The effect of subcellular localization on the efficiency of EGFR‐targeted VHH photosensitizer conjugates

Graphical abstract Figure. No caption available. Abstract Photodynamic therapy (PDT) is an emerging method to treat light‐accessible malignancies. To increase specificity and allow dose reduction, conjugates of photosensitizers (PS) with antibodies against tumor‐associated antigens have been developed for photoimmunotherapy (PIT). However, so far it is unclear whether cellular internalization of these conjugates after binding affects PIT efficacy. The use of low molecular weight llama single domain antibodies (VHHs, nanobodies) for PIT is preferred above full size antibodies because of better tumor penetration. Therefore, we functionalized the VHH 7D12, directed against the epidermal growth factor receptor (EGFR), with a PS (IRDye700DX). To assess the impact of cellular internalization on activity, the VHHs were additionally conjugated to a cell‐penetrating peptide (VHH[PS]‐CPP). Here we show that upon illumination with near‐infrared (NIR) light, both VHH[PS] and VHH[PS]‐CPP conjugates specifically induce cell death of EGFR expressing cancer cell lines and of EGFR‐expressing cells derived from surgically obtained ascites from patients with high‐grade serous ovarian cancer. However, VHH[PS] conjugates were significantly more effective compared to internalizing VHH[PS]‐CPP suggesting that cell surface association is required for optimal therapeutic activity.


Introduction
Photodynamic therapy (PDT) is a method to induce cell death through administration and activation of a photo-sensitizer (PS). When activating the PS with light of the appropriate wavelength, the PS is transferred from its ground state into an excited triplet state [1] that can return to the ground state via transmitting its energy to molecular oxygen, leading to the formation of reactive oxygen species (ROS). In general, these ROS are formed through type II photo-oxidative reactions that result in the formation of singlet oxygen, a highly toxic and shortlived radical that induces peroxidation and breakdown of lipids, proteins and nucleic acids [2].
For tumors that are amenable for local light application, PDT has a number of advantages as compared to other therapies. It is less invasive than surgery and, because of its local character, more selective than chemotherapy. Furthermore, the direct cell killing effects of PDT prevent development of resistance, as is seen with chemotherapy and most targeted therapies [3] and may also induce a vaccination effect because PDT-induced necrotic cell death releases neo-epitopes that may challenge the immune system [4]. PDT has been tested in clinical trials for cancer of the bladder, skin, head and neck. Various PSs are now approved as PDT drugs [5].
There are, however, some issues with PDT that still need to be solved. Currently used PSs suffer from low water solubility and dark toxicity. To enhance solubility, significant efforts have been made in engineering drug delivery systems that allow tumor-specific targeting of PS [6][7][8][9][10][11][12]. Even though antibodies are highly potent targeting vehicles, antibody-PS conjugates for photoimmunotherapy (PIT) [13,14] have the disadvantage that they circulate for weeks, increasing the risk of dark toxicity and phototoxicity in light-exposed skin [15]. circulation time, conjugates of PS with small-sized recombinant llama antibodies (VHHs) have been used to induce epidermal growth factor receptor (EGFR)-dependent cell death [14,16]. The small size of these delivery systems also allows faster tumor accumulation and better tissue penetration as compared to full-size antibodies, and improves the carrier-to-drug molecular weight ratio [17].
The half-life of singlet-oxygen in biological systems is < 40 ns, restricting its toxic action radius to < 20 nm [2]. Studies using photosensitizers with different physicochemical properties that behave differently with respect to cell uptake have shown that intracellular localization greatly influences the cellular response to light-induced activation. PSs that localize to the mitochondria or cytoplasm are described to induce apoptosis, while necrosis is induced when the plasma membrane is the site of action [18,19]. Furthermore, more efficient membrane binding and membrane photooxidation increases phototoxicity of the PS [20].
Next to relying on internalization of targeted receptors, internalization of VHHs can also be triggered by conjugation to cell-penetrating peptides (CPPs) [21], a class of peptides that mediate cellular uptake of molecules that otherwise do not enter the cell [22]. Conjugation of the anti-EGFR VHH 7D12 to the CPP hLF, derived from human lactoferrin [23], endows the VHH with the capacity to enter the cell [24].
To investigate how the efficacy of PIT is influenced by the subcelullar localization of the PS we used a site-selective bioconjugation protocol to functionalize the EGFR targeting VHH 7D12 [25,26] with the PS IRDye700DX. Using sortase A transpeptidation we conjugated hLF to VHH 7D12 [PS] , yielding 7D12 [PS] -hLF. We then investigated the efficacy of both constructs on light-induced cell death and show that the CPP-free non-internalizing variant is more active in inducing cell death.

Sortase A expression
E. coli ER2566 cells transformed with the plasmid pGBMCS-SortA, encoding His 6 -tagged sortase A with an N-terminal deletion of 59 amino acids [30] (Addgene, Cambridge, MA, USA, plasmid #21931) were grown to log phase at 37°C, and protein expression was induced with 1.0 mM isopropyl β-D-thiogalactoside (IPTG, Serva, Heidelberg, Germany) at 30°C for 3 h. Cells were harvested, resuspended in ice-cold 50 mM Tris pH 8.0/300 mM NaCl containing a protease inhibitor cocktail (Roche, Basel, Switzerland) and lysed by sonication using a Bandelin Sonopuls HD2070 sonicator (Bandelin electronic GmbH & Co, Berlin, Germany). After centrifugation the supernatant was incubated with Ni-NTA sepharose (IBA, Goettingen, Germany) in 50 mM phosphate pH 7.4/500 mM NaCl for 1 h at 4°C. After washing, His 6 -tagged sortase A was eluted from the beads with 500 mM imidazole. The eluate was dialyzed against 50 mM Tris pH 7.5/150 mM NaCl in a 3.5 kDa dialysis membrane (Spectrum Labs, Los Angeles, CA, USA). Protein purity was analyzed by SDS-PAGE gel electrophoresis under reducing conditions and liquid chromatography mass spectrometry (LC-MS, Shimadzu HPLC and Thermo Finnigan LCQ Fleet) on a C4 column. Protein concentration was determined by absorbance at 280 nm using the Nanodrop spectrophotometer (LI-COR, Lincoln, NE, USA).

VHH expression
The cDNA encoding anti-EGFR VHH 7D12 (a gift from Paul van Bergen en Henegouwen, Utrecht University, The Netherlands) was recloned into the vector pHENIX-C-LPETG-His 8 -Vsv, resulting in 7D12-C-LPETG-His 8 -Vsv, the cysteine providing a handle for maleimide conjugation, LPETG being a sortase A consensus recognition sequence and the His 8 -Vsv tag allowing Ni-based purification and Vsv-based detection. 7D12-C-LPETG-His 8 -Vsv expression was induced via standard methods in E. coli strain TG1. Cells were harvested by centrifugation at 2830g for 20 min at 4°C, resuspended in ice cold Tris/EDTA/sucrose (TES) buffer (200 mM Tris pH 8.0/0.5 mM EDTA/20% w/v sucrose/ protease inhibitor cocktail) and incubated for 20 min on ice, followed by centrifugation (4424g, 20 min, 4°C). The supernatant was collected and the pellet was resuspended in TES buffer supplemented with 15 mM MgSO 4, and incubated on ice for 20 min. After centrifugation, both supernatants were pooled and protein was purified with Ni-NTA sepharose as described for sortase A.
In parallel, IRDye700DX-NHS (LI-COR, Lincoln, NE, USA from now on referred to as PS-NHS) was incubated with H 2 N-PEG 3 -N 3 (Jena Bioscience, Jena, Germany) in a 3:1 molar ratio in 100 mM phosphate buffer pH 8.6/150 mM NaCl for 7 h at RT in a thermoshaker at 450 rpm, yielding PS-N 3 . Subsequently 7D12-C [DBCO] -LPETG-His 8 -Vsv was incubated with PS-N 3 in a 1:2 molar ratio o/n at RT in a thermoshaker at 450 rpm, yielding 7D12-C [PS] -LPETG-His 8 -Vsv. This conjugate was purified from unconjugated PS-N 3 by centrifugation in a 10 kDa MWCO centrifugal unit employing 4 washing cycles with 50 mM Tris pH 7.5/ 150 mM NaCl. The protein was analyzed by SDS-PAGE gel electrophoresis under reducing conditions and with electrospray ionization time-of-flight mass spectrometry (ESI-ToF) on a JEOL AccuTOF-CS (JEOL, Tokio, Japan). Protein concentration was determined by measuring absorbance at 689 nm (ε = 165,000 M −1 cm −1 , assuming 1:1 complete labeling) using the Nanodrop spectrophotometer.

Functionalization of 7D12-C [PS] -LPETG-His 8 -Vsv with hLF
GGG-hLF or GGG-hLF (Fluo] (synthesized by EMC microcollections, Tübingen, Germany) was dissolved in 50 mM HEPES pH 8.0 at 5-10 mM, to ensure intramolecular disulfide bridge formation that is required for CPP activity [31]. The carboxyfluorescein moiety was coupled to a C-terminal lysine residue. Subsequently sortase A (80 µM final concentration) and GGG-hLF (100 µM final concentration) were added to 7D12-C [PS] -LPETG-His 8 -Vsv (20 µM final concentration) in 50 mM Tris pH 7.5/150 mM NaCl/10 mM CaCl 2 . The sortase reaction was allowed to proceed for 5 h at 30°C in a thermoshaker at 450 rpm, after which the reaction mixture was incubated with pre-washed Ni-NTA sepharose beads to remove the His 6 -tagged sortase A, G-His 8 -Vsv and unreacted 7D12-C [PS] -LPETG-His 8 -Vsv. Excess GGG-hLF was removed from the Ni-NTA agarose supernatant by filtration in a 10 KDa MWCO centrifugal unit employing 6 cycles against 20 mM phosphate pH 7.5/500 mM NaCl. Purity and concentration of the VHH conjugate were analyzed by SDS-PAGE gel electrophoresis and ESI-ToF. Protein concentrations were determined by measuring absorbance at 689 nm using the Nanodrop spectrophotometer.

Cell uptake assays with confocal microscopy
Cellular uptake of the different fluorescein-labeled 7D12 conjugates was examined with confocal microscopy on a TCS SP5 microscope (Leica Microsystems, Mannheim, Germany) equipped with an HCX PL APO 63 × 1.2 water immersion lens. During imaging, cells were maintained at 37°C. The 488 nm laser line of the argon laser was used for excitation and emission was collected between 500 and 550 nm. A431, E98 and SK-OV-3 cells (30,000, 50,000 and 30,000 cells per well, respectively) were plated in 8-well Lab-Tek borosilicate coverglass chambers (NUNC, Thermo Fisher Scientific, Waltham, MA, USA) and allowed to adhere and proliferate for 48 h. Then cells were incubated for 30 min at standard culture conditions in 200 µl phenol red-free RPMI/10% FCS, supplemented with 2 µM of 7D12-C [Fluo] -LPETG-His 8 -Vsv, 7D12-C-LPETGGG-hLF [Fluo], or GGG-hLF [Fluo] . Subsequently, cells were washed twice with phenol red-free RPMI/10% FCS containing 20 mM HEPES and were imaged by confocal microscopy. Acidic pH in lysosomes reduces fluorescein fluorescence. To investigate if this effect lead to an underestimation of 7D12-hLF [Fluo] uptake, we treated cells with 65 μM chloroquine, which increases the lysosomal pH and thereby enhances fluorescein fluorescence [32]. The samples were imaged once more and, to quench extracellular fluorescence, 0.4% trypan blue (Sigma Aldrich, Saint Louis, MO, USA) was added to the wells before imaging for a third time [33].

Study of cell uptake mechanism with confocal microscopy
To detect clathrin-dependent endocytosis, cells were co-incubated with the constructs as described above and 100 µg/ml transferrin-Alexa Fluor 633 (Life Technologies, Thermo Fisher Scientific, Waltham, MA, USA). After 30 min, the cells were washed and colocalization of transferrin and the constructs was detected by confocal microscopy. Fluorescein was excited with the argon laser at 488 nm and emission was detected between 500 and 550 nm. Alexa Fluor 633 was excited with the 633 nm helium-neon laser and fluorescence detected between 650 and 715 nm.

In vitro PDT assays with adherent cell cultures
A431, E98 and SK-OV-3 cells were cultured in clear 96-well plates (Costar, Greiner-Bio One, Essen, Germany). At 80% confluency cells were incubated for 30 min with different concentrations of 7D12-C [PS] -LPETG-His 8 -Vsv, 7D12-C [PS] -LPETGGG-hLF or equimolar concentrations of PS alone in DMEM/10% FCS. Controls were incubated with DMEM/10% FCS only. Cells were washed twice with warm DMEM/ 10% FCS. Immediately after washing, plates were illuminated with 100 mW/cm 2 for 600 s, reaching a total light dose of 60 J/m 2 , using a standardized light emitting diode device (690 ± 10 nm) as described in [34]. To determine dark toxicity, cells were incubated with the highest used concentration of the conjugates without subsequent illumination. After overnight incubation a sulforhodamine-B-assays (SRB) assay was performed as described in [35] to determine total protein content. Results were expressed as cell viability relative to controls (untreated illuminated cells). Half maximal inhibitory concentration (IC50) of the various conjugates were determined in GraphPad Prism 5.02 (LaJolla, CA, USA).
To examine selectivity of PDT-induced cytotoxicity with the different conjugates, 5 × 10 5    were illuminated with a 60 J/m 2 total light dose. After overnight incubation at 37°C, spheres were fixed in Unifix (Klinipath, Duiven, The Netherlands) at RT, and embedded in agar. 4 µM sections were stained with hematoxylin and eosin.

Ex vivo PDT assays with clinical ascites samples
All experiments with patient materials were performed according to institutional guide lines. To examine whether PIT with our conjugates could be an option for treatment of ovarian cancer we tested our constructs on cells, freshly derived from malignant ascites of patients diagnosed with high-grade serous ovarian carcinoma. Ascites was filtered through a 70 µm cell strainer. Propidium iodide and bound antibodies were visualized using the EVOS microscope.

Statistics
Experiments were performed at least in duplicate, and within experiments all measurements were done in triplicate. IC50 values were determined in Graphpad Prism and statistical significance was checked with a Student's T-test. To check for significance of other data a oneway ANOVA with post hoc Bonferroni was performed in Graphpad Prism; * = p < .05, ** = p < .01, *** = p < .001.

Preparation and characterization of 7D12-C [PS] conjugates
Targeted PDT of cancers requires high tumor selectivity and specificity, concomitant with a lack of dark toxicity. Approaches that improve tumor selectivity of the PS by conjugation to antibodies still have major drawbacks due to the long circulation half-lives of these conjugates, and poor penetration of these large constructs into poorly perfused tumor areas. To tackle these problems, smaller targeting moieties like VHHs are interesting alternative PS carriers. Here, the anti-EGFR VHH 7D12 was used as a clinically relevant model VHH to examine how subcellular localization of PSs affects PDT efficacy. It was published before that 7D12-IRdye700DX conjugates can be prepared successfully via NHS-based conjugation to lysines [14]. This approach however, carries a risk of overlabeling of lysines that are involved in antigen binding, resulting in reduced affinity. Also, this procedure yields a heterogeneous mixture of VHHs, containing a fraction of unlabeled VHHs that may act as competitor for the labeled ones. We therefore chose to perform a site-selective reaction at an introduced cysteine at the carboxyterminus (Fig. 1A). Because IRDye700DX was not available as a maleimide conjugate at the time of this study, we first coupled a dibenzocyclo-octyn (DBCO)-functionality to the VHH and an azide functionality to the PS. This enabled us to use the highly specific and bio-orthogonal click reaction between azide and DBCO [36]. 7D12-C [DBCO] -LPETG-His 8 -Vsv and 7D12-C [PS] -LPETG-His 8 -Vsv were produced successfully as verified with SDS-PAGE electrophoresis (Fig. 1B) and LC-MS (Fig. 1C). LC-MS verified conjugation of only one maleimide-DBCO and PS molecule per VHH, as we have seen before for maleimide-fluorescein [37]. This method is generally applicable to all VHHs, as long as these do not carry unpaired cysteines in the complementarity determining region, and since all reactions occur distant from the VHHs' antigen binding site, this approach is predicted to retain VHH affinity.

Cellular binding and uptake of 7D12 conjugates
Previous research has shown that hLF is taken up by endocytosis at concentrations below 10 µM [23], and that functionalization of 7D12 with hLF causes increased internalization of the VHH [24]. Here we confirmed these patterns of cellular uptake in A431 and SK-OV-3 cells (Fig. 2). Cells were incubated with the conjugates at a concentration of 2 μM at which little vesicular uptake of hLF alone was observed (supplementary Fig. 1A). 7D12-C [Fluo] -LPETG-His 8 -Vsv showed a predominant membrane staining whereas for 7D12-C-LPETGGG-hLF [Fluo] membrane staining was weaker and a more pronounced vesicular fluorescence was observed. SK-OV-3 cells showed very weak membrane staining after incubation with 7D12-C [Fluo] -LPETG-His 8 -Vsv, but after incubation with 7D12-C-LPETGGG-hLF [Fluo] intracellular vesicular staining was observed, which was more clearly visible after quenching extracellular fluorescence with trypan blue (Fig. 2). No binding to or uptake in E98 was observed for both 7D12 conjugates (supplementary Fig. 1B). This latter finding further confirmed that cell association was primarily VHH and not hLF driven.
It is generally accepted that CPPs lack cell line selectivity, however, such experiments are typically conducted at medium micromolar concentrations [38]. The dissociation constant of the binding of hLF to cell surface glycosaminoglycans is in the low micromolar range [31], while reported Kd values of 7D12 binding to cell-associated EGFR is in the order of 10-20 nM [14,17,26]. Cell ELISAs, performed at 4°C to study binding in the absence of internalization, confirmed high affinity binding of 7D12-C [PS] -LPETG-His 8 -Vsv and 7D12-C [PS] -LPETGGG-hLF to EGFR overexpressing A431 cells (Kd = 11.85 ± 1.203 and 25.53 ± 2.514 nM, respectively; not shown). At the low micromolar and subnanomolar concentrations that we employed in the uptake and PDT assays, respectively, binding of the different conjugates is therefore expected to be determined by the 7D12 moiety, rather than hLF. This finding has important implications for in vivo applications of CPP-based strategies. Except for those tissues that show a high propensity for CPP uptake [39], uptake of VHH-CPP conjugates is directed by the presence of EGFR. In this case, the CPP module is a modulator of subcellular trafficking rather than an unspecific driver of uptake.

PIT efficacy of 7D12-C [PS] constructs in adherent cell cultures
Cell killing assays showed that 7D12-C [PS] conjugates were very potent and specific PDT agents, without inducing dark toxicity (Fig. 4A and C). Incubation with PS alone, followed by illumination did not induce cell killing under the conditions used (Fig. 4B). Although the 7D12-C [PS] conjugates that we produced via a two-step click reaction contained only one PS per VHH, these induced EGFR-specific cell death very efficiently, with an IC50 value for 7D12-C [PS] -LPETG-His 8 -Vsv on A431 cells of 87.8 ± 4.3 pM after NIR light illumination with 60 J/m 2 .   Interestingly, in previous reports it was demonstrated that VHH-PS conjugates with higher affinity for EGFR had higher PIT efficacy and it was suggested that increased internalization due to higher affinity was responsible for increased photosensitivity [14]. Furthermore, internalizing antibody-PS conjugates were shown to be more effective in PIT than non-internalizing antibody-PS [41]. Our data, however, show that increased cell surface association is a more important determinant. Interestingly, no phototoxicity was induced in non-EGFR expressing E98 cells in line with the low cell binding capacity mediated by the CPP alone.
To further assess the selectivity of the 7D12-C [PS] conjugates, co-culture experiments with A431, E98 and SK-OV-3 were performed. These cell types were distinguished by pre-labeling them with different membrane associated dyes. Labeled and dead cells were visualized immediately, 2 h and 16 h after illumination ( Fig. 4D and supplementary Fig. 2). Only EGFR expressing cells were killed after PDT (Fig. 4D).
No cell death was observed in the controls (no PS-conjugates added, data not shown). Parallel experiments with monocultures of the cells showed that cell death was induced immediately after illumination for the A431 cells, and after 2-16 h for the SK-OV-3 cells (supplementary Fig. 3). The cell death of A431 immediately after PDT suggested that necrosis was induced. Neither A431 and SK-OV-3 cells were stained with Annexin-V (not shown), and they did not induce caspase-3 dependent apoptosis at 4 h after PDT, which further indicates that cells indeed died by necrosis (supplementary Fig. 4). Unconjugated IR-Dye700DX did not lead to phototoxicity in all used cell types, in agreement with previous studies, and did not display dark toxicity. This  makes it a suitable compound for targeted PDT [42].

Efficient PIT with ovarian carcinoma SK-OV-3 spheroids and EGFR positive cells in clinical ascites
To assess effectiveness of the conjugates in a cellular model with more resemblance to the in vivo tumor situation, conjugates were incubated with SK-OV-3 spheroids. PIT with SK-OV-3 spheroids showed efficient cell killing after incubation with 120 nM of the 7D12-C [PS] -LPETG-His 8 -Vsv construct and a light dose of 60 J/m 2 , illustrated by the majority of cells with picnotic nuclei in the HE stained sections, and only a minority of cells in the centre that appeared viable (Fig. 5A, enlargement in insert). In the 7D12-C [PS] -LPETGGG-hLF treated cells, also dead cells with picnotic nuclei were observed but to a lesser extent. Control spheres showed only viable cells, indicating that cell death also in the centre is due to penetration of the VHH conjugates into the spheroid.
Finally, cells derived from clinical ascites from women with stage III or IV high-grade serous ovarian carcinoma (n = 7) were incubated with 15 nM 7D12-C [PS] constructs and illuminated with 60 J/m 2 NIR light. Phototoxicity was restricted to the EGFR-positive subpopulation of cells, as could be observed by costainings for dead cells (propidium iodide) and EGFR (Cetuximab). The absence of costaining for dead cells and a population of cells positive for another tumor marker (EpCAM) indicated selectivity for EGFR (Fig. 5C). SRB assays after incubation with 7D12 [PS] constructs and illumination showed significant decreases in overall cell viability compared to controls (P < .001 and P < .03 for 7D12-C [PS] -LPETG-His 8 -Vsv and 7D12-C [PS] -LPETGGG-hLF, respectively (Fig. 5D). The survival of about 60-70% of the cells is in agreement with the notion that ascites also contain numerous non-EGFRexpressing cells (as shown in Fig. 5C) that are expected to survive the treatment.
With the use of optical fibers NIR light can be delivered to otherwise inaccessible tumors, as has been shown for peritoneal metastasis of ovarian cancer using Verteporfin [43]. In addition, peroperative fluorescent image-guided surgery is gaining momentum [44] and could be very well extended with compounds comparable to those described here.

Conclusion
Targeted PDT is a rapidly increasing field which promises to become of great importance for the treatment of light-accessible malignancies. Targeted light-induced cancer cell death has great advantages as compared to currently applied targeted treatments which mostly only delay progression of cancers, giving these ample time to develop resistance. We here describe the construction of a highly defined VHH-PS via a controlled bioconjugation method that is easily applicable to other VHHs. Despite the fact that every VHH carries only one PS molecule, these constructs are highly selective and specific, requiring very low concentrations to induce efficient cell killing via necrosis. In 3Dspheroid models, efficient cell killing was observed, indicating that these low-molecular weight constructs have good tissue penetrating properties. Furthermore, we show that conjugation to a CPP endows the VHH-PS with an increased capacity to enter the cell via clathrinmediated endocytosis at the expense of a slightly reduced affinity. These changed properties decreased PDT efficacy, suggesting that association with the cell membrane is needed for optimal therapeutic activity.