NIR Photodynamic Destruction of PDAC and HNSCC Nodules Using Triple-Receptor-Targeted Photoimmuno-Nanoconjugates: Targeting Heterogeneity in Cancer

Receptor heterogeneity in cancer is a major limitation of molecular targeting for cancer therapeutics. Single-receptor-targeted treatment exerts selection pressures that result in treatment escape for low-receptor-expressing tumor subpopulations. To overcome this potential for heterogeneity-driven resistance to molecular targeted photodynamic therapy (PDT), we present for the first time a triple-receptor-targeted photoimmuno-nanoconjugate (TR-PIN) platform. TR-PIN functionalization with cetuximab, holo-transferrin, and trastuzumab conferred specificity for epidermal growth factor receptor (EGFR), transferrin receptor (TfR), and human epidermal growth factor receptor 2 (HER-2), respectively. The TR-PINs exhibited up to a 24-fold improvement in cancer cell binding compared with EGFR-specific cetuximab-targeted PINs (Cet-PINs) in low-EGFR-expressing cell lines. Photodestruction using TR-PINs was significantly higher than the monotargeted Cet-PINs in heterocellular 3D in vitro models of heterogeneous pancreatic ductal adenocarcinoma (PDAC; MIA PaCa-2 cells) and heterogeneous head and neck squamous cell carcinoma (HNSCC, SCC9 cells) containing low-EGFR-expressing T47D (high TfR) or SKOV-3 (high HER-2) cells. Through their capacity for multiple tumor target recognition, TR-PINs can serve as a unique and amenable platform for the effective photodynamic eradication of diverse tumor subpopulations in heterogeneous cancers to mitigate escape for more complete and durable treatment responses.

The approximations of ligands (Cet, HT, TZ) attached on the surface of untargeted-PSNs were derived as described previously [65]. Fluorescence emission (Exc = 480 nm, Emi = 517 nm) of all purified PINs and BPD-PC concentration (nM) within each PINs were used to derive the conjugation efficiency (%) of the ligand (Cet, HT, TZ) to the untargeted-PSN.
Untargeted-PSNs and PINs were characterized with regards to their hydrodynamic diameter (nm), polydispersity index (PDI), and ς-potential (mV) using the Zetasizer Nano ZS Dynamic Light Scattering Instrument (Malvern Instruments, Ltd., Houston, TX, USA). Measurements were performed in triplicates and values were reported as mean and standard deviation.

Cellular Binding of PINs
Single-cell suspensions of 50,000 cells/microcentrifuge tubes were incubated with 250 nM BPD-PC equivalent of untargeted-PSN or PINs formulations in the respective serum-containing culture media at 37 • C for 30 min in the dark. For the approximation of expression levels of EGFR, TfR, and HER-2, MIA PaCa-2 and SCC-9 cells were incubated with 10 µg/mL of AF-conjugated proteins (Cet-AF, HT-AF, or TZ-AF) in the respective serum-containing culture media at 37 • C for 30 min in the dark.
Following incubation, the cells were centrifuged at 1000× g for 5 min and the supernatant was removed. Cell pellets were resuspended in 200 µL of pre-cooled 1× DPBS, agitated 5 times with a pipette to form single-cell suspensions, and transferred to flow cytometry tubes. The fluorescence intensity of cell-associated BPD-PC and Alexa Fluor 488 was measured using the BD FACSAriaTM II flow cytometer (BD Biosciences ® , Woburn, MA, USA). Ten thousand events were recorded and gated for each group using a 405 nm laser and a 610 nm dichroic long-pass filter for BPD and a 450/40 nm filter for AL488. Median BPD-PC emission was quantified using the FlowJo ® software (V10, Franklin Lakes, NJ, USA). Data are presented as mean ± SEM from three biological replicates for each group. Fold improvement in binding with targeting is defined as the cellular binding of a targeted nanoconstruct with respect to the cellular binding of untargeted nanoconstructs and is calculated as Fold improvement = Cellular binding of PINs Cellular binding of untargeted PSNs (2)

In Vitro PINs Internalization Studies
MIA PaCa-2 and SCC-9 cells at 70-90% confluence were seeded in 24-well, glass-bottom, black-walled plates at a density of 1 × 10 5 cells/well. Adherent cells were then incubated for 6 h with Cet-PINs, Cet-TZ-PINs, or TR-PINs formulations at 250 nM BPD-PC equivalent concentration in the respective serum-containing cell media and kept in the dark at 37 • C. Prior to imaging, cells were washed twice with 1× DPBS and were stained with 50 nM LysoTracker ® Red DND-99 (Invitrogen, Carlsbad, CA, USA) at 37 • C in the dark. Hoechst ® 33,342 (Invitrogen, Carlsbad, CA, USA) was used to stain the nuclei of the cells prior to fluorescence imaging. Images were acquired using a confocal microscope (Olympus FluoView-1000 confocal microscope) through a 60× objective (1.2NA, Water). The nuclei, lysosomes, and BPD-PC were visualized using 405 (Hoechst and BPD) and 559 nm (LysoTracker) laser excitation, respectively, with appropriate filters (Hoechst: 425-475 nm; LysoTracker: 580-650 nm; BPD-PC: 655-755 nm).
Flow cytometry was also used for the quantification of intracellular uptake after 6h incubations. Cells were seeded in 24-well, glass-bottom plates at a density of 2 × 10 5 cells/well, incubated with untargeted-PSN or PINs formulations at 250 nM BPD-PC equivalent concentration in media for 6 h, washed twice in 1× DPBS harvested with trypsin, and transferred to flow cytometry tubes following subsequent washing with 1× DPBS and pipette agitation as described before. Ten thousand events were recorded and gated for each group. BPD-PC emission was quantified using the FlowJo ® software (V10, Franklin Lakes, NJ, USA). Data are presented as mean ± SEM from three biological replicates for each group. An increase in cellular uptake of PINs as compared with untargeted-PSN was calculated as Suspended 3D nodules of MIA PaCa-2 and SCC-9 cells were grown and cultured in 96-well, black-walled, round-bottom ultralow attachment plates (Corning ® Costar ® , Corning, NY, USA) at 37 • C. MIA PaCa-2 cells were seeded at a density of 2.5 × 10 3 cells per well and SCC-9 cells were seeded at a density of 5 × 10 3 cells per well for 48 h to self-assemble into single 3D nodules. Nodules were then incubated with untargeted-PSN or PINs formulations at varying concentrations of BPD-PC. After 6 h of incubation, nodules were washed three times with 100ul of the respective serum-containing cellular media and irradiated with 40 J/cm 2 of 690 nm laser light (Intense, North Brunswick, NJ, USA) at an irradiance of 150 mW/cm 2 . At 72 h following photodynamic activation, cells were co-stained with LIVE (Calcein AM; Invitrogen, Carlsbad, CA, USA) and DEAD (propidium iodide) reagents to analyze the viability of treated cells. Prior to staining, nodules for total killing control were fixed using a 10% formalin solution in 1× DPBS (2-4 min) and cell membranes were permeabilized with 0.1% Triton X-100 incubation (60 min) and washed with 0.1 M Glycine (3 times). Nodules were then incubated with calcein AM (Invitrogen, Carlsbad, CA, USA) and Propidium Iodide (Sigma-Aldrich, St. Louis, MO, USA) at standard culture conditions according to the manufacturer's protocol.
Fluorescence signals were recorded using an Olympus FV-1000 confocal microscope through a 0.16NA 4x air objective at λexc = 488 nm/λem = 520 nm (calcein) and λex = 559 nm/λem = 630 nm (PI). Brightfield images were acquired under 559 nm light. The acquisition was standardized for each nodule. All experimental conditions were performed with an n of 8-12 nodules. Comprehensive high-throughput image analysis (CALYPSO) was used to generate heat map images and for quantifying the fractional viability [81].

Design, Preparation, and Characterization of Photoimmuno-Nanoconjugates (PINs)
Untargeted-photosensitizing-nanoconstructs (PSNs) were prepared from anionic DOPGcontaining DPPC liposomes and a lipid-anchored derivative of benzoporphyrin derivative (BPD-PC), as described previously [65,80]. The anionic charge is required to minimize the variability in uptake between multiple cell lines [65]. The untargeted-PSNs also contained DSPE-PEG 2000 with a dibenzocyclooctyle (DBCO) functional group to further allow for the covalent conjugation of the targeting ligands through copper-free click chemistry. Liposomal nanoconstructs hold great promise as drug delivery vehicles for emerging treatment regimens due to their ability to carry multiple payloads that can be tuned with regard to their hydrophilicity or hydrophobicity. Furthermore, their ability to incorporate multiple surface-targeting ligands of varying natures with finely tunable surface densities is a particularly important attribute required for precision medicine.
The conjugation efficiency of the individual ligands bound to the surface of the untargeted-PSNs was quantified by labeling cetuximab (Cet) with Alexa Fluor 488, holo-transferrin (HT) with Alexa Fluor 647, and trastuzumab (TZ) with Alexa Fluor 680. Untargeted-PSNs and photoimmuno-nanoconjugates (PINs), including HT-targeted PINs (HT-PINs), TZ-targeted PINs (TZ-PINs), Cet-targeted PINs (Cet-PINs), both Cet-and TZ-targeted PINs (Cet-TZ-PINs), and triple-receptor-targeted PINs (TR-PINs) (Figure 1), exhibit an average hydrodynamic size of 130.57 ± 9.2, and polydispersity index (PDI) of 0.06 ± 0.01, which is suggestive of a narrow size distribution and monodisperse nanoconstructs. The constructs all exhibited a ζ-potential between −16.7 and −18.6 mV, demonstrating that an anionic charge is maintained in all PINs prepared (Table 1). A consistent ζ-potential is important for minimizing variability in uptake that is not associated with the nature of the targeting ligand or ligands.

Cellular Binding Specificity of Photoimmuno-Nanoconjugates (PINs)
We have recently shown for the first time that our chemically tuned NIR light-activated Cet-PINs targeted to a single receptor, EGFR, selectively binding, permeating, and destroying tumor cells in a 3D heterocellular pancreatic ductal adenocarcinoma (PDAC) model more efficiently than untargeted-PSNs [65]. In this study, we have further modified the design of Cet-PINs to direct the construct towards additional tumor-associated receptors (HER-2 and TfR) that are over-expressed in several cancers including PDAC and HNSC. The cellular binding was measured by the quantitation of the BPD-PC fluorescence intensity from the nanoconstructs. A431 (high EGFR) [82], T47D (high TfR) [83,84], SKOV-3 (high HER-2) [85,86], and CHO-WT (EGFR null) [87] cells were incubated with untargeted-PSNs or targeted PINs (Cet-PINs, HT-PINs, TZ-PINs, Cet-TZ-PINs, TR-PINs) to determine the cellular binding specificity using flow cytometry. Cet-PINs, HT-PINs, and TZ-PINs exhibit higher cellular association in high-receptor-expressing cancer cells than the untargeted-PSNs ( Figure 2). Figure 2. Cellular binding specificity of photoimmuno-nanoconjugates (PINs) to tumor cells over-expressing EGFR, TfR, and HER-2 receptors. Flow cytometry histograms and bar graphs representing the specificity of PINs conjugated to the tumor-specific ligands cetuximab (to target EGFR), holo-transferrin (to target TfR), or trastuzumab (to target HER-2). Binding specificity of (a) Cet-PINs to A431 cells (high EGFR), (b) HT-PINs to T47D cells (high TfR), and (c) TZ-PINs to SKOV-3 cells (high HER-2) is presented with respect to the untargeted-PSNs for each cancer cell line and the control CHO-WT cell line (null for EGFR, TfR, HER-2). (mean ± S.E.M.; unpaired t-test, n = 3 for each cell line; **** = p ≤ 0.0001).
As is consistent with our previous findings [65], Cet-PINs improved nanoconstruct binding to A431 cells (high EGFR) by 24-fold (Figure 2a), as compared with untargeted-PSNs. Although elegant prior work has shown that liposomal Foscan ® targeted with transferrin exhibited no cellular specificity [71], our HT-PINs demonstrated an 8-fold improvement in T47D cell (high TfR) binding, as compared with untargeted-PSNs ( Figure 2b). This discrepancy with the prior work is most likely due to the nanoconstruct membrane-stabilizing effect that lipid anchoring of BPD has in our studies, that prevents the non-specific transfer of the photosensitizer when the construct is not targeted [65,80]. TZ targeting also improved the binding of TZ-PINs to SKOV-3 cells (high HER-2) by 13.5-fold ( Figure 2c). As expected, no significant binding of Cet-PINs, HT-PINs, or TZ-PINs to CHO-WT cells was observed due to the absence of expression of all three receptors [87,88] (Figure 2a-c).

Triple Receptor Targeting Enhances PIN Binding and Cellular Uptake in MIA PaCa-2 PDAC Cells and SCC-9 HNSCC Cells
We hypothesize that heterogeneous tumors such as PDAC and HNSCC exhibiting diverse patterns of tumor-associated cell surface receptors (EGFR, TfR, HER-2) over-expression, can be selectively targeted using PDT directed against EGFR, TfR, and HER-2 concurrently. TR-PINs would enable the specific recognition of multiple cell surface targets and would increase the specificity of drug delivery and treatment efficacy in heterogeneous tumor environments, thereby ultimately mitigating treatment escape.
The advantages of triple-receptor targeting are not limited to only an enhanced diversity of cancer cell surface binding (Table 3). This strategy also significantly increases the ability of PDAC and HNSCC tumor cells to internalize TR-PINs in vitro. It was found that the simultaneous targeting of EGFR, HER-2, and TfR receptors demonstrate significantly higher cellular binding of TR-PINs (Figure 4b), relative to the EGFR, TfR, and HER-2 over-expression in MIA PaCa-2 and SCC-9 cells. Triple-receptor targeting resulted in 41-fold (MIA PaCa-2 cells) and 33-fold (SCC-9 cells) improvements in binding with targeting when compared with the untargeted-PSNs. Furthermore, a 77% (MIA PaCa-2) and 80% (SCC-9) increase in binding was observed with TR-PINs in comparison with Cet-PINs ( Figure 4b).
As expected, no notable changes in cellular binding in CHO-WT cells were observed using TR-PINs due to the lack of expression of all three receptors (Figure 4b).  The subcellular localization of PINs in MIA PaCa-2 and SCC-9 cells was observed using confocal microscopy. Cells were incubated with PINs and nuclei and lysosomes were stained after 6 h of incubation. All PINs were found to localize to endo-lysosomal compartments, exhibiting punctate intracellular BPD-PC signals in MIA PaCa-2 ( Figure 5a) and SCC-9 cells, respectively (Figure 5b). This is consistent with our previous findings for BPD-PC nanoconstructs [65,94]. Intracellular uptake of PINs was quantified using flow cytometry following 6 h of incubation with MIA PaCa-2 and SCC-9 cells. The trend in the uptake levels at 6 h incubation correspond to the cellular binding of PINs and demonstrate a significant increase in uptake of TR-PINs in MIA PaCa-2 and SCC-9 cells, compared with Cet-PINs and Cet-TZ-PINs. Quantitation of BPD-PC fluorescence intensities using flow cytometry demonstrate a 1.5-fold (45%) and 1.7-fold (73%) increase in cellular uptake of Cet-TZ-PINs and TR-PINs, respectively, in MIA PaCa-2 cells, as compared with Cet-PINs. Further, a 1.2-fold (24%) and 1.4-fold (39%) increase in cellular uptake of DR-PINs and TR-PINs, respectively, was observed in SCC-9 as compared with Cet-PINs. These results suggest that triple targeting enables the TR-PINs to bind and internalize in MIA PaCa-2 and SCC-9 cells more efficiently with respect to Cet-PINs, delivering higher levels of intracellular BPD-PC for molecular-targeted PDT.

Singlet Oxygen Measurements
As the PINs are nanosystem-designed for effective PDT-mediated killing, they must retain their ability to generate cytotoxic reactive molecular species, such as singlet oxygen ( 1 O 2 ), when functionalized with various targeting ligands. 1 O 2 is the predominant cytotoxic molecular species produced during the photosensitization of BPD and its lipid-anchored derivatives [1,80].
To monitor photogenerated 1 O 2 from NIR-activated PINs, two 1 O 2 probes, SOSG and DADB, were used in this study. While oxidation of SOSG with 1 O 2 increases the probe's fluorescence, the endoperoxide photooxidation product of DADB is non-fluorescent and exhibits a decay in the fluorescence intensity upon reaction with 1 O 2 . Further, DADB is lipophilic and partitions in the phospholipid bilayer of BPD-PC nanoconstructs, probing the immediate production of 1 O 2 [80,95], whereas SOSG is a membrane-impermeable probe that measures global 1 O 2 in the entire solution. We have confirmed in our previous study that BPD-PC fluorescence is negligible at the wavelengths used to monitor DADB (505 nm) and SOSG (525 nm) emission signals, confirming that both probes are appropriate for measuring 1 O 2 production from BPD-PC nanoconstructs [80]. Figure 6a shows a light dose-dependent increase in the SOSG fluorescence intensity following laser light irradiation. Fluorescence intensity of the SOSG in the mixture (PINs + SOSG) increased with increasing light doses (0 J/cm 2 -100 J/cm 2 ), representing the generation of 1 O 2 and conversion of SOSG to its photo-oxidized product. No difference in the rate of photogenerated 1 O 2 (as measured by increased emission of SOSG) was observed following the irradiation of the untargeted-PSNs, Cet-PINs, and TR-PINs, (Figure 6a) suggesting that the global average of 1 O 2 in the solution is unaltered by the degree of ligand functionalization. However, when 1 O 2 generation was measured only in the hydrophobic membrane compartments of the PSNs and PINs using DADB (Figure 6c), differences were observed. The relative rate of 1 O 2 production was calculated using the equation described in Section 2.7. A 1.5-fold higher rate of 1 O 2 production (DADB fluorescence decay (0.09%/J cm −2 )) was observed with both the untargeted-PSNs and Cet-PINs, as compared with the TR-PINs (0.06%/J cm −2 ) (Figure 6d). The TR-PINs have a total of 89.6 ligands per construct, whereas the Cet-PINs have a total of 27.1 ligands per construct. Given that 1 O 2 also reacts with aromatic amino acids that are abundant in the surface-bound targeting ligands [96], this increased number of membrane surface ligands from Cet-PINs to TR-PINs might explain the 1.5-fold reduction in the rate of 1 O 2 production in the TR-PIN membrane, as determined by the DADB measurements.

NIR Light-Mediated Photodynamic Treatment of PDAC and HNSCC Monocellular and Heterocellular 3D Models of Heterogeneity
As shown earlier, the TR-PINs exhibit expanded cancer cell binding specificities and enhanced cellular uptake in MIA PaCa-2 and SCC-9 cells and have the potential for simultaneously targeting heterogeneous tumor subpopulations in PDAC and HNSCC. As such, we further evaluated the NIR phototoxicity of the TR-PINs in PDAC (MIA PaCa-2) and HNSCC (SCC-9) 3D nodules with varying cell surface receptor expression levels of EGFR, TfR, and HER-2. Considering that EGFR over-expression is prevalent in PDAC and HNSCC, we compared the NIR phototoxicity of Cet-PINs with TR-PINs (specific for the additional receptors HER-2 and TfR) in the MIA PaCa-2 and SCC-9 3D nodules. Treatment efficacy was also evaluated in T47D and SKOV-3 3D nodules as a control for low-EGFR-expressing cells. Furthermore, T47D and SKOV-3 cells (low EGFR) were included in PDAC (MIA PaCa-2) and HNSCC (SCC-9) 3D nodules to recapitulate heterogeneous tumor cell subpopulations that would evade EGFR-targeted Cet-PINs. Targeted PDT efficacy was also evaluated in the heterocellular 3D models of heterogeneity.
Firstly, the NIR phototoxicity of Cet-PINs and TR-PINs was assessed in monocellular 3D nodules of MIA PaCa-2 and SCC-9. The nodules were incubated for 6 h with untargeted-PSNs, Cet-PINs, or TR-PINs at 0-2000 nM BPD-PC equivalent, 48 h after seeding in round-bottom ultralow attachment plates as described in the Experimental Section. The nodules were then irradiated with 40 J/cm 2 of 690 nm laser light at an irradiance of 150 mW/cm 2 . This time point was selected based on our previous study showing that Cet-PINs exhibited the highest level of specificity in 3D nodules at 6 h incubation time [65]. At 72 h following PDT treatment, the nodules were co-stained with LIVE (Calcein AM) and DEAD (propidium iodide) reagents prior to single-plane confocal imaging. For viability assessment of the 3D nodules in each experimental group, quantitative fractional viability heatmap images (Figure 7a,c) were generated using a comprehensive image analysis procedure for structurally complex organotypic cultures (CALYPSO) [81]. Untargeted-PSNs did not show any significant phototoxicity even at the highest concentration of 2000 nM of BPD-PC equivalent (Figure 7b,d). However, at a concentration of 2000 nM of BPD-PC equivalent, TR-PINs were most effective, and significantly reduced the viability of SCC-9 nodules to 45% and MIA PaCa-2 nodules to 24% (Figure 7b,d). Cet-PINs were equally effective at all concentrations of BPD-PC equivalent in the MIA PaCa-2 nodules and SCC-9 nodules.
MIA PaCa-2 nodules were more responsive to targeted PDT than the SCC-9 nodules at all concentrations of BPD-PC equivalent (Figure 7b,d). Importantly, in the absence of photoactivation, neither Cet-PINs nor TR-PINs exerted any toxic effects on cancer cells ( Figure S2) at the concentration range of 0-2000 nM of BPD-PC equivalent, which is consistent with our previous findings using Cet-PINs [65].
Triple-targeted PDT using TR-PINs was equally as effective as single-receptor EGFR-targeted Cet-PINs in the EGFR over-expressing MIA PaCa-2 and SCC-9 nodules. However, in the low-EGFR-expressing control nodules that over-express TfR (T47D) and HER-2 (SKOV-3), the TR-PINs were significantly more effective than the EGFR-targeted Cet-PINs (Figure 8a, Figure S3). The T47D nodule viability decreased by 67% after PDT with the TR-PINs (500 nM of BPD-PC equivalent), which was significantly more effective than the Cet-PINs. The SKOV-3 nodule viability decreased by 24% after PDT with the TR-PIN, whereas the Cet-PINs were ineffective at the same concentration (500 nM of BPD-PC equivalent). These T47D and SKOV-3 nodules represent low-EGFR-expressing tumors that would typically escape single-receptor EGFR-targeted PDT but would respond to the TR-PINs we report in this study.
As discussed earlier, tumors comprise of heterogeneous cells with various receptor expression profiles. Thus, even though a large proportion of tumor cells can be eradicated by single-receptor-targeted therapy, low-receptor-expressing subpopulations may persist and contribute to tumor recurrence. As such, we attempted to recapitulate heterogeneous PDAC and HNSCC tumors in vitro by forming heterocellular 3D models. MIA Paca-2 and SCC-9 nodules were formed with the addition of low-EGFR-expressing T47D or SKOV-3 cells that represent tumor subpopulations which we have shown to evade EGFR-targeted PDT using Cet-PINs (Figure 8a). While T47D and SKOV-3 cells express low levels of EGFR, treatment escape can be circumvented using TR-PINs by exploiting the over-expression of TfR in T47D cells and HER-2 in SKOV-3 cells. PDT treatment response in the heterogeneous heterocellular 3D nodules was then evaluated using Cet-PINs and TR-PINs ( Figure 8c) and was compared with the treatment response in the monocellular 3D nodules. While the efficacy of TR-PINs was identical to that of the EGFR-targeted Cet-PINs in EGFR-over-expressing MIA PaCa-2 and SCC9 nodules, the TR-PINs were significantly more effective than the Cet-PINs in the heterogeneous heterocellular 3D nodules (Figure 8b,d). In heterocellular MIA PaCa-2 nodules containing SKOV-3 cells, TR-PINs provided a 17% greater reduction in viability than the Cet-PIN, and a 25% greater reduction in viability in heterocellular MIA PaCa-2 nodules containing T47D cells (Figure 7b). In heterocellular SCC9 nodules containing SKOV-3 cells, TR-PINs provided a 34% greater reduction in viability than the Cet-PIN, and a 14% greater reduction in viability in heterocellular MIA PaCa-2 nodules containing T47D cells (Figure 8d). would typically escape single-receptor EGFR-targeted PDT but would respond to the TR-PINs we report in this study. As discussed earlier, tumors comprise of heterogeneous cells with various receptor expression profiles. Thus, even though a large proportion of tumor cells can be eradicated by single-receptortargeted therapy, low-receptor-expressing subpopulations may persist and contribute to tumor recurrence. As such, we attempted to recapitulate heterogeneous PDAC and HNSCC tumors in vitro by forming heterocellular 3D models. MIA Paca-2 and SCC-9 nodules were formed with the addition of low-EGFR-expressing T47D or SKOV-3 cells that represent tumor subpopulations which we have shown to evade EGFR-targeted PDT using Cet-PINs (Figure 8a). While T47D and SKOV-3 cells express low levels of EGFR, treatment escape can be circumvented using TR-PINs by exploiting the over-expression of TfR in T47D cells and HER-2 in SKOV-3 cells. PDT treatment response in the heterogeneous heterocellular 3D nodules was then evaluated using Cet-PINs and TR-PINs ( Figure  8c) and was compared with the treatment response in the monocellular 3D nodules. While the efficacy of TR-PINs was identical to that of the EGFR-targeted Cet-PINs in EGFR-over-expressing MIA PaCa-2 and SCC9 nodules, the TR-PINs were significantly more effective than the Cet-PINs in the heterogeneous heterocellular 3D nodules (Figure 8b,d). In heterocellular MIA PaCa-2 nodules containing SKOV-3 cells, TR-PINs provided a 17% greater reduction in viability than the Cet-PIN,

Discussion
Intratumoral heterogeneity can limit the efficacy of therapies directed towards single tumor cell surface receptors. Diverse patterns of tumor-associated cell surface receptor expression (EGFR, TfR, HER-2) have been found in PDAC and HNSCC patients. Specifically, studies have reported the over-expression of EGFR and HER-2 in patient pancreatic cancer tissue (45-95% and 43-69%, respectively) [60,97] and patient head and neck squamous cell carcinoma tissue (up to 90% and 68%, respectively) [25,98,99]. Thus, EGFR and HER-2 are both attractive targets for molecular targeted activatable therapies, such as PDT. However, one study using PDAC patient tissue found that HER-2 over-expression was concurrent with EGFR over-expression in 24% of patients, and in patients with no EGFR expression, no HER-2 over-expression was observed [100]. These findings emphasize that, in a clinical setting, targeting EGFR and HER-2 simultaneously using Cet-TZ-PINs may not be sufficient for effective and complete tumor eradication in all patients, and thus the TR-PINs we present here, with an expanded specificity for a third tumor receptors are critical. Considering that TfR over-expression has been found in PDAC [24] and HNSCC [32] and has also been found to play a prominent role in cancer cell proliferation, it is an important additional target for our TR-PIN-mediated PDT approach [44].
Thus, targeting three tumor receptors simultaneously may promote complete tumor eradication using spatiotemporally controlled activatable therapies such as PDT. Moreover, the emerging and promising use of bi-specific and multi-specific monoclonal antibodies (mAbs) directed towards multiple receptors is an additional motivation, which we further advance in the field by leveraging high-payload nanosystems [101,102]. In this study, we have developed NIR-activable triple-receptor-targeted photoimmuno-nanoconjugates (TR-PINs) with three ligands, conjugated to a single photosensitizing nanoconstruct to simultaneously target heterogeneous tumor cell subpopulations with differential expression levels of EGFR, TfR, and HER-2. These TR-PINs are proposed to increase the specificity and effective photodynamic eradication of tumor subpopulations in heterogeneous cancers with differential receptor expression levels.
We have assessed the cellular binding of the nanoconstructs by the quantitation of BPD-PC fluorescence emission using flow cytometry. Our results show that the cellular binding with TR-PINs was superior to both the Cet-PINs and Cet-TZ-PINs in tumor cells (A431, MIA PaCa-2, SCC-9, SKOV-3, T47D) with different origins and varying expression levels of tumor cell surface receptors. TR-PINs exhibit varying binding specificities to tumor cells (Table 3), thus the fold improvement in the cellular binding of TR-PINs (with respect to untargeted-PSN constructs) was found to be significantly higher than the calculated sum of the fold-improvements in binding with each PIN, with respect to untargeted-PSN constructs. This is likely due to the combined effect of the multiple ligands when conjugated on the surface of a single nanoconstruct to target multiple receptors simultaneously. This multiplicative increase in binding is likely due to multi-avidity effects, which can be advantageous; however, multi-target specificity remains to be the priority for targeting heterogeneity in this study.
The advantages of triple-receptor targeting are not only limited to enhanced surface binding to a greater proportion of heterogenous cancer cell subpopulations. This strategy also significantly increases the ability of PDAC and HNSCC tumor cells to internalize TR-PINs in vitro. PINs were found to localize to endo-lysosomal compartments when incubated with cells. Prior work using similar constructs also shows lysosomal localization of liposomal BPD-PC [94]. The staining with DAPI and lysotracker for confocal images does not show any membrane-bound constructs. Considering that, by 6 h, most of the PINs would have been internalized, as receptor-mediated endocytosis occurs from as early as 20 min following incubation [103], our prior work [65] corroborates the observations in this study whereby Cet-PINs were observed on the membrane of MIA PaCa-2 cells at 1 h incubation, but no membrane binding was observed in MIA PaCa-2 or OVCAR-5 cells at 6h following internalization. We also previously explored the relative binding and penetration of the Cet-PINs through 3D nodules [65]. Binding specificity of the Cet-PINs in 3D nodules of MIA PaCa-2 was quantified at different time points (1 h, 6 h, 24 h). Cet-PINs exhibited the highest binding specificity (12.5-fold) for the MIA PaCa-2 3D nodules with respect to untargeted constructs at 6 h after incubation. In the current study, the fold improvement in binding with Cet-PINs is 23-fold compared with the untargeted-PSN in single-cell suspensions of MIA PaCa-2 cells. As such, it is apparent that the selectivity in uptake decreases from 2D to 3D cultures. These observations require further investigation before definitive and generalizable conclusions can be drawn.
We further evaluated the NIR phototoxicity in MIA PaCa-2 and SCC-9 nodules. TR-PINs are equally as effective as Cet-PINs in the MIA PaCa-2 and SSC-9 monocellular 3D nodules, even though the TR-PINs exhibit 1.7-fold and 1.4-fold higher cellular uptake, respectively. The absence of further improved PDT efficacy by triple targeting is likely to be a result of the 1.5-fold quenching of 1 O 2 in TR-PINs, as compared with Cet-PINs, when measured using DADB (Figure 6d). The results suggest that there is a minimum threshold in targeted cellular uptake after which the triple targeting will become more effective than the single targeting. These findings also suggest that improving the overall outcome of TR-PINs in heterogeneous tumors, as compared with Cet-PIN, may only become evident when the differences in the heterogeneous receptor expression exceed the 1.5-fold quenching of 1 O 2 observed.
As described earlier, tumors are heterogeneous masses of cancer cells with differential receptor expression profiles. Single-receptor-targeted therapies treat tumors with an assumption that they are a homogenous mass of cancer cells, and thus are generally only able to eradicate a specific proportion of tumor cells. The residual surviving tumor cells are thus likely to persist, proliferate aggressively, and promote tumor relapse and progression. Thus, the response to single-receptor-targeted therapies can be highly variable in these heterogeneous tumors. To address this heterogeneity-driven resistance to targeted therapies, we attempted to recapitulate heterogeneity in vitro, by forming 3D heterocellular models of MIA Paca-2 and SCC-9 cells, with the addition of low-EGFR-expressing T47D or SKOV-3 cells, that represent tumor subpopulations which we have shown to evade EGFR-targeted PDT using Cet-PINs (Figure 8a). While T47D and SKOV-3 cells express low levels of EGFR, treatment escape can be circumvented using TR-PINs by exploiting the over-expression of TfR in T47D cells and HER-2 in SKOV-3 cells. PDT treatment response in the heterogeneous heterocellular 3D nodules was then evaluated using Cet-PINs and TR-PINs ( Figure 8c) and was compared with the treatment response in the monocellular 3D nodules. While the efficacy of TR-PINs was identical to that of the EGFR-targeted Cet-PINs in EGFR-over-expressing MIA PaCa-2 and SCC9 nodules, the TR-PINs were significantly more effective than the Cet-PINs in the heterogeneous heterocellular 3D nodules (Figure 8b,d). In heterocellular MIA PaCa-2 nodules containing SKOV-3 cells, TR-PINs provided a 17% greater reduction in viability than the Cet-PINs, and a 25% greater reduction in viability in heterocellular MIA PaCa-2 nodules containing T47D cells (Figure 8b). In heterocellular SCC9 nodules containing SKOV-3 cells, TR-PINs provided a 34% greater reduction in viability than the Cet-PINs, and a 14% greater reduction in viability in heterocellular MIA PaCa-2 nodules containing T47D cells (Figure 7d).
The significance of these findings is that they demonstrate that triple-receptor targeting using TR-PINs does provide more complete photodestruction of heterogeneous tumor nodules, which would otherwise partially evade single-receptor EGFR-targeted PDT. More complete PDT responses in heterogeneous tumors would thereby mitigate treatment escape, recurrence, and resistance to targeted therapy. Furthermore, TR-PINs would potentially be effective in a broader range of PDAC and HNSCC patients, in addition to patients with various other cancer indications where EGFR, HER-2, and TfR over-expression are implicated.

Conclusions
In this study, we show for the first time that heterogeneous heterocellular 3D models of PDAC and HNSCC can be more effectively destroyed using triple-receptor-targeted TR-PINs (EGFR-, HER-2-, and TfR-specific) that would otherwise partially evade single-receptor EGFR-targeted PDT. The significance of these findings specifically for PDT is that heterogeneous tumor subpopulations can also be effectively targeted using the TR-PINs, irrespective of the increase in cellular binding of the TR-PINs. PDT dosimetry can be modulated by tailoring the light dose applied, and thus, while cellular delivery is important, molecular specificity towards heterogeneous tumor cell subpopulations and discrimination between tumor tissue and healthy tissue remains critical.
Future work will further explore the encapsulation of multiple treatment modalities within a single TR-PIN construct. As such, heterogeneous tumor subpopulations would be simultaneously targeted with multiple treatment regimens that exhibit non-overlapping modes of cytotoxicity. Furthermore, PDT-based regimens using the TR-PIN platform will be evaluated in complex heterogeneous in vivo tumor models, such as patient-derived xenografts, to further mitigate the risk of treatment escape that often leads to tumor recurrence after an initial response.