Tumor stressors induce two mechanisms of intracellular P-glycoprotein-mediated resistance that are overcome by lysosomal-targeted thiosemicarbazones

Multi-drug resistance (MDR) is a major obstacle in cancer treatment due to the ability of tumor cells to efflux chemotherapeutics via drug transporters e.g. , P-glycoprotein (Pgp; ABCB1). While the mechanism of Pgp-mediated drug efflux is known at the plasma membrane, the functional role of intracellular Pgp is unclear. Moreover, there has been intense focus on the tumor micro-environment as a target for cancer treatment. This investigation aimed to dissect the effects of tumor micro-environmental stress on sub-cellular Pgp expression, localization and its role in MDR. These studies demonstrated that tumor micro-environment stressors ( i.e. , nutrient starvation, low glucose levels, reactive oxygen species (ROS), and hypoxia) induce Pgp-mediated drug resistance. This occurred by two mechanisms, where stressors induced: (1) rapid Pgp internalization and re-distribution via intracellular trafficking (within 1 h); and (2) hypoxia-inducible factor-1α (HIF-1α) expression after longer incubations (4-24 h), which up-regulated Pgp was lysosomal biogenesis.

Previous studies have shown that Pgp expression is up-regulated in tumor cells through activation of the hypoxia inducible factor-1 (HIF-1α) pathway (5)(6)(7). Moreover, it has been established that tumors typically over-express HIF-1α (8). Hypoxia is typical of the tumor micro-environment (~1% oxygen) (9,10), which is markedly different to the environment of normal tissues where average oxygen levels are approximately 5% (10,11). Apart from hypoxia, tumor cells are also exposed to stressors, including glucose and nutrient starvation, as well as increased reactive oxygen species (ROS) generation (12). The increased ROS production in the tumor micro-environment (12) is linked to drug resistance via its ability to increase Pgp expression (6,13). Furthermore, apart from hypoxia, HIF-1α expression can also be increased by stress stimuli in the tumor microenvironment (8). Hence, the stress-inducing conditions of the tumor micro-environment may play a role in inducing drug resistance through up-regulation of Pgp, which could occur via the HIF-1α pathway (5)(6)(7).
It has been well characterized that plasma membrane Pgp is responsible for the efflux of drugs, such as DOX (14) (Fig. 1Aii). However, recently, our laboratory has established that Pgp is also endocytosed from the plasma membrane into the cell (14)(15)(16). Additionally, Pgp has been identified in the lysosomal membrane, where it is functional in terms of transporting Pgp substrates (e.g., DOX) into the lysosomal lumen ( Fig. 1Aii) (14)(15)(16). In fact, intracellular Pgp was demonstrated by several techniques to associate with the lysosome, but not the mitochondrion or nucleus (14). As part of endocytosis, the Pgp-containing plasma membrane buds inward to form an early endosome, which then "matures" to form a lysosome. This process results in the transport of Pgp substrates, such as DOX, into the lysosome (Fig. 1Aii) (14).
During endocytosis, the catalytic activesite and ATP-binding domain of Pgp are still exposed to the cytosol and enable Pgp to "pump" substrates from the cytosol into lysosomes (14). Hence, when a Pgp substrate, such as DOX, enters the cell, the drug is not only effluxed out by Pgp on the plasma membrane, but is also sequestered within acidic lysosomes by lysosomal Pgp (3) (Fig. 1Aii). If the Pgp substrate is charged at acidic pH (such as occurs with DOX), lysosomal trapping results (14). This effect prevents drugs from reaching their molecular targets (e.g., the nucleus in the case of DOX), leading to increased resistance to the agent (i.e., lysosomal "safe house" effect; Fig.  1Aii) (14).
The present investigation dissected the effects of stress stimuli in the tumor microenvironment on sub-cellular Pgp expression, localization and its role in Pgp-mediated drug resistance. Considering this, for the first time, we demonstrate that well known tumor stress stimuli, namely: hypoxia, nutrient starvation and ROS (12,25,26), rapidly (within 1 h) induce Pgp-re-distribution into lysosomes. This effect results in lysosomal sequestration of chemotherapeutics that are Pgp substrates. Additionally, longer incubations (over 4-24 h) with these stressors resulted in a second mechanism that led to increased Pgp expression via a HIF-1α dependent process and also lysosomal biogenesis. These two mechanisms lead to Pgp-mediated resistance that can be uniquely targeted by therapeutic exploitation using specially designed thiosemicarbazones, such as Dp44mT, that target lysosomal Pgp.
In these investigations, our first experiments assessed parental KB31 cells, which express very low Pgp levels relative to KBV1 (+Pgp) cells that highly express Pgp (6,14).
Notably, for all studies herein, KBV1 (+Pgp) cells were grown in medium supplemented with low levels of vinblastine (VBL; 0.5 μg/mL) for maintenance of a partial MDR phenotype (i.e., "half-resistant" KBV1 (+Pgp) cells) (Fig. 1Ci). This "half-resistant" cell model was specifically utilized throughout our experiments, instead of the "fully-resistant" KBV1 (cells grown in VBL at 1 μg/mL (14,15,17)), to enable increased Pgp expression upon exposure to stress stimuli. As shown in Fig. 1Cii, upon exposure to hypoxia (1% O2) for 24 h/37 o C, Pgp expression in halfresistant KBV1 cells increased to a similar level to those that were fully-resistant. Of note, the two closely migrating Pgp bands observed in the western blots correspond with the immature (140 kDa) and the mature glycosylated form (170 kDa) of Pgp (27). These two forms were only clearly observed in KBV1 cells, but not other cell-types. 4 glucose), serum starvation ((-) serum), or H2O2 stress (100 μM; (+) H2O2) in the presence of normoxia ( Fig. 2A), or hypoxia (Fig. 2B).
As previously observed (6,14), during normoxia, negligible levels of Pgp were observed in KB31 (very low Pgp) cells (Fig.  2Ai). Interestingly, Pgp was not increased in response to glucose starvation, serum starvation, or H2O2 stress, despite the slight, but significant (p < 0.05) increase in HIF-1α expression over time (4, 8 and 24 h) following exposure to these stressors, compared to their respective 0 h controls (Fig. 2Ai).
To further mimic the tumor microenvironment, the stressors, namely: glucose starvation, serum starvation, or H2O2 stress, were imposed on KB31 (very low Pgp) or KBV1 (+Pgp) cells under hypoxia (i.e., 1% O2; Fig.  2Bi, ii). Indeed, hypoxia in tumors can be the result of the abnormally fast metabolism of cancer cells and the poor vascularization of solid tumors (11,28). In these studies, glucose starvation, serum starvation, or H2O2 stress, in combination with hypoxia increased or significantly (p < 0.001-0.05) increased Pgp expression in KB31 (Fig. 2Bi) and KBV1 (+Pgp; Fig. 2Bii) cells at 4, 8 and 24 h compared to their 0 h controls. The only exception to this was for KBV1 cells under H2O2 stress, where a significant (p < 0.001-0.01) increase in Pgp expression was observed after 4 and 8 h, but not 24 h (Fig. 2Bii).
Overall, these results in Fig. 2 demonstrate a positive relationship between tumor stress-induced HIF-1α and Pgp expression.
Micro-environmental Stressors Up-Regulate Pgp and HIF-1α Expression in a Variety of Tumor Cell-Types Endogenously Expressing Pgp -To further assess the effects of micro-environmental tumor stressors on endogenously Pgp-expressing cells, we tested a variety of cancer cell-types, including DMS-53 lung carcinoma, DU-145 prostate carcinoma, MDA-MB-231 breast carcinoma, PANC-1 pancreatic carcinoma and PC3 prostate carcinoma cell lines (Fig. 3A, B). Based on the studies above using KB31 and KBV1 cells (Fig.  2), where an 8 h incubation showed a marked response to stressors, this time point was chosen to investigate the effects of stressors on tumor cell-types endogenously expressing Pgp under normoxia and hypoxia (Fig. 3A, B).
Incubation of DMS-53 cells under hypoxia for 8 h (i.e., 8 h control) led to a marked and significant (p < 0.01-0.05) up-regulation of Pgp and HIF-1α expression compared to the control (0 h; Fig. 3B). Notably, no further increase in Pgp or HIF-1α expression in DMS-53 cells was observed when hypoxia was combined with additional stressors compared to the 8 h control under hypoxia alone (Fig. 3B).
Under hypoxia, DU-145, MDA-MB-231, PANC-1 and PC3 cells, all demonstrated a significant (p < 0.001-0.05) up-regulation of Pgp and HIF-1α under control hypoxia at 8 h relative to 0 h (Fig. 3B). The combination of hypoxia with stressors (8 h) resulted in a similar or even greater increase in Pgp or HIF-1α expression compared to hypoxia alone (Fig. 3B). In general, the increase in both Pgp and HIF-1α expression relative to the respective control was greater under all conditions in the presence of hypoxia relative to normoxia (cf. Fig. 3A and 3B).
In summary, the results in Fig. 3A, B demonstrate that micro-environmental stressors induce Pgp and HIF-1α expression similarly across multiple endogenous Pgp-expressing tumor cell-types.
Micro-environmental Stressors Up-Regulate Pgp via the HIF-1α pathway -Our studies above suggested that tumor microenvironmental stressors induced Pgp expression potentially via a HIF-1α-dependent pathway (Figs. 2, 3). Moreover, the effect was more pronounced under hypoxia. This would agree with previous studies suggesting the role of HIF-1α in Pgp expression (5)(6)(7). Considering this, to directly examine the role of HIF-1α in Pgp upregulation after exposure to microenvironmental stressors, HIF-1α was silenced using siHIF-1α relative to the non-targeting control (siNC) in KBV1 (+Pgp) cells (Fig. 4). Following incubation of cells with siNC or siHIF-1α for 48 h, a range of stressors was implemented under hypoxia for 8 h relative to the 0 h control, as described above (see Fig. 3).
Hypoxia led to a marked and significant (p < 0.001-0.05) up-regulation of both Pgp and HIF-1α expression in siNC-treated cells after an 8 h incubation in the presence or absence of stressors relative to the 0 h control (Fig. 4). In contrast to cells incubated with the siNC, HIF-1α was undetectable in siHIF-1α-treated cells for all conditions under hypoxia (Fig. 4). In agreement with the lack of HIF-1α expression in siHIF-1αtreated cells, Pgp expression was markedly down-regulated relative to siNC in the 8 h control and after serum starvation and H2O2 stress (Fig. 4). These data indicate that HIF-1α silencing decreased Pgp expression. Notably, under glucose starvation in siHIF-1α-treated cells, both Pgp and β-actin levels were barely detectable (Fig. 4), despite equivalent protein loads being applied. However, under the combined conditions of hypoxia and glucose starvation in siHIF-1α-treated cells, a high degree of necrotic cell death was observed (data not shown). Given that necrosis is characterized by extensive proteolysis (30,31), this would 6 explain the low Pgp and β-actin levels under this condition.
Overall, these studies in Fig. 4 indicate that a high degree of stress-mediated Pgp expression was dependent on HIF-1α. Indeed, HIF1-α may transcriptionally induce Pgp expression via the functional hypoxia responsive element within the promoter region of the PGP gene (32,33).
Tumor Micro-Environmental Stressors Lead to Increased Pgp Re-distribution to Lysosomes -Up-regulation of Pgp through the HIF-1α pathway in response to stress has been linked to bolstering the ability of tumor cells to survive chemotherapy (34). Furthermore, stressors in the tumor micro-environment have also been implicated in triggering endocytosis as a cell survival mechanism (35), which occurs through a HIF-1α-associated pathway (36). This endocytic process may allow re-distribution of Pgp into endosomes and lysosomes (14,16).
To investigate how tumor microenvironmental stressors affect Pgp expression as well as sub-cellular trafficking and localization of Pgp, western analysis and confocal microscopy were implemented using KBV1 (+Pgp) cells (Fig. 5). These procedures assessed Pgp levels and also its co-localization with a well-characterized lysosome-specific marker, namely lysosome-associated membrane protein 2 (LAMP2) (37), and the nuclear marker, 4',6diamidino-2-phenylindole (DAPI) (38). Notably, a short incubation time (1 h) with the different stressors (i.e., (-) glucose), (-) serum or (+) H2O2) was initially chosen to avoid an increase in Pgp expression. This allowed assessment of sub-cellular trafficking and re-distribution of Pgp, rather than overall Pgp up-regulation. In fact, using these conditions, western analysis demonstrated no significant (p > 0.05) change in Pgp or LAMP2 expression with all stressors after a 1 h incubation under normoxia (Fig. 5A).
Using confocal microscopy, Pgp expression (green) was observed in control KBV1 (+Pgp) cells under normoxia (Fig. 5Bi). Under control normoxia, basal levels of LAMP2-stained lysosomes (visible as cytoplasmic red puncta) and Pgp (seen predominantly as green staining on the plasma membrane) were evident (Fig. 5Bi, Bii). The colocalization of the red and green signals in the merge (yellow) was similar comparing the 0 h time point (Fig. 5Bi, C) and after a 1 h incubation (Fig. 5Bii, C). Analysis of these images using ImageJ resulted in Mander's overlap coefficients (39) of 0.68 and 0.67 at 0 h and 1 h, respectively (Fig. 5Bi,ii). Under these conditions, there is a LAMP2-staining pattern in close proximity to the plasma membrane resulting in some co-localization with Pgp. This overlap is expected, as based on our model (Fig.  1A, B), the process of Pgp internalization via endocytosis occurs continuously with lysosomal co-localization occurring at a steady-state level. Hence, the 0-or 1-h incubation provides a "snapshot" of the basal, steady-state level of Pgp internalization and its co-localization with the LAMP2-lysosomal marker (Fig. 5Bi,ii). However, after incubation of KBV1 cells with each stressor for 1 h (i.e., (-) glucose, (-) serum, or (+) H2O2), a pronounced and significant (p < 0.001) increase in Pgp co-localization with the lysosomal marker, LAMP2, was observed as a yellow punctate lysosomal pattern in the merge ( Fig. 5Biii-v, C) compared to the relevant 1 h control (Fig. 5Bii, C). Analysis of the images in Fig. 5Biii-v using ImageJ resulted in an increase in the Mander's overlap coefficients (39) to 0.94, 0.96 and 0.93 for glucose starvation, serum starvation and H2O2 stress, respectively ( Fig.  5Biii-v). Collectively, these observations indicate that Pgp is redistributed to co-localize 7 intracellularly with LAMP2-containing lysosomes after a short exposure (1 h) to stress stimuli.
The effect of glucose starvation, serum starvation, or H2O2 stress was also investigated in KBV1 (+Pgp) cells under hypoxia (1 h) using western blotting and confocal microscopy ( Fig.  5D-F). Under hypoxia, western blotting demonstrated no significant (p > 0.05) increase in Pgp or LAMP2 expression after a 1 h exposure to hypoxia and stressors relative to the 0 h control (Fig. 5D). Confocal microscopy showed that in KBV1 (+Pgp) cells, Pgp was poorly co-localized with LAMP2 (Mander's overlap coefficient: 0.50) at the 0 h control time point, where the cells had not been exposed to hypoxia (Fig. 5Ei, F). However, subjecting cells to hypoxia for 1 h (Fig. 5Eii, F) led to a significant (p < 0.05) increase in Pgp colocalization with LAMP2 (forming a yellow punctate lysosomal pattern in the merge) compared to the 0 h control condition (Fig. 5Ei, F). Analysis of these images using ImageJ resulted in Mander's overlap coefficients of 0.50 and 0.81 at 0 h and 1 h, respectively (Fig.  5Ei,ii). This increase in overlap during hypoxia suggested Pgp re-distribution to the lysosomal compartment ( Fig. 5Ei,ii).
Similarly, exposing KBV1 (+Pgp) cells to tumor micro-environmental stressors and hypoxia for 1 h produced a more intense pattern of co-localization of LAMP2 and Pgp in the merge ( Fig. 5Eiii-v, F). This effect was similar to that observed using the same stressors under normoxia (Fig 5Biii-v). Indeed, following the incubation with all stressors for 1 h, there was a pronounced and significant (p < 0.001) increase in Pgp co-localization with the LAMP2 lysosomal marker in the merge ( Fig. 5Eiii-v, F) compared to the relevant 1 h control (Fig. 5Eii, F).
These results in Fig. 5B, C, E, F demonstrate that Pgp expressing cells respond rapidly (within 1 h) to stressors increasing Pgp co-localization with lysosomes under normoxia and hypoxia. These studies were consistent with those in Fig.  5Bi,ii under control normoxic conditions in the absence of stressors. In contrast, under hypoxia, western blotting demonstrated a significant (p < 0.001-0.01) increase in Pgp expression from 8-24 h and a significant (p < 0.001) increase in LAMP2 levels after a 4-24 h incubation (Fig.  6D).
Furthermore, confocal microscopy 8 demonstrated that hypoxia induced a significant (p < 0.001) increase in LAMP2 from 4-24 h and Pgp levels from 8-24 h and significantly (p < 0.001) increased their co-localization at 4, 8 and 24 h relative to the 0 h control ( Fig. 6Ei-iv, F). The increase in LAMP2 was consistent with lysosomal biogenesis after exposure to stressors and was also confirmed using another lysosomal marker, cathepsin D (16,35) (Fig. S1).
In summary, in addition to the redistribution of Pgp observed after a short incubation of 1 h under various stressors (

Stressors Increase Internalization of Pgp from the Plasma Membrane into
Cathepsin-Stained Lysosomes -To further dissect the mechanism of the increased intracellular Pgp levels after incubation with micro-environmental stressors, studies assessed Pgp internalization from the plasma membrane to intracellular compartments. This was achieved using pulse-chase analysis, and combining data obtained from examining 125 I-labeled anti-Pgp antibody (Ab) internalization and immunofluorescence examining co-localization of anti-Pgp Ab and a well characterized lysosomal marker, cathepsin D (Fig. 7A).
In initial studies to validate the pulse chase experiments and the role of temperaturedependent endocytosis (40) in Pgp internalization from the plasma membrane, the uptake of 125 I-labeled anti-Pgp Ab was examined using KBV1 (+Pgp) cells. The plates were first cooled on ice to limit endocytosis (40) and 125 Ilabeled Ab added for 1 h/4 o C to label cellsurface Pgp. These plates were then washed on ice and incubated with the protease, Pronase (1 mg/ml/4 o C), to examine membrane-bound (Pronase-sensitive) and internalized 125 I-anti-Pgp Ab (Pronase-resistant; (40)(41)(42)). To determine if Pgp is internalized at 37 o C, the 125 I-Ab labeled plates at 4 o C were washed and then pre-warmed media was added and the cells incubated for 1 h/37 o C. Cells were then treated with Pronase to separate the internalized compartment. These studies demonstrated that for the 4 o C control, most 125 I-labeled anti-Pgp Ab remained on the cell membrane, with little internalization, while at 37 o C, a 2-3 fold increase in 125 I-Ab internalization was observed (Fig. S2). While these observations with the 125 I-Ab demonstrate increased internalization of Pgp, they do not provide evidence of co-localization of Pgp specifically to lysosomes.
Hence, a similar pulse-chase protocol under the same conditions then assessed colocalization of Pgp from the membrane to the cathepsin D-defined lysosomal compartment (Fig. 7A). As found for the 125 I-labeled anti-Pgp Ab-binding studies, for the 4 o C control, most Pgp remained on the plasma membrane with limited internalization and co-localization ( Fig.  7A). This latter observation is probably due to the limitation of endocytosis at this temperature (40). Notably, co-localization did not significantly (p > 0.05) increase under the 37 o C control condition compared to its control condition at 4 o C. However, Pgp co-localization with cathepsin D was markedly and significantly (p<0.001) stimulated by incubation with the stressors at 37 o C compared to its 37 o C control. In fact, stressors resulted in a pronounced increase in co-localization of Pgp with the cathepsin D lysosomal marker (Mander's overlap coefficient: 0.87-0.88; Fig. 7A).
Initial analysis of these images ( Fig.  7Bi-v) was then performed using the software, 9 ImageJ, and its analytic tool, Plot Profile (43). This type of analysis displays intensities of different channels (i.e., green for Pgp and red for cathepsin D) across the white line that crosses the cell body in the merged image in Fig. 7A. Notably, for the 4 o C control, the intensity of the Pgp fluorescence peaks sharply on the outside boundaries of the cell and then decreases markedly in the intracellular compartment, while cathepsin D was confined only to the cell body (Fig. 7Bi). Examining the 37 o C control, the intensity of Pgp channel begins to increase in the intracellular compartment relative to the external membrane (Fig. 7Bii).
These results are consistent with the increase in internalization of the 125 I-Pgp Ab at 37 o C relative to 4 o C (Fig.  S2). Hence, there is an alteration in distribution of Pgp between the plasma membrane and the intracellular compartment (Fig. 7Bii), but no change in total Pgp after only 1 h of incubation ( Fig. 5A-C). Upon the addition of stressors, the intracellular Pgp and cathepsin D channels strongly merge ( Fig. 7Biii-v).
Additional analyses using ImageJ was performed by masking the cathepsin D channel to demonstrate the distribution of Pgp and enable quantitation of relative Pgp fluorescence intensity inside and outside the cathepsin Ddefined lysosomal region (Fig. 7C). In direct accordance with the previous analyses ( Fig. 7A,  B), these studies demonstrated a marked and significant (p < 0.001) increase in Pgp inside the lysosome-defined region upon incubation at 37 ℃ with the stressors relative to that found with the 37 ℃ control (Fig. 7C). All these data in Fig. 7 agree with a model of Pgp internalization and trafficking from the plasma membrane to the lysosome upon exposure to micro-environmental stressors.

Tumor Micro-Environmental Stressors Increase Functional Lysosomal Pgp under both
Normoxia and Hypoxia, Redirecting DOX into Lysosomes and Away from Nuclei -Our laboratory has previously demonstrated that Pgp actively mediates transport of Pgp substrates into lysosomes (14)(15)(16). In particular, DOX was demonstrated to be sequestered by Pgp into the lysosome and away from the nucleus, resulting in resistance to this agent ( Fig. 1Aii; (14,16,24)). Considering this, we next assessed if tumor micro-environmental stressors could potentiate this effect. To assess the functionality of the redistributed lysosomal Pgp under stressful conditions, intracellular trafficking and localization of the intrinsically fluorescent Pgp substrate, DOX (44), was examined by live-cell immunofluorescence microscopy (Figs. 8,9).
In these studies, KBV1 (+Pgp) cells were incubated with tumor micro-environmental stressors under both normoxic (Fig. 8) and hypoxic ( Fig. 9) control conditions (0 h), leading to DOX localization in the nucleus ( Fig. 8Ai  and 9Ai). The addition of the potent Pgp inhibitor, Elacridar (Ela; 0.2 µM; (14,15)), resulted in no change in the overlap between DOX and LAMP2 relative to the 0 h control (Fig. 8Aii,B and 9Aii,B cf. to 8Ai,B and 9Ai, B). However, Ela led to a significant (p < 0.01-0.05) increase in DOX nuclear accumulation ( Fig. 8Aii, C and 9Aii, C) compared to that observed with DOX alone (Fig. 8Ai, C and Fig.  9Ai, C). This result was probably due to the ability of the Pgp inhibitor to block Pgpmediated DOX export out of the cell and also Pgp-induced lysosomal DOX accumulation, thereby allowing DOX to access its nuclear targets.
However, when KBV1 (+Pgp) cells were glucose starved for 1 h either under normoxia or hypoxia, there was a significant (p < 0.01) increase in DOX overlap with LAMP2stained lysosomes (Fig. 8Aiii, B and 9Aiii, B).

10
The Pgp inhibitor, Ela, reversed the effect of glucose starvation, so that DOX then accumulated in the nucleus and not LAMP2stained lysosomes (Fig. 8Aiv, C and 9Aiv, C).
Similarly to the results observed with glucose starvation under normoxia and hypoxia, serum starvation or H2O2 stress also induced significantly (p < 0.001-0.01) more overlap of DOX with lysosomal LAMP2 (see yellow overlap in the merge; Fig. 8Av,vii, B and 9Av,vii, B) compared to control cells (0 h; Fig.  8Ai, B and 9Ai, B). Again, this effect was reversed with Ela, resulting in a significant (p < 0.001-0.01) increase in DOX nuclear staining ( Fig. 8Avi, viii, C and 9Avi, viii, C) relative to DOX alone ( Fig. 8Av, vii, C and 9Av, vii, C). The re-distribution of DOX to the nucleus after inhibition of Pgp by Ela was in good agreement with our previous study (14). In fact, we demonstrated that selective Pgp inhibitors such as Ela or valspodar (45)(46)(47)(48) acted similarly to Pgp silencing to inhibit the entrance of DOX into lysosomes, leading to its uptake into nuclei (14). Considering the demonstrated potency and selectivity of Ela (14)(15)(16)(17)24,(45)(46)(47)(48), this inhibitor was used throughout this study, as it gave results similar to both valspodar and Pgp siRNA.
In contrast to KBV1 (+Pgp) cells (Figs. 8,9), when KB31 cells (very low Pgp) were utilized to study DOX overlap with lysosomal LAMP2 in the presence and absence of Ela and stressors, there was none observed under either normoxia ( Fig. S3) or hypoxia (Fig. S4). Further, compared to the untreated control, the addition of Ela had no significant (p > 0.05) effect on the relative distribution of Pgp in KBV1 (+Pgp) cells in the presence or absence of stressors (Fig. S5). Collectively, these controls indicated that Ela acted as a specific Pgp inhibitor in KBV1 (+Pgp) cells and did not nonspecifically affect DOX or Pgp distribution.
Together, the studies and controls in Figs. 8,9,and Figs. S3, S4, and S5 are consistent with the hypothesis that in Pgpexpressing cells, stressors typically present in the tumor micro-environment (49) result in redistribution of functional Pgp to LAMP2-stained lysosomes. This in turn could lead to increased lysosomal accumulation of Pgp substrates, such as DOX, which prevents its interaction with its nuclear targets Tumor Micro-Environment Stressors Increase Pgp-Mediated Lysosomal Damage by Thiosemicarbazones -The studies above demonstrate that tumor micro-environmental stressors result in rapid re-distribution of Pgp into lysosomes (Fig. 5), followed by sustained up-regulation of Pgp (Figs. 2Aii, 2Bi,ii, 3, 6) and lysosomal biogenesis over longer incubations (Fig. 6D, E). This combined effect results in transport of Pgp substrates into lysosomes (Figs. 8,9). Considering this, we next investigated if micro-environmental stressor-induced redistribution of Pgp into the lysosome could also be used to potentiate LMP and cell death in resistant cancer cells. To test this we used the novel anti-tumor agent, Dp44mT ( Fig. 1Bi) that was previously shown to "hijack" lysosomal Pgp (15)(16)(17)24), inducing more LMP and cell death in resistant cancer cells compared to cells with lower Pgp-expression ( Fig. 1Bii) (15,18,24).
In these studies, lysosomal stability was examined using the lysosomotropic, metachromatic fluorophore, acridine orange (AO) (50) by live-cell immunofluorescence microscopy. High lysosomal concentrations of AO result in a red fluorescence, while lower cytosolic and nuclear AO concentrations generate green fluorescence (50). Both KB31 (very low Pgp) and KBV1 (+Pgp) cells were pre-incubated for 1 h under normoxia with either control medium, or stressors, namely: glucose starvation, serum starvation, or H2O2 (Fig. 10). Cells were then incubated with Dp44mT (25 µM) in the presence or absence of the Pgp inhibitor, Ela (0.2 µM), in the continued absence or presence of these micro-environmental stressors for 24 h/37 o C.
Examination of KB31 (very low Pgp) cells incubated with Dp44mT (25 µM) in control media (Fig. 10Av) or together with tumor microenvironmental stressors (Fig. 10Avi-viii), demonstrated a granular red AO fluorescence similar to the respective controls ( Fig. 10Ai-iv As observed for KB31 (very low Pgp) cells, KBV1 (+Pgp) cells exposed to stressors under normoxia (Fig. 10Axiv-xvi), showed no significant (p > 0.05) change in lysosomal stability relative to the control (Fig. 10Axiii, B). Further, Dp44mT (25 µM) had no significant (p > 0.05) effect on KBV1 (+Pgp) cells under control normoxia (24 h; Fig. 10Axvii, B). In contrast, incubation of KBV1 (+Pgp) cells with Dp44mT under glucose starvation, serum starvation, or H2O2 stress, resulted in a marked and significant (p < 0.001) loss of red fluorescence and the disappearance of redstained AO vesicles consistent with increased lysosomal damage and LMP ( Fig. 10Axviii-xx,  B). However, when adding the Pgp inhibitor, Ela (0.2 μM), to cells treated with Dp44mT and the stressors, the distinctive lysosomal red fluorescence pattern was again observed (Fig.  10Axxii-xxiv, B). This observation indicates that inhibiting Pgp activity with Ela prevented the ability of Dp44mT to damage lysosomes when cells were under stress.
These results in Fig. 10 indicate that stressors potentiate Pgp-mediated lysosomal damage induced by Dp44mT. Importantly, this occurred only in Pgp-expressing KBV1 cells and was not observed in KB31 (very low Pgp) cells.
To further mimic the tumor microenvironment, these treatments using stressors in Fig. 10A, B were repeated using the same conditions, except under hypoxia (Fig. 11A, B). As per the observations under normoxia (Fig.  10Av), Dp44mT alone did not induce LMP under hypoxia in KB31 (very low Pgp) cells (Fig. 11Av). However, in contrast to normoxia (Fig. 10Avi-viii), the red-stained lysosomes disappeared when KB31 (very low Pgp) cells were incubated under hypoxia with Dp44mT and either glucose starvation, serum starvation, or H2O2 ( Fig. 11Avi-viii). This indicated lysosomal damage under these conditions. The Pgp inhibitor, Ela (0.2 μM), reversed this latter effect of Dp44mT ( Fig.  11Ax-xii vs. Fig. 11Avi-viii) by inhibiting Pgpmediated transport of Dp44mT into lysosomes, and thus, preventing lysosomal damage (15)(16)(17). Hence, these observations demonstrated that stressors up-regulate Pgp, even in very low Pgp expressing KB31 cells, as shown in western blotting studies (Fig. 2Bi). This up-regulation of Pgp is probably responsible for the increased lysosomal damage via Dp44mT (Fig. 11Aviviii).
Similar to normoxia (Fig. 10Axviii-xx), incubation of KBV1 (+Pgp) cells under hypoxia with Dp44mT and either glucose starvation, serum starvation, or H2O2, resulted in a loss of red fluorescence that is indicative of LMP ( Fig.  12 11Axviii-xx). In contrast, the lysosomal red fluorescence appeared upon the addition of Ela to cells treated with Dp44mT and the stressors under hypoxia (Fig. 11Axxii-xxiv).
These results in Fig. 11 demonstrate that hypoxia combined with other stressors can potentiate the Pgp-mediated lysosomal damage of Dp44mT not only in highly Pgp-expressing KBV1 cells, but also in cells such as KB31 that express very low Pgp levels under basal conditions.

Tumor
Stressors Induce Pgp-Resistance to DOX -To determine how tumor micro-environmental stressors affect Pgpmediated drug resistance, the cytotoxicity of DOX was characterized in KB31 cells (very low Pgp), KBV1 (+Pgp) cells and an endogenously Pgp-expressing cell line, DMS-53 (29,51), under normoxia, or hypoxia (Table 1A-C; Fig. S6A-F). All cell-types were pre-incubated for 1 h/37 o C with either normoxia or hypoxia (1% O2) and the different micro-environmental stressors found in tumors (i.e., glucose starvation, serum starvation and H2O2). The cells were then incubated for a further 24 h/37 o C under normoxia or hypoxia with these stress conditions with DOX (0.2-100 µM) in the presence and absence of the Pgp-inhibitor, Ela (0.2 µM). The concentration of DOX required to inhibit proliferation by 50% (i.e., IC50 value) was then calculated.
Initially, as a negative control, very low Pgp expressing KB31 cells (14) underwent a 24 h incubation with DOX under normoxia in the presence and absence of Ela (Table 1A). There was no significant (p > 0.05) difference in the IC50 of KB31 (low Pgp) cells incubated under normoxic control conditions with DOX in the presence and absence of the Pgp-inhibitor, Ela, (0.2 μM; Table 1A). Similarly, Ela had no effect on the IC50 of DOX in KB31 cells, that were glucose starved (0 μM), serum starved, or exposed to H2O2 (100 μM) stress for 24 h under normoxia (Table 1A). Interestingly, the IC50 of DOX and DOX + Ela under serum starvation were significantly (p < 0.05) increased relative to DOX alone or DOX + Ela, respectively under control conditions (Table 1A). In this case, Ela had no effect on the IC50, suggesting the increased IC50 value was independent of Pgp. Considering this, the high IC50 may be due to the G1/S arrest caused by serum starvation, which may decrease DOX activity (52). The lack of effect of Ela on the cytotoxicity of DOX under normoxia, was probably because the stressors did not alter Pgp expression under these conditions (Fig. 2Ai).
To further test the effect of stressors on the cytotoxicity of DOX, KB31 (very low Pgp) cells were subjected to hypoxia alone, or hypoxia with glucose starvation, serum starvation, or H2O2 stress (100 μM; Table 1A). The IC50 for DOX under these conditions was slightly higher than that under normoxia alone (Table 1A), which is consistent with an upregulation of Pgp (Fig. 2Bi), and thus, the reduced efficacy of DOX. The addition of Ela to KB31 (very low Pgp) cells under control conditions, glucose starvation, serum starvation or H2O2 stress during hypoxia led to a decrease, or a significant (p<0.05) decrease (for serum starvation) in the IC50. This observation indicates an increase in DOX cytotoxicity, suggesting a role for Pgp in resistance against this drug (Table 1A). Together with the results in Fig.  2Bi, these data indicate that hypoxia plays a major role in Pgp expression and activity, which contributes to drug resistance towards DOX.
In order to compare the effect of Pgp on drug efficacy, the IC50 of DOX in the presence or absence of Ela (0.2 μM) was also investigated in 13 KBV1 (+Pgp) cells under normoxia or hypoxia (Table 1B). Implementing control normoxia, an IC50 value could not be established for DOX, as it was >100 µM (Table 1B). This result suggested a role for the high Pgp levels in KBV1 cells conferring resistance against DOX relative to KB31 (very low Pgp) cells (Table 1B). This was substantiated by studies showing that Ela (0.2 μM) caused a pronounced decrease in the IC50 of DOX in KBV1 (+Pgp) cells from >100 µM to 18.1 + 2.8 µM (Table 1B).
This latter trend was also observed during glucose starvation, serum starvation and H2O2 stress under normoxia, where the IC50 of DOX was >100 µM, while the addition of Ela (0.2 μM) again markedly sensitized KBV1 (+Pgp) cells to drug treatment resulting in IC50 values from 23.9-34.5 µM (Table 1B). Furthermore, the cytotoxicity of DOX in KBV1 (+Pgp) cells under hypoxia alone and in combination with stressors was similar to that under normoxia being >100 µM (Table 1B). Under all conditions, the addition of Ela (0.2 μM) led to a pronounced decrease in the IC50 to 18.8-31.6 μM, indicating an increase in DOX cytotoxicity probably due to inhibition of Pgp (Table 1B).
Similar results to those for KBV1 cells (Table 1B) were also obtained using DMS-53 lung cancer cells ( Table 1C) that express relatively high endogenous Pgp levels (Fig. 3). Under normoxic control conditions, Ela led to a marked and significant (p<0.01) decrease in the IC50 of DOX from 90.1 + 7.0 µM to 30.2 + 2.4 µM (i.e., increase in DOX cytotoxicity; Table  1C). All stress conditions showed a marked increase in DOX cytotoxicity in the presence of Ela (Table 1C). Under hypoxia in the presence or absence of stressors, the IC50 of DOX increased or remained similar to that observed under normoxia, suggesting decreased or comparable efficacy of DOX (Table 1C). Similarly to normoxia, the addition of Ela under control hypoxia led to an increase in DOX cytotoxicity (i.e., from >100 µM to 43.0 + 6.0 µM), which is probably due to the inhibitor suppressing Pgp activity (Table 1C). This marked increase in DOX cytotoxicity in the presence of Ela was also observed with glucose starvation, serum starvation and H2O2 stress under hypoxia (Table 1C).
Overall, in the presence or absence of stressors for KBV1 (+Pgp) and DMS-53 (+Pgp) cells under normoxia and hypoxia, or for KB31 (very low Pgp levels) cells under hypoxia, Ela increased DOX cytotoxicity (Table 1A-C), suggesting a role for Pgp in conferring DOX resistance. Notably, Ela could only increase DOX cytotoxicity in KB31 cells (Table 1A) under stressful conditions (i.e., hypoxia in the presence or absence of additional stress stimuli) where an increase in Pgp expression occurred (cf. Fig. 2Ai and Bi).
Tumor Stressors Up-Regulate Pgp, Increasing Sensitivity to Dp44mT -As hypoxia and the ensuing Pgp expression leads to resistance to DOX ( Table 1), one strategy to combat this major clinical problem (53) could be the use of agents such as Dp44mT, that overcome Pgp-mediated resistance by targeting lysosomes (Fig. 1Bii) (15,16). As stressors such as hypoxia, exacerbate DOX resistance ( Table  1) by up-regulating Pgp ( Fig. 2A cf. 2B; Fig. 3A cf. 3B), it was crucial to establish how Dp44mT efficacy is influenced by tumor stressors (Fig.  12). This was assessed using the same experimental conditions described above for Table 1 implementing KB31 (Fig. 12A), KBV1 (Fig. 12B), and DMS-53 cells (Fig. 12C) comparing normoxia or hypoxia.
14 For KB31 (very low Pgp) cells under normoxia (Fig. 12Ai), the IC50 of Dp44mT was significantly (p < 0.05) increased (i.e., increased resistance and decreased cytotoxicity) only under glucose starvation, but was not significantly (p > 0.05) affected by the other stressors. The reason for this unique effect of glucose starvation is unclear, but as the Pgp inhibitor, Ela, had no influence on the IC50 of Dp44mT (Fig. 12Ai), this effect does not appear to be due to Pgp activity.
Using KB31 (very low Pgp) cells under hypoxia in the presence of the other stressors (Fig. 12Aii), Dp44mT was significantly (p < 0.05) more cytotoxic relative to the hypoxic Dp44mT control. Moreover, Ela significantly (p < 0.001-0.01) decreased the sensitivity to Dp44mT (i.e., increased IC50) using all conditions under hypoxia (Fig. 12Aii). These data with Ela indicate the role of Pgp activity in mediating the cytotoxicity of Dp44mT. That is, these results are consistent with Ela inhibiting Pgp-mediated transport into the lysosome and preventing lysosomal damage and cytotoxicity by Dp44mT when exposed to hypoxia and stressors (see Figs. 1Bii, 10A). The fact that this inhibition by Ela only occurred under hypoxia, and not normoxia, is consistent with the increased Pgp expression under hypoxia in KB31 (very low Pgp) cells (Fig. 2Ai cf. Fig.  2Bi).
Collectively, the results in Fig. 12 suggest that tumor stressors potentiate Dp44mT cytotoxicity, especially under hypoxia. This is probably due to the increased Pgp re-distribution to, and expression in, lysosomes (Figs. 2, 5, 6) leading to Dp44mT-mediated lysosomal damage (Figs. 10, 11) and increased cytotoxicity in resistant cancer cells (Fig. 1Bii). In contrast, for DOX, the increase of Pgp in lysosomes after stressors decreases DOX cytotoxicity (i.e., increased IC50; Table 1) due to the trapping of DOX in lysosomes, preventing access to its sensitive nuclear targets (Fig. 8, 9; so called "safe house" effect; see Fig. 1Aii).

DISCUSSION
In this study, we demonstrate for the first time that micro-environmental stressors such as hypoxia, serum, glucose starvation and H2O2 markedly increase drug resistance to known chemotherapeutics such as DOX. Furthermore, we demonstrate that stressors characteristic of the tumor micro-environment greatly affected drug resistance not only by upregulating Pgp, but also by re-localizing Pgp to lysosomes. This investigation also showed that stressors up-regulate HIF-1α, which is involved in the transcriptional up-regulation of Pgp (5)(6)(7)32).
Under normoxia, HIF-1α is a short-lived protein due to its proteasome-mediated degradation that is enhanced by the irondependent prolyl hydroxylases (PHD) 1-3 (54,55). However, tumor micro-environmental stressors such as hypoxia, glucose starvation and increased ROS production (12), can decrease PHD activity, resulting in increased HIF-1α expression (55,56). Our data demonstrate that micro-environmental stressors play a marked role in up-regulating Pgp and HIF-1α expression in a range of cell-types. Moreover, studies implementing silencing of HIF-1α led to suppression of stress-dependent Pgp upregulation. These results indicate that HIF-1α is involved in Pgp up-regulation in response to cellular stress (5-7).
Stressors in the tumor have also been implicated in triggering endocytosis, which is initiated as a cell survival mechanism (16,35). In the current studies, a short (1 h) exposure of Pgp expressing KBV1 cells to tumor stressors under normoxia or hypoxia did not lead to upregulation of Pgp (Fig. 5A). However, this short exposure to stress stimuli resulted in a major change in Pgp distribution, resulting in significant co-localization with lysosomes ( Fig.  5B-C, 5E-F). Indeed, we demonstrated that a short-term exposure to stressors increase internalization of Pgp from the plasma membrane, into the lysosomal compartment ( Fig. 7A-C). Further to this, under prolonged hypoxia, an increase in both Pgp and lysosomal-LAMP2 expression was observed in addition to their co-localization (Fig. 6D-F). The increase of the lysosomal marker, LAMP2, in response to prolonged stress was in line with the ability of the cell to adapt to stressors, resulting in up-regulation of autophagy (57) and lysosomal biogenesis (16,58).
The stress-induced increase in lysosomal Pgp was also found to be functional and led to increased trapping of DOX in lysosomes, resulting in decreased targeting of DOX to the nucleus (Figs. 8, 9). In fact, tumor stressors led to DOX significantly re-distributing to LAMP2positive lysosomes in Pgp-expressing cells. The crucial role of stress-induced lysosomal Pgp was further supported by the observation that the Pgp inhibitor, Ela, prevented DOX sequestration in lysosomes, allowing DOX to reach one of its major molecular targets (i.e., nuclear DNA; Figs. 8,9). Hence, not only does Pgp facilitate drug efflux out of cells, but when this drug pump is endocytosed, it can also promote trapping of substrates in lysosomes (14,16). For DOX, this decreases its anti-tumor efficacy due to the "safe house" effect, which prevents interaction of DOX with its molecular targets in the nucleus (see Fig. 1Aii). Furthermore, the hydrolytic environment of the lysosome and its role in the dynamic process of autophagy could facilitate the degradation of DOX. The lysosome is the site of ~40 hydrolytic enzymes and its acidic pH that mediates the degradation of myriad substrates (59) provides an optimized "reaction vessel" for facilitating drug breakdown. Hence, the ability of the lysosome to safely sequester DOX away from its key molecular targets in the nucleus (this study) and also its dynamic role in hydrolytic degradation could both play a role in DOX resistance.
As an alternative strategy for overcoming Pgp-mediated drug resistance, Dp44mT has been shown to "hijack" lysosomal Pgp pumps to induce lysosomal damage via ROS generation (15). This investigation demonstrated that tumor stressors under normoxia or hypoxia result in increased 16 Dp44mT Pgp-mediated lysosomal damage (Figs. 10,11). Interestingly, when incubating low Pgpexpressing KB31 cells with hypoxia, these cells became sensitive to Dp44mT-induced lysosomal damage under tumor stressors. The lysosomaldamaging effects were directly linked to Pgp activity, as the potent Pgp inhibitor, Ela (14), rescued lysosomal damage (Figs. 10, 11). Thus, our studies demonstrated that stress stimuli increased lysosomal Pgp, which can potentiate lysosomal damage for agents targeting this organelle.
DOX induces cell death through DNA intercalation, etc. (60). However, the therapeutic use of this drug can be severely limited in patients due to Pgp-mediated resistance (61). In the current studies, Pgp was induced by hypoxia and other stressors, leading to decreased DOX cytotoxicity even in very low Pgp expressing KB31 cells. Opposite to the effect of DOX, under hypoxia, tumor stressors potentiated the effectiveness of Dp44mT to induce cytotoxicity to very low Pgp expressing KB31 cells (Fig.  12A). The activity of Pgp played a central role in this potentiated cytotoxicity of Dp44mT, as the Pgp inhibitor, Ela, decreased Dp44mT-mediated cytotoxicity under all stressors in KBV1 and DMS-53 cells (Fig. 12B, C). Hence, the marked differences in the pharmacology of Dp44mT and DOX enable Dp44mT to take advantage of the tumor micro-environment to utilize lysosomal Pgp to kill resistant cancer cells.
In conclusion, the classical anti-cancer agent, DOX, presents limited cytotoxicity to Pgp-expressing cells in the stressful tumor micro-environment.
The tumor microenvironment-inducible Pgp prevents DOX from reaching one of its major molecular targets, namely the nucleus (6,16) (Fig. 13). However, the stressors typically found under these conditions resulted in potentiated cytotoxicity of novel lysosomal targeting agents, such as Dp44mT, via increased Pgp-mediated targeting of lysosomes. In fact, the stressful tumor microenvironment increases functional Pgp on the lysosomal membrane and facilitates entry of Dp44mT into this organelle, where it induces LMP and cell death. This property is a distinctive advantage of these thiosemicarbazones over standard chemotherapy, and is important, as this class of agents has entered multi-center clinical trials in patients with advanced cancers (22,23).
Immunofluorescence -For assessment of intracellular Pgp co-localization, cells (1 x10 5 cells/ml) were grown on coverslips (24 h/37 o C) and incubated for 1 h/37 o C under control conditions or with stress stimuli such as glucose starvation, serum starvation, or H2O2 (100 μM) under normoxia or hypoxia. These steps were similarly carried out for experiments assessing Pgp function, although the cells were then incubated with DOX (10 μM; 2 h/37 °C). For both studies, this procedure was followed by paraformaldehyde fixation (4% (w/v), 15 min/20 °C) and digitonin permeabilization (100 μM, 10 min/20 °C). Importantly, the mild detergent digitonin was utilized to specifically avoid dissolving the lysosomal membrane (65) In the case of the primary incubation with anti-LAMP2 Ab, this was followed by treatment (1 h/25°C) with Alexa Fluorconjugated secondary antibodies (1:1000, Cat. #: A-21200 and A-21201, Invitrogen). The coverslips where then mounted onto DAPI (0.5 μM; Invitrogen). Stained samples were examined using a Zeiss LSM 510 Meta confocal microscope (Zeiss, Oberkochen, Germany) equipped with FITC (excitation, 495 nm; emission, 516 nm) and Texas Red (excitation, 577 nm; emission, 592 nm) filters and captured with Zeiss LSM 510 META software (Zeiss). Fluorescence intensity and the Mander's overlap for image co-localization were measured using ImageJ 4.7v software (National Institutes of Health, Baltimore, MD). Scan parameters were kept constant during acquisition of all images. Minimal image processing (i.e., brightness and contrast only) was carried out on all confocal images presented. Slight adjustments in brightness and contrast were made in images in order to allow low level fluorescent signals to be observed more clearly. However, all image analysis measurements, including co-localization analyses were carried out on raw data. For each slide assessed, 3 separate field of views were used and at least 3 cells were counted per field. The fields of view were chosen randomly with no selection criteria being applied. Each condition was performed at least 3 times independently.

Pulse-Chase Analysis to Assess
Internalization of Pgp -Examination of the internalization of Pgp from the plasma membrane to intracellular compartments was performed via a pulse chase protocol (67) using either: radioisotope labeling, or immunofluorescence.
In initial studies, the uptake of 125 Ilabeled anti-Pgp Ab was examined using KBV1 (+Pgp) cells. These cells were seeded on 35 mm 3 plates and allowed to grow overnight. They were first cooled to 4 o C on ice to limit endocytosis (40) and the 125 I-labeled antibody was then added for 1 h/4 o C on ice. For the 4 o C control condition, cells were washed six times on ice with ice-cold PBS to remove unbound Ab. The cells were then incubated for 30 min on ice with the protease, Pronase (1 mg/mL) to separate membrane-bound proteins from internalized proteins using standard methods (40)(41)(42). For the 1 h control/stress conditions at 37 o C, the labeled cells at 4 o C were washed six times on ice with ice-cold PBS. Then, pre-warmed media was added and incubated for 1 h/37 o C. Cells were then treated with Pronase as above to separate the membrane (Pronase-sensitive) from the internalized (Pronase-resistant) compartment. At the end of the incubation, the cell monolayer was then detached from the plate using a plastic spatula and centrifuged at 14,000 rpm/3 min/4 °C. The supernatant that contains membranebound 125 I was removed and the cell pellet containing internalized 125 I was re-suspended in 1 mL of PBS. The levels of 125 I in both the supernatant and cell suspension were measured using a γ-counter (Wallac Wizard 3, Turku, Finland).
For immunofluorescence studies, KBV1 cells were seeded onto 24 well plates containing coverslips and were left to grow overnight. KBV1 cells were seeded onto 24 well plates containing coverslips and were left to grow overnight. The plates were first cooled to 4 o C on ice to limit endocytosis (40) and the anti-Pgp (Mab 4E3; Abcam; Cat. # ab10333) antibody was then added at 1:100 dilution and incubated for 1 h/4 o C on ice.
For the 4 o C control condition, cells were thoroughly washed three times with ice-cold PBS on ice to remove unbound Ab. This was followed by a further wash with ice-cold PBS. The cells were incubated with 4% (w/v) paraformaldehyde for 10 min at room temperature to enable fixation. Cells were then washed twice with PBS and permeabilized by incubation with 200 µM digitonin.
For the 1 h control/stressors at 37 o C, the plates were thoroughly washed three times with ice-cold PBS on ice. Then, pre-warmed media containing the stressors (control, glucosefree medium, serum-free medium or 100 µM H2O2) was added to the respective wells and incubated for 1 h/37 o C. Cells were washed once with PBS followed by incubation with 4% (w/v) paraformaldehyde for 10 min at room temperature to ensure fixation. Cells were then washed twice with PBS and incubated with 200 µM digitonin.
For both the 0 h control at 4 o C and 1 h control/stressors plate at 37 o C, cells were washed three times with PBS every 5 min. Cells were then blocked with 10% BSA in PBS-T for 30 min and incubated with cathepsin D antibody (Abcam; Cat. # ab72915) at 1:100 dilution overnight at 4 ℃ with agitation. Cells were incubated with both anti-mouse Alexa-Fluor® 488 (green; for Pgp Ab) and anti-rabbit Alexa-Fluor® 594 (red; for Cathepsin D Ab) secondary antibodies at a 1:1,000 dilution for 1 h in the dark with agitation. Cells were then washed three times with PBS every 5 min.
Coverslips were mounted on slides with ProLong Gold DAPI and sealed with nail polish and left to dry. Images were taken using the confocal microscope implemented above and analyzed using ImageJ software.
Plot Profile Analysis to Assess Cellular Internalization of Pgp -Plots were generated from confocal images using the Plot Profile analysis function in ImageJ software (National Institutes of Health). The merged confocal image was split into its individual channels and a line for performing Plot Profile analysis was drawn through the cell, excluding the DAPI-stained nucleus. Cells chosen for analysis were representative of the overall pattern of expression of both cathepsin D and Pgp for each condition. The plot generated displays the intensity of protein expressed as a gray value for each point along the distance of the line measured in μm.
To examine the proportion of Pgp in and outside of lysosomes under various incubation conditions, a "mask" was created from the cathepsin D channel and was applied to the Pgp channel using ImageJ software. Corrected total cell fluorescence was then calculated by established methods (68) for Pgp inside and outside the masked region using ImageJ.

Assessment of Lysosomal Membrane
Permeability -Microscopic visualization of lysosomes was performed using the classical lysosomotropic stain, acridine orange (AO), which accumulates within the lysosomal compartment (69,70). In these studies, KB31 (-Pgp) and KBV1 (+Pgp) cells were pre-incubated under glucose starvation, serum starvation, or H2O2 stress under normoxia or hypoxia (1 h/37 °C), followed by a further incubation under similar conditions for 24 h/37 °C in the presence or absence of Dp44mT (25 μM) or Dp44mT (25 μM) and Ela (0.2 μM). The cells were then washed with serum-free media and further incubated with AO (20 μM; 12 min/37 °C), followed by a second wash. Notably, AO accumulates in lysosomes to induce a red lysosomal fluorescence and a weak green cytosolic fluorescence following blue light excitation. The live-cell samples were examined with a Zeiss Axio Observer.Z1 wide-field fluorescence microscope (Zeiss) equipped with FITC and Texas Red filters. Images were captured with an AxioCam camera and AxioVision Rel. 4.7 Software (Zeiss).
MTT Proliferation Assay -Cellular proliferation was examined using the [3-(4, 5dimethylthiazol-2-yl)-5-(3carboxymethoxyphenyl)-2-(4-sulfophenyl)-2Htetrazolium] (MTT) assay (71). Cells were seeded into 96-well micro-titre plates at 3 x 10 3 cells/well which allowed exponential growth for the duration of the assay. The cells were grown overnight and the agents then added to give concentrations between 0.2 μM to 100 μM. For assessing the impact of MDR inhibitors on cellular proliferation, cells were treated with potent Pgp-inhibitor, Ela (Sigma-Aldrich; 0.2 μM), which was utilized in conjunction with other test agents for 24 h/37 o C.
After this incubation, 20 μL of MTT (5 mg/mL; Sigma-Aldrich) in PBS was added to each well and the plates incubated for 2 h/37 o C. The cells were then solubilized by adding 70 μL of DMSO for 60 min/20 o C, while incubating on a shaking incubator and the plates read at 570 nm on a Victor TM Multilabel Counter plate reader (Perkin Elmer, Waltham, MA). The formation of the insoluble purple formazan was directly proportional to the number of viable cells (72). Concentration-response curves were fitted in Prism 6.0 (Graphpad Software, San Diego, CA) to obtain IC50 values.
Data Analysis -Results were expressed as mean ± SD (3 experiments). Statistical analysis was performed using Student's t-test. Data was considered statistically significant when p < 0.05.

ABBREVIATIONS
3-[4,5-dimethylthiazol-2-yl]-2,5diphenyltetrazolium bromide, MTT; acridine orange, AO; doxorubicin, DOX; dimethyl sulfoxide, DMSO; di-2-pyridylketone 4,4dimethyl-3-thiosemicarbazone, Dp44mT; elacridar, Ela; fluorescein isothiocyanate, FITC; hydrogen peroxide, H2O2; hypoxia-inducible factor-1α, HIF-1α; lysosomal membrane permeabilization, LMP; lysosome-associated membrane protein 2, LAMP2; multi-drug resistance, MDR; murine embryonic fibroblast, MEF; reactive oxygen species, ROS; Pglycoprotein, Pgp; prolyl hydroxylases, PHDs.       (Aii) Schematic showing that DOX is effluxed out of cells by P-glycoprotein (Pgp), but can also be transported into endosomes and lysosomes by Pgp in these organelles (14). Storage of DOX in the lysosome contributes to drug resistance to this agent, as DOX is sequestered away from its molecular targets in the nucleus i.e., lysosomal "safe house" effect (14). (Bi) Line drawing of the structure of Dp44mT. (Bii) Schematic demonstrating that Pgp facilitates Dp44mT transport out of cells and into endosomes/lysosomes (15)(16)(17)19). However, Dp44mT overcomes Pgp-mediated drugresistance by forming copper complexes that potently generate reactive oxygen species (ROS) (15,17,18). Generation of ROS causes lysosomal-membrane permeabilization (LMP) and apoptosis that leads to the death of resistant cancer cells, and thus, overcomes resistance (15,17,18).         Panels i-xii) and KBV1 (+Pgp; Panels xiii-xxiv) cells under normoxia were pre-incubated for 1 h with either control medium, or stressors, namely: glucose starvation, serum starvation, or H2O2 stress (100 µM). Cells were then incubated with Dp44mT (25 µM) in the presence or absence of the Pgp inhibitor, Ela (0.2 µM), in the continued absence or presence of these stressors (under normoxia) for 24 h/37 o C. Lysosomal stability was examined using live cell immunofluorescence imaging of the lysosomotropic fluorophore, acridine orange (AO), which is sequestered and retained in intact lysosomes. At high lysosomal concentrations of acridine orange, an orange fluorescence is visualized, while lower cytosolic and nuclear concentrations produce a green fluorescence. Images are typical of three independent experiments with data analysis in (B) representing mean ± SD (n = 3). Relative to respective KBV1 control: ***p < 0.001; or relative to respective treatment with Dp44mT alone: † †p < 0.01; or relative to respective treatment of KB31 cells with Dp44mT alone: ## p < 0.01; ### p < 0.001. Scale bar: 10 µm Figure 11. Under hypoxia, micro-environmental stressors potentiate Dp44mT-mediated lysosomal damage in both low-and high-Pgp expressing cells. (A) KB31 (very low Pgp; Panels ixii) and KBV1 (+Pgp; Panels xiii-xxiv) cells under hypoxia were pre-incubated for 1 h with either 30 control medium, or stressors, namely: glucose starvation, serum starvation, or H2O2 stress (100 µM). Cells were then incubated with Dp44mT (25 µM) in the presence or absence of the Pgp inhibitor, Ela (0.2 µM), in the continued absence or presence of these stressors (under hypoxia) for 24 h/37 o C. Lysosomal stability was examined using live cell immunofluorescence imaging of the lysosomotropic fluorophore, acridine orange (AO), which is retained within intact lysosomes. At high lysosomal concentrations of acridine orange, a red fluorescence is visualized, while lower cytosolic and nuclear concentrations produce a green fluorescence. Images are typical of three independent experiments with data analysis in (B) representing mean ± SD (n = 3). Relative to respective control: *p < 0.05, **p < 0.01, ***p < 0.001; or relative to the respective treatment with Dp44mT alone: † † †p < 0.001. Scale bar: 10 µm Cellular proliferation was measured using the MTT proliferation assay. Results are typical of three independent experiments with data analysis representing mean ± SD (n = 3). Relative to the respective Dp44mT control: *p < 0.05, **p < 0.01, ***p < 0.001; relative to the respective Dp44mT treatment with Ela: #p < 0.05, ##p < 0.01, ###p < 0.001. Figure 13. Schematic model illustrating the 2 major mechanisms of Pgp regulation by which tumor micro-environmental stressors increase drug resistance. Tumor micro-environment stressors (i.e., serum starvation, low glucose levels, ROS, and hypoxia) induced Pgp-mediated resistance by two mechanisms: (1) re-distribution of Pgp to lysosomes after short-term (1 h) stress; and (2) increased Pgp expression via HIF-1α accompanied by lysosomal biogenesis after long-term (4-24 h) stress. (A) Basal Pgp expression in un-stressed (normoxic) cancer cells; (B) short-term exposure to stressors causes a rapid increase in Pgp re-distribution to lysosomes; (C) long-term exposure to stressors also re-distributes Pgp to lysosomes, but additionally increases Pgp levels via HIF-1α and lysosomal biogenesis. Within the lysosome, Pgp allows import of Pgp substrates into this organelle. (D) From a therapeutic standpoint, stressors decrease drug sensitivity/toxicity (i.e., increases drug-resistance) to DOX via increased Pgp expression and Pgp-mediated lysosomal drug trapping (i.e., lysosomal "safe house" effect). In contrast to DOX, tumor micro-environmental stress increases drug sensitivity/toxicity of Dp44mT towards Pgp-expressing cells. Significantly, the stressors induce Pgp-mediated resistance to DOX, while Dp44mT directly utilizes Pgp to overcome this resistance and kill tumor cells.