Supplementary Fig. S1B), we used that cell dose for xenograft formation in the efficacy studies of G2.2 below.

Selective targeting of CSCs is a paradigm-shifting approach. G2.2, an NSGM of HS06 (23), was identified earlier as a potent and selective CSC inhibitor using a novel in vitro tandem, dual screen strategy (22). Buoyed by this advancement, we embarked on the in vivo evaluation of its therapeutic potential. HT-29 xenografts were induced by injecting 10e5 Dual hi cells s.c. in a group of 20 NCr nude mice. Mice were randomized to either vehicle or one of the doses of G2.2 (25→200 mg/kg, 3×/week for 5 weeks) delivered i.p. after palpable tumors formed. We observed a dose-dependent inhibition of tumor volume in G2.2-treated animals compared with vehicle controls with maximal potency observed at doses of 200 mg/kg (Fig. 2A; Supplementary Fig. S2A) without any gross toxic effects (see toxicity studies below). The tumor volumes at day 43 displayed a robust >75% decrease in the G2.2 (200 mg/kg)-treated mice compared with vehicle controls (Fig. 2A) resulting in significantly improved survival using humane endpoint criteria (Supplementary Fig. S2A). Using 200 mg/kg of G2.2 as an optimal dose, we performed additional studies to understand its effect on CSCs growth better. Comparable with significant tumor volume changes, we observed a robust reduction in a complement of CSC markers (CD133, DCLK1, LRIG1, and LGR5) as well as self-renewal factor (BMI-1; Fig. 2B). Similarly, there was approximately 5-fold and 3.5-fold reduction in the numbers of LGR5⁺ and Dual hi cells, respectively, in G2.2-treated xenografts compared with vehicle control (Fig. 2C; Supplementary Fig. S2B). Indeed, a similar reduction (4.5-fold) in Dual hi cells was also observed in HT-29 spheroids treated with G2.2 (100 μmol/L; Supplementary Fig. S2B), confirming inhibition of Dual hi CSCs by G2.2 in vitro and in vivo. We utilized our earlier observation that CSCs/progenitors are enhanced several-fold in spheroid culture and assessed the xenografts-derived cells for retention of 2°→4° spheroid growth inhibition profiles to distinguish between nonself-renewing progenitors and self-renewing CSCs ex vivo (22, 27). Indeed, cells derived from G2.2-treated xenografts showed a robust 7-fold decrease in 2°→4° spheroid formation, measured a week (day 43) after the last dose of G2.2, compared with vehicle controls (Fig. 2D; Supplementary Fig. S2C). More importantly, akin to our in vitro studies (22), we observed a robust induction of apoptosis in G2.2-treated xenografts compared with vehicle controls (Supplementary Fig. S2D), which was mainly restricted to Dual hi (50%) compartment with a minor effect in Dual lo (10%) compartment, suggesting selective targeting of CSCs in vivo by G2.2 (Fig. 2E; Supplementary Fig. S2D). Hence, G2.2 seems to be a first of its kind, novel NSGM that inhibits tumor growth by CSCs’ inhibition-dependent mechanism in vivo.

Despite robust growth inhibition of HT-29 xenograft by G2.2, residual tumors remain (Fig. 2A). It is essential to determine if rapid resistance to therapy develops in vivo. We conducted additional studies on G2.2 sensitivity in residual G2.2-treated xenograft-derived cells in vitro (200 mg/kg cohort) and in vivo (100 mg/kg cohort). Post–G2.2 (200 mg/kg)-treated xenograft-derived 3° spheroids were treated with G2.2 (100 μmol/L) or vehicle and observed for 4° spheroids (presence of the drug) as well as 5° and 6° spheroids (absence of further G2.2 treatment). Indeed, G2.2 showed robust inhibition of 4°→6° spheroids (Supplementary Fig. S2D). Furthermore, G2.2 (100 mg/kg 3 times/week × 3 weeks i.p.) showed potent inhibition of secondary xenografts (>50%) generated from post–G2.2 (100 mg/kg)-treated xenograft-derived cells in NCr nude mice (Fig. 2F) as well as ex vivo CSCs’ self-renewal (2°→4° spheroids; Supplementary Fig. S2F) in cells derived from these secondary xenografts compared with respective vehicle-treated controls. Hence, G2.2 treatment does not induce rapid development of treatment resistance in vivo. Finally, G2.2 (175 mg/kg 3 times/week × 3 weeks i.p.) showed robust inhibition of another Dual hi–generated colon cancer xenograft (HCT 116) growth as well as ex vivo CSC phenotype (Fig. 2G; Supplementary Fig. S2G and S2H) as in HT-29 cells, suggesting likely generalizable effect of G2.2 on colon CSCs.

Our subsequent study was designed to understand the efficacy of G2.2 in a CSC-enriched xenograft model. We utilized a dual enrichment strategy. In the first step, we performed in vitro enrichment, as described in the above model, by developing Dual hi HT29 cell–derived xenografts in nude mice. Combination of 5-fluorouracil and oxaliplatin (FUOX), which is the most frequently used chemotherapy for colon cancer, enriches CSCs in vitro (28). Hence, we performed a second step of CSC enrichment in vivo. This involved randomization of HT29 Dual hi–induced xenograft-bearing nude mice to i.p. administration of FUOX or vehicle weekly for 3 weeks. As expected, the dual enrichment strategy resulted in a further increase in the Dual hi cells as well as 2°→3° spheroids in chemotherapy-treated mice (11% vs. 6%) compared with vehicle controls (Fig. 3A; Supplementary Fig. S3A and S3B). The overall lower CSC population in the xenograft (than at the time of original inoculation) can be attributed to the ability of CSCs to proliferate into non-CSC progenitor cells (29), as well as self-renew to maintain its own population in vivo postinoculation. The chemotherapy-treated CSC-enriched xenograft-bearing mice were then randomized (second randomization) to receive either G2.2 (200 mg/kg 3×/week × 10 injections i.p.) or vehicle. Hence, the study had three groups including (1) vehicle (Veh); (2) chemotherapy followed by vehicle (CVeh), and (3) chemotherapy followed by G2.2 (CG2.2). The results showed that there was 2-fold rapid increase in tumor volume in the CVeh mice compared with Veh (Fig. 3A) accompanied by a progressive increase in Dual hi population (Fig. 3C), which most probably arose from enhancement in CSC-mediated proliferation due to our dual enrichment strategy. More importantly, we observed a robust 7.6-fold (P < 0.01) inhibition in tumor volume (day 30; Fig. 3A) resulting in improved survival using humane endpoint criteria (Fig. 3B). The tumor volume changes were accompanied by a robust 4.3-fold reduction in the fraction of Dual hi cells (Fig. 3D; Supplementary Fig. S3C) as well as a complement of CSC makers/self-renewal factor (CD133, DCLK1, LRIG1, LGR5, BMI1) in CG2.2 xenografts compared with CVeh (Fig. 3C; Supplementary Fig. S3D). Indeed, phenotypic CSC self-renewal studies with 2°→3° colonospheres (at both days 30 and 42 after treatment initiation) showed a similar profile as the change in tumor volume, characterized by an increase in CVeh, but a robust decrease in CG2.2-treated animals compared with Veh controls (Fig. 3E; Supplementary Fig. S3E). Overall, the data validate the in vivo efficacy of G2.2 in inhibiting chemotherapy-enriched CSC xenograft growth model.

As stated before, G2.2 is a structural mimetic of the natural HS06 sequence, as demonstrated in a recent study (23). We also reported that HS06 selectively inhibits CSC self-renewal through an isoform-specific activation of p38α MAPK (11). Thus, our expectation was that G2.2 would also induce activation of p38 MAPK. Indeed, like HS06, G2.2 induced early and sustained activation of p38 MAPK, but caused inhibition of ERK1/2 and JNK—the other related MAPK members (Supplementary Fig. S4A), in HT29 spheroids. Interestingly, inactive analogs of G2.2, G1.4, and G4.1 failed to activate p38 (Supplementary Fig. S4B) in HT29 spheroids. Measurement of phosphorylation level of p38 in colorectal (HCT116 and HT29), pancreatic (Panc-1), and breast (MDA-MB-231) cancer spheroids (enriched in CSCs) also revealed that G2.2 enhanced p38 activation nearly 1.5 to 2.7-fold. At the same time, p38 MAPK activation was not found in the corresponding monolayer counterparts (Fig. 4A), suggesting selectivity of G2.2 toward CSCs.

We performed immunoprecipitation with anti-pp38 antibody followed by Western blotting with isoform-specific p38 antibodies following vehicle- or G2.2 treatment in HT29 spheroids to elucidate isoform specificity of p38 induction. G2.2 treatment resulted in activation of α and β isoforms (α > β), inhibition of the δ isoform, and no discernible effect on the γ isoform (Supplementary Fig. S4C). The results indicated isoform-specific activation of p38 MAPK by G2.2. Importantly, pretreatment with SB203580 (SB), a pharmacologic inhibitor of p38α/β (24), mostly resulted in a near-complete reversal of G2.2-mediated inhibition of CSC self-renewal as evident in 3° spheroids formation (Fig. 4B) as well as CSC markers (CD44, EPCAM) and self-renewal (BMI-1) factors (Fig. 4C). These results imply that p38α/β MAPK activation mediates the effects of G2.2 on CSC self-renewal. Furthermore, genetic depletion of p38α (siRNA) in HT-29 spheroids (Fig. 4D; Supplementary Fig. S4D and S4E), as well as function p38α inhibition using a dominant-negative vector (p38α agf) in both HT-29 and HCT 116 spheroids (Fig. 4E and F; Supplementary Fig. S4F), produced results identical to SB203580. Overall, these findings point to a highly specific modulation of p38 MAPK in the regulation of CSCs’ self-renewal by G2.2. Moreover, these findings are virtually identical to the effects observed with HS06 strongly implying functional mimicry of HS06 by G2.2 (11).

To test p38α/β activation-based anti-CSC mechanism of G2.2 in vivo, we performed mice xenograft experiments similar to that described in Fig. 1 using G2.2 (inducer) and SB (inhibitor) as p38α/β modulators. Mice induced with 1 × 10⁵ Dual hi HT29 cell xenografts were randomized to i.p. administration of (1) vehicle (Veh), (2) G2.2 (200 mg/kg); (3) SB (10 mg/kg), and (4) SB 3 hours prior to G2.2 (SG) 3 times a week for 3 weeks. As expected, G2.2 displayed a substantial decrease in tumor volume (4.6-fold) when compared with vehicle (Fig. 5A), which was also corroborated by a similar reduction in 3° spheroids (Fig. 5B), Dual hi CSC population (Fig. 5C), and levels of CSC markers (CD133, LGR5) and self-renewal (BMI1) factor (Fig. 5D). Intriguingly, SB alone caused a modest decrease in tumor volume (Fig. 5A). The latter observation can be attributed to a chemotherapy-like targeting of non-CSCs as we observed enrichment of Dual hi CSCs, increase in 2° spheroids, and levels of CSC/self-renewal markers (Fig. 5B–D). Indeed, as observed in the in vitro studies, administration of SB followed by G2.2 (SG) produced near-complete reversal of G2.2′s effect on tumor volume, 2° spheroids, Dual hi population, and CSC/self-renewal makers (Fig. 5A–D; Supplementary Fig. S5A). Indeed, G2.2 caused activation, SB induced inhibition, and SG caused no change in levels of pp38 (the activated form of p38 MAPK; Fig. 5D). These findings support the conclusion that activation of p38 α/β largely drives in vivo anti-CSC effects of G2.2.

Given the robust anti-CSC efficacy of G2.2 in vivo, we proceeded to elucidate the toxicity profile in vivo to assess its true therapeutic potential. The toxicity was determined at different levels including (1) gross, (2) vital organ damage, and (3) progenitor cell growth. We also studied its effect on coagulation as GAGs and NSGMs may inhibit coagulation enzymes (30, 31). At a gross level, i.p. administration of G2.2 (200 mg/kg, 3 times/week × 5 weeks) in the above studies did not produce any significant weight loss in mice (Supplementary Fig. S6A). Mice showed normal behavior with no external signs of distress, allergy-induced rashes, diarrhea, or significant deviation from expected repertoire throughout the course of the treatment. A thorough assessment of critical vital organ damage was conducted by measuring (1) morphology (H&E) and (2) serum chemistry in G2.2- and vehicle-treated mice. No significant changes in tissue morphology were observed in G2.2-, compared with vehicle-, treated mice organs (Fig. 6A) except in the liver. The liver in G2.2-treated mice showed a minor 5% dropout in hepatocytes (Supplementary Fig. S6B). However, serum chemistry revealed no significant alterations in various electrolyte levels or biomarkers indicating hepatic (e.g., Ast, Alt), renal (creatinine), and muscle damage (Cpk; Fig. 6B). Hence, we conclude that the changes in the liver morphology are unlikely to be of any major clinical significance.

GAGs are known to interact with many proteins of the coagulation system. Hence, we studied the anticoagulation potential of G2.2 through APTT and PT studies, which are routinely used to assess the anticoagulation state of blood/plasma. The APTTs of plasma collected from G2.2- compared with vehicle-treated mice were found to be substantially similar, whereas a modest delay in the extrinsic clotting time (PT) was observed for G2.2-treated plasma (Fig. 6C).

As CSCs and normal stem/progenitors (NSCs) utilize common pathways of self-renewal, it was essential to examine G2.2′s effect on the NSCs. The NSC population in the bone marrow and the intestines are abundant and differentiate continuously to fulfill functions of blood cells while also replenishing the surface lining in the bone marrow and the intestines (32, 33). We first examined the proportion of LGR5⁺ cells in the colonic mucosa using flow cytometry. Intriguingly, G2.2 had no discernible effects on the number of adult colonic NSCs (Fig. 6D). To further understand the effects of G2.2 on the proliferative function of NSCs, we examined ex vivo intestinal organoids and bone marrow colony formation in G2.2- and vehicle-treated nude mice by harvesting the respective organs, as reported in the literature (34, 35). Chronic administration of G2.2 (200–400 mg/kg) had none to a minimal inhibitory effect on small intestinal and colonic organoid (1°/2) formation or bone marrow–mixed colony formation compared with vehicle control (Fig. 6E). The results suggested no untoward effect on the function of the intestine and bone marrow NSCs/progenitors.

Thus overall, G2.2 exhibited minimal systemic effects in nude mice including preserving adult colonic/marrow NSC function while exhibiting potent anti-CSC effects toward colon CSCs. These findings support G2.2 as a potential therapeutic agent.

GAGs have been known to play essential roles in several cancer-related processes including fine-tuning of growth factor receptor signaling, tumor angiogenesis, and metastasis (9, 13, 14). In fact, we recently presented evidence that a heparan sulfate hexasaccharide (HS06) sequence, but not any other longer or shorter sequence including UFH, LMWH, ODSH, and fondaparinux, has the unique ability to selectively inhibit CSCs from a variety of organ types (11). HS06 is particularly useful because it carries minimal anticoagulant potential as compared with UFH, LMWH, and fondaparinux. More importantly, the delicate chain-length dependence of anti-CSC activity suggests finely tuned biology that may be amenable for therapeutic targeting.

Developing HS06 as an anti-CSC agent is challenging. HS06 synthesis or isolation from UFH is an incredibly painstaking as well as an expensive proposition. Instead, it might be advisable to develop HS06 mimetics that are easier to synthesize and purify. Hence, we resorted to the approach of studying NSGMs, which are not only easy to synthesize but also structurally homogeneous and likely to be more selective in protein recognition because of their hydrophobic aromatic scaffold. Hence, the discovery of G2.2, a structural mimetic of HS06 (23), as a robust and selective anti-CSC agent is a significant advancement in developing GAG-based anticancer agents. In fact, relapse following chemotherapy is generally attributed to the enrichment of CSCs after chemotherapy (36, 37). To this end, our studies provide the proof-of-the-concept that inhibition of CSCs is a viable strategy to halt tumor regrowth after chemotherapy treatment.

Polymeric GAGs are known to interact with a plethora of proteins to regulate many pathophysiologic responses (14, 26, 38–42). G2.2 (1,701 Da), in contrast, is much smaller and more homogenous than polymeric GAGs (15,000–50,000 Da). Despite these advantages, one could theoretically expect G2.2 to interact with several GAG-binding proteins in the manner of polymeric GAGs. Thus, G2.2′s toxicity profile was essential to characterize. In combined morphologic and biochemical studies of vital organ functions, G2.2 demonstrated excellent tolerance and lack of significant toxicity at anticancer doses. Concerning its anticoagulation profile, there was a modest increase in PT (INR < 2) but no significant effect on APTT with chronic administration of G2.2 at 200 mg/kg. Importantly, the latter was not associated with any major bleeding. Despite these positive results, it is important to note that toxicity studies were conducted in immunocompromised mice. It will be important to perform similar toxicity studies in immunocompetent mice in the future as GAGs are known to modulate immune function in vivo (43).

Although G2.2 robustly inhibited cancerous intestinal stem cells, it has no apparent ill-effects toward normal intestinal stem cell function. This property of G2.2, selective targeting of CSCs, while sparing NSCs, is highly desirable in a selective anti-CSC agent. Often, the therapies that are developed to target CSCs, e.g., Wnt−β-catenin, notch, and hedgehog inhibitors, are likely to affect NSC function as many of these pathways are shared between CSCs and NSCs albeit with different degrees of dependence. In this respect, G2.2 is likely to be a unique agent. It appears to be inducing a common singling that differentially regulates CSCs and NSCs, thus providing a greater therapeutic window with respect to toxicity toward NSCs. Can p38 MAPK activation as the common signaling achieve such differential response? Although this question remains to be fully answered, the data discussed below support the latter hypothesis.

Mammalian p38 MAPKs are activated in the wide range of stimuli, ranging from physiologic processes such as cell differentiation to pathologic states including cancer. Activation of p38 MAPK has been shown to display often opposing roles in different organs, cell types, and pathophysiologic conditions (10). In fact, the tumor-suppressor role of p38α/β in various cancers has been reported. More recently, others and we have highlighted the importance of p38 activation in suppression of the CSC self-renewal (11, 44). Intriguingly, recent reports supporting the role of p38 activation in promoting NSC differentiation in the intestine (12) and/or bone marrow (45) have come to light. In the future, a differential role of G2.2 against CSCs and NSCs will have to be assessed in mice bearing murine colonic tumors using chemically induced or genetic models of colon cancer. In fact, such studies should ideally be performed in LGR5 reporter mice as one will be able to visualize the effect of the molecule on both CSCs and NSCs.

In conclusion, our studies offer a paradigm-shifting approach of selectively targeting CSCs to prevent tumor regrowth following traditional chemotherapy that often enriches CSCs. In addition, our mechanistic studies bring to the light importance of p38α/β signaling as a therapeutic target to achieve a degree of selectivity toward CSCs. Moreover, G2.2′s highly selective phenotypic properties bode well for NSGM technology to deliver additional promising therapeutic agents for cancer and other pathophysiologic conditions where GAGs play a key role.

B.B. Patel and U.R. Desai has ownership interest (including stock, patents, etc.) in U.S. Patent application 2016/0280676. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R.S. Boothello E.I. Abdelfadiel, U.R. Desai, B.B. Patel

Development of methodology: R.S. Boothello, C. Sharon, H.R. Lippman, B.B. Patel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.S. Boothello, C. Sharon, E.I. Abdelfadiel, S. Morla, D.F. Brophy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.S. Boothello, N.J. Patel, B.B. Patel

Writing, review, and/or revision of the manuscript: R.S. Boothello, E.I. Abdelfadiel, D.F. Brophy, U.R. Desai, B.B. Patel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.J. Patel, E.I. Abdelfadiel, H.R. Lippman, U.R. Desai, B.B. Patel

Study supervision: U.R. Desai, B.B. Patel

Other (performed experiments): C. Sharon

This work was supported in part by Veteran Affairs Merit Award 5I01BX000837 awarded to B.B. Patel; National Heart, Lung and Blood grants HL107152, HL090586, and HL128639 awarded to U.R. Desai; and Massey Cancer Center Pilot Project Fund A35365 to B.B. Patel and U.R. Desai. We also thank the computing resources made available through the S10RR027411 grant from the National Center for Research Resources to Virginia Commonwealth University. Services and products in support of the research project were generated by the VCU Massey Cancer Center Flow Cytometry Shared Resource and Cancer Mouse Model Shared Resource, supported, in part, with funding from NIH-NCI Cancer Center Support Grant P30 CA016059. Service and products for research were also generated by Penn Diabetes Research Center grant P30-DK19525 metabolism core.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.