Extracellular Proton Concentrations Impacts LN229 Glioblastoma Tumor Cell Fate via Differential Modulation of Surface Lipids

Background Glioblastoma multiforme (GBM) is a highly aggressive form of brain cancer with marginal survival rates. GBM extracellular acidosis can profoundly impact its cell fate heterogeneities and progression. However, the molecules and mechanisms that enable GBM tumor cells acid adaptation and consequent cell fate competencies are weakly understood. Since extracellular proton concentrations (pHe) directly intercept the tumor cell plasma membrane, surface lipids must play a crucial role in pHe-dependent tumor cell fate dynamics. Hence, a more detailed insight into the finely tuned pH-dependent modulation of surface lipids is required to generate strategies that can inhibit or surpass tumor cell acid adaptation, thereby forcing the eradication of heterogeneous oncogenic niches, without affecting the normal cells. Results By using image-based single cell analysis and physicochemical techniques, we made a small-scale survey of the effects of pH ranges (physiological: pHe 7.4, low: 6.2, and very low: 3.4) on LN229 glioblastoma cell line surface remodeling and analyzed the consequent cell fate heterogeneities with relevant molecular targets and behavioral assays. Through this basic study, we uncovered that the extracellular proton concentration (1) modulates surface cholesterol-driven cell fate dynamics and (2) induces ‘differential clustering’ of surface resident GM3 glycosphingolipid which together coordinates the proliferation, migration, survival, and death reprogramming via distinct effects on the tumor cell biomechanical homeostasis. A novel synergy of anti-GM3 antibody and cyclophilin A inhibitor was found to mimic the very low pHe-mediated GM3 supraclustered conformation that elevated the surface rigidity and mechano-remodeled the tumor cell into a differentiated phenotype which eventually succumbed to the anoikis type of cell death, thereby eradicating the tumorigenic niches. Conclusion and significance This work presents an initial insight into the physicochemical capacities of extracellular protons in the generation of glioblastoma tumor cell heterogeneities and cell death via the crucial interplay of surface lipids and their conformational changes. Hence, monitoring of proton–cholesterol–GM3 correlations in vivo through diagnostic imaging and in vitro in clinical samples may assist better tumor staging and prognosis. The emerged insights have further led to the translation of a ‘pH-dependent mechanisms of oncogenesis control’ into the surface targeted anti-GBM therapeutics.

Note that the treatment with Temozolomide (an FDA approved anti-glioma drug) and high concentrations of H 2 O 2 induced partial DNA laddering and nucleic acid smearing, hence served as the positive controls in this assay. B), C) and D) Shows expression profiles of apoptosis-associated proteins. Cleaved PARP1 was weakly expressed at pH 3.4, cleaved caspase-8 too showed lower expression at pH 3.4 in comparison to other pHs but cleaved caspase-3 showed higher expression at pH 3.4. However, the data overall did not show any appreciable and consistent trendlines at very low vs. low and physiological pH conditions, suggesting that cells were not progressing towards apoptosis. It is to be noted that cleaved caspases are associated with non-apoptotic functions too. Also note that cleaved caspases were observed to be sequestered near the inner surface of the plasma membrane at low and very low pHs, a localization that cannot induce apoptosis. The absence of tumor cell apoptosis at various pHe was also in concert with the lack of DNA fragmentation and nuclear pyknosis (Fig 2D) in all pH ranges. Besides, the positive response of tumor cells in live cellbased assays such as glucose uptake, gel contraction and macropinocytosis also confirmed 3 that the cells did not die at pH 3.4 in a timescale of 7-24 hours. Error Bar=S.D; No. of independent experimental replicates=3; *:p≤0.05, **: p≤0.01, ***:p≤0.001.

S3 Fig: Autophagy and Senescence marker analysis at different pHe in LN229 glioblastoma cells.
A) Beclin1 and B) LC3, two major autophagy associated proteins, did not show any consistent trendlines in cells exposed to different pHe. Beclin1 was reduced at very low pH whereas LC3 expression was enhanced. The mixed trendlines suggest that overall autophagy may be induced at all pH ranges and may assist in the survival of tumor cells. C) Shows that when senescence is induced by H 2 O 2 treatment (positive control), there was a high cytoplasmic accumulation of beta-galactosidase enzymatic product (bluish-green in color).
However, when tumor cells were exposed to various pH ranges, active enzyme localization was detected either on the plasma membrane or in acid organelles but such localizations are associated with the non-senescent condition. Error Bar=S.D; No. of independent experimental replicates=3; *:p≤0.05, **: p≤0.01, ***:p≤0.001. Necrotic Zone (pH <5.5-3.4), Pseudo-palisading cells layer (pH approx. 5.5 or less) and Cellular Tumor Zone (pH <7.0-6.2) from the same glioma tissue were compared for the expression of LAMP2 (an acidosis marker) and HMGCR (a rate limiting cholesterol synthesizing enzyme). Tabulated results showed that both proteins were highly expressed in necrotic and pseudo-palisading areas which are associated with very low pH. Moderate expression was also observed in Cellular Tumor Zones. The patient data was extracted from Human Protein Cancer Atlas (http://www.proteinatlas.org/). Necrotic Zone (pH <5.5-3.4), Pseudo-palisading cells layer (pH approx. 5.5 or less) and Tumor Core (pH <7.0-6.2) from the same glioma tissue were compared with respect to the expression of LAMP2 (an acidosis marker) and SREBF2 (a transcription factor that 4 upregulates the synthesis of HMGCR, a rate-limiting enzyme in cholesterol synthesis).

A), B)
Results showed that both LAMP2 and SREBF2 were highly expressed in necrotic and pseudo-palisading zones (which are associated with very low pH). Moderate expression was detected in Cellular Tumor regions. The patient data was extracted from Human Protein Cancer Atlas (http://www.proteinatlas.org/). p-values were represented with one star, 2 stars or 3 stars to denote p≤0.05, 0.01, 0.001 respectively.

S7 Fig: Staining, Intensity and Localization analysis of OCT4 in different glioma zones
Nuclear localization of Oct4 indicates proliferative and stem like properties of expressing cells. A) and B) Shows that necrotic and pseudo-palisading zone cells were low in nuclear and high in cytoplasmic expression for Oct4. C), D) Cellular Tumor Zone showed medium-5 high nuclear and medium cytoplasmic expression of Oct4. The patient data was extracted from Human Protein Cancer Atlas (http://www.proteinatlas.org/).

S8 Fig: Staining and Intensity analysis of GFAP in different glioma zones
GFAP is a cytoplasmic marker of astrocytes and is also identified as a membrane-associated cytosketetal protein. It is highly expressed in matured/differentiated astrocytes. A), B), C), D) and E) Shows that GFAP was strongly expressed in high-grade glioma and was found to be enriched in necrotic and peri-necrotic zones in comparison to cellular tumor zone (see panel F). Note that necrotic zone cells have a particularly high surface expression of GFAP. The patient data was extracted from Human Protein Cancer Atlas (http://www.proteinatlas.org/).

S9 Fig: Extracellular pH impacts cytoskeleton associated remodeling in a cholesterolsensitive manner
Rhodamine phalloidin binds to F-actin and identifies cortical, lamellipodial and stress fiber associated actin. α-actinin is a major F-actin associated protein and enable cell motility. A) Shows representative images of F-actin and α-actinin co-immunostained tumor cells exposed to various pHs. α-actinin was observed to penetrate well into the cortical actin areas in tumor cells exposed to physiological and low pHs (shown with white arrowheads). The actin stress fibers were diminished and severed in very low pH condition and showed prominent delocalization of α-actinin with cortical actin. B) Co-dynamics of surface beta-1 integrins and underlying alpha tubulins (palmitoylated) influence migration competencies. Cellular levels of alpha-tubulin along with respective SD's are indicated in each panel. Colocalization of beta-1 integrin and alpha-tubulin on the surface is indicated by Pearson's coefficient (R-value) and also highlighted with white arrows. Tumor cells at low pH in different conditions (pH 6.2 +veCD; pH 3.4 -veCD and +veCD) appeared to be supported by alpha tubulins although integrin expression was majorly reduced. This might be responsible for the migratory potential of tumor cells at low pH. C) Rac1, a small GTPase showed submembrane (white arrowheads) to nuclear localization in various pH microenvironments. High sub-membrane localization is associated with better migration competencies. This GM3 glycan tilt angle (Ɵ) denotes the extent of glycan bending to the membrane normal from the axis of the fatty acyl chains. It ranges from 0-90 ο . Torsion angles (φ) allow rotation of glycans on the acyl chain axis. It ranges from 0-360 ο .We did not find any significant difference in the torsion angles in our pH simulation studies; hence it was not further discussed in the main results. A) Shows LN229 glioblastoma cell line exposed to various pHs in -veCD-veHQ, +veCD and +veHQ conditions.Cells were fixed with 1.5% PFA at room temperature and then probed for surface GM3 clusters. Results show that the GM3 clustering was comparable to the experiment that was done at low temperature (live cells, on ice). B) Cells were treated with 0.5% saponin to disrupt lipid rafts (cholesterol and sphingolipid enriched domains). Results show that the GM3 surface clustering was dramatically reduced at all pHs upon saponin treatment, supporting the observations that cholesterol may stabilize GM3 surface ligations. A) and B) Shows that GM3 undergoes minimal lactonization in low pH conditions: Tumor surface GM3 signal was quantified in basal (-veCD) and high cholesterol (+veCD) conditions at various pH ranges. The neuraminidase treatments were performed on cells that were either kept on ice (cold treatment to prevent endocytosis of lipids concerned) or at room temperature. The GM3 signal detection was performed post fixation. Since GM3 lactonized forms are documented to be neuraminidase resistant, the major loss of GM3 signal from the surface, upon digestion of the neuraminic/sialic acid moiety, indicated that the observed GM3 clusters (as described in Fig10) were predominantly due to GM3-GM3 ligations and not due to the formation of GM3 lactones. Cleavage of the sialic acid moiety from GM3 converts this lipid into its precursor form known as lactosylceramide. C) and D) Shows an increase in surface lactosylceramide signal upon neuraminidase digestion with concomitant loss of GM3 signal in the same conditions (see A, B) which indicates that the extent of GM3 lactones formed on the cell surface in low pH condition contributed only minimally to GM3 clustered pattern. More specifically when neuraminidase treatment was performed at cold temperature, the estimated % of GM3 clusters due to lactonization was as follows: 16.7% at pH 6.2 and 23.2 % at pH 3.4 in -veCD condition; 17.1 % at pH 6.2 and 28.4% at pH 3.4 in +veCD condition. However, when neuraminidase treatment was performed at RT and GM3 signal detection was then followed, the estimated % of GM3 clusters due to lactonization was found to be 4.4% at pH 6.2 and 5.4 % at pH 3.4 in -veCD condition; 0.2% at pH 6.2 and 0.6 % at pH 3.4 in +veCD condition. The % of GM3 signal at respective pHs was calculated over the untreated controls signals at the same pH. Hence, GM3 clustering effects were majorly due to GM3-GM3 ligations. The resulting signal of both GM3 and lactosylceramide in the cells that were digested with neuraminidase at RT were lower in comparison to the cells that were kept in cold due to the enhanced endocytosis of the lipids at RT or due to higher activity of neuraminidase. Error Bar=S.D; No. of independent experimental replicates=3. A), B) and C) Shows that intracellular cyclophilinA was very high at pH 3.4 but failed to be released extracellularly, whereas, a high level of vesicular release was observed in the physiological (7.4) and much more at the low (6.2) pH condition. Error Bar=S.D; No. of independent experimental replicates=3; *:p≤0.05, **: p≤0.01, ***:p≤0.001.

GM3 and GFAP supra-clustering.
A), B), C) and D Insets in DIC images show that upon CyPA inhibitor treatment, cyclophilin A vesicular release was inhibited, and it was trapped inside the cells. DIC images represent 9 that within 30-33 hours of CyPA treatment (within 3-6 hours of 2 nd shot) massive cellular blebbing, cell rounding and loss of anchorage occurred. Several tumor cells were levitated and underwent secondary necrosis within this timeframe. A), B), C) and D) CyPA inhibitor treated LN229 tumor cells showed massive GM3 clustered patterns over the untreated cells in oncogenic pH microenvironments in both unperturbed cholesterol and cholesterol high conditions. The patterns of GM3 clusters colocalized well with the intracellular membranous GFAP protein clustering. Note that the Pearson's colocalization coefficient [R-values] between GM3 and GFAP in cyclophilin A inhibitor treated conditions was >0.7±S.D, hence was highly significant. The experiment was repeated 3 times. F-actin was visualized via rhodamine conjugated phalloidin which showed disruption of actin stress fibers in CyPA inhibitor vs. the control condition.