Engineering photodynamics for treatment, priming and imaging

Abstract

Photodynamic therapy (PDT) is a photochemistry-based treatment approach that relies on the activation of photosensitizers by light to locally generate reactive oxygen species that induce cellular cytotoxicity, in particular for the treatment of tumours. The cytotoxic effects of PDT are depth-limited owing to light penetration limits in tissue. However, photodynamic priming (PDP), which inherently occurs during PDT, can prime the tissue microenvironment to adjuvant therapies beyond the direct PDT ablative zone. In this Review, we discuss the underlying mechanisms of PDT and PDP, and their application to the treatment of cancer, outlining how PDP can permeabilize the tumour vasculature, overcome biological barriers, modulate multidrug resistance, enhance immune responses, increase tumour permeability and enable the photochemical release of drugs. We further examine the molecular engineering of photosensitizers to improve their pharmacodynamic and pharmacokinetic properties, increase their molecular specificity and allow image guidance of PDT, and investigate engineered cellular models for the design and optimization of PDT and PDP. Finally, we discuss alternative activation sources, including ultrasound, X-rays and self-illuminating compounds, and outline key barriers to the clinical translation of PDT and PDP.

Introduction

Photodynamic therapy (PDT) is a photochemistry-based treatment modality that involves the light activation of photosensitizers to generate reactive oxygen species (ROS), such as singlet oxygen (¹O₂), and free radicals¹. ROS can be cytotoxic, deleterious or biomodulatory to biological targets, depending on the light dose and photosensitizer concentration², and they can thus be used for the treatment of diseases. PDT was first approved by the US Food and Drug Administration (FDA) for bladder cancer, followed by indications in ophthalmology, dermatology and oncology³. Although PDT has been clinically proven as a minimally invasive treatment for certain local tumours (such as, prostate cancer, brain cancer and bladder cancer) (Fig. 1a) with high spatiotemporal control and minimal-to-no side effects, its widespread clinical implementation for cancers and metastatic diseases remains limited. This limitation is likely due to various factors, including perceptions from studies using first-generation photosensitizers with limited tumour specificity, training barriers to bring optical technologies (for example, lasers) into mainstream oncology, insufficient knowledge (and misconceptions) about the biomodulatory effects of PDT beyond the tissue ablation zone, the need for preclinical models to optimize PDT dosimetry and study its biological responses, and the complicated regulatory pathway for what is considered a combination product (drug and device). Bioengineering approaches can address some of these concerns to help integrate PDT into clinical workflows. Here, the science and clinical application of using photodynamic activation to achieve intended biological responses in target cells and tissues is referred to broadly as ‘photodynamics’.

Fig. 1 |. Photodynamic therapy.

a, Major advances in photodynamic therapy (PDT) and related engineering approaches since the first application of PDT in patients in 1976²²⁰. The following references are cited in the timeline: refs. 17,53,61–63,65,66,76,164,220–222. b, Examples of PDT agents include monoclonal antibody–photosensitizer (mAb–PS) conjugates (photoimmunconjugates)¹⁷, F(ab’)₂–PS conjugates⁶⁶, nanobody–PS conjugates⁷⁶, virus-like particle drug conjugates⁸⁸, porphysomes⁶³, photoactivable multi-inhibitor nanoliposomes⁶¹, photoimmunconjugate nanoliposomes^62,86 and photoimmunconjugate nanoparticles⁸⁷. c, Endoscopic ultrasound (EUS)-guided Visudyne-PDT procedure in a patient with locally advanced pancreatic cancer⁵⁵. The patient first receives an intravenous administration of Visudyne (liposomal verteporfin commercialized by Novartis AG; 0.4 mg per kg) at 60–90 min before the EUS procedure. Under EUS guidance, a diffuser is then placed into the tumour, delivering light at 50 J cm⁻¹ for 333 s. Computed tomography images are obtained after PDT to identify tumour necrosis⁵⁵ and photodynamic priming (PDP) effects⁵⁴. d, General procedure in clinical PDT, including formulation selection and administration routes of photosensitizers, illumination procedures, technologies involved in photoactivation and image-guidance in treatment planning, monitoring, and evaluation. AK, actinic keratosis; ALA, 5-aminolevulinic acid; AMD, age-related macular degeneration; F(ab’), fragment antigen-binding region; HpD, haematoporphyrin derivative; LED, light-emitting diode; PCI, photochemical internalization; PIT, photoimmunotherapy.

Multifunctionality can be imparted on photosensitizers, for example, by tuning and regulating their pharmacokinetic and pharmacodynamic behaviour, adding molecular specificity for tumour tissue, improving their transport across cellular barriers, and enabling stimuli-responsive controlled release with image guidance. In 2000, Visudyne (a liposomal formulation of verteporfin, Novartis AG), along with diode lasers at 689 nm (Coherent and Zeiss lasers), was approved by the FDA as the first liposomal photosensitizer delivery system for macular degeneration⁴. In 2020, the first antibody–photosensitizer conjugate, cetuximab saratolacan sodium⁵, with a near-infrared laser system (BioBlade, Rakuten Medical Inc.), was approved by the Ministry of Health, Labour and Welfare in Japan for photoimmunotherapy of head and neck cancer (Fig. 1a). Iterations and amalgamations of nanoscale systems and molecule-targeted nanoconstructs (Fig. 1b) may further improve the clinical translation of PDT (Fig. 1c,d).

Cancer therapy should consider resistance mechanisms, physical tumour barriers, microenvironmental factors and immunologic heterogeneity to improve treatment responses. Photodynamic priming (PDP) is a collateral phenomenon of PDT beyond the direct PDT ablative zone caused by the exponential decrease of light intensity on its way through tissues (although the incident optical energy drops to ~37% within a few millimetres, diffused red light penetrates up to several centimetres). PDP can thus transiently alter the tumour microenvironment, including modulating vascular permeability, cellular processes and subcellular compartments of tumour and non-tumour cells, as well as altering extracellular matrix components and the local tumour immune landscape (Fig. 2). Therefore, PDP is a powerful tool to control vascular permeability⁶, target drug transporter proteins⁷, modulate evolutionary processes in disease progression⁶, remodel the tumour stroma^8,9 and potentially modulate the immune landscape¹⁰. PDP may thus enable other treatment modalities (such as, chemotherapy, radiation and immunotherapy) and potentially reverse chemoresistance. To identify and optimize photodynamic modalities in preclinical settings, high-throughput heterocellular models have been developed that recapitulate key physiological and architectural cues of disease (Fig. 3a); for example, 3D and organoid cultures enable high-throughput dosimetry analyses, mechanistic studies and evaluation of treatment response (Fig. 3b).

Fig. 2 |. Manipulation of tissue architecture and microenvironment by photodynamics.

Light irradiation of solid tumours results in a gradient of fluences that decrease exponentially along the tumour cross-section. A zone of acute tumour tissue death arises in the areas of high light fluences. Beyond that zone of photodynamic therapy (PDT)-induced tissue death, a finite photodynamic priming (PDP) volume exists, which is characterized by transient priming events, such as loosening of stroma and permeabilization of vasculature, as well as mechanistic sensitization to adjuvant therapies and immunological modulation. The combined effects of PDT and PDP lead to local tumour destruction and allow the control of locoregional and distant metastases through the abscopal effect. DAMPs, damage-associated molecular patterns; ICI, immune checkpoint inhibitor; RTKi, receptor tyrosine kinase inhibitor.

Fig. 3 |. Models and analysis of photodynamic therapy.

a, Tissue-engineered 3D cellular models can be used to optimize photodynamic therapy in vitro, including monotypic and heterotypic spheroid cultures, architecturally complex tumouroids grown on 3D-printed matrix substrates, tumouroids containing immunological subtypes (T cells), and organoids derived from patient tumours. b, Biomaterials can be utilized to develop 3D cellular models. Spheroids can be cultured in ultra-low-attachment microwell plates. Substrates such as Matrigel, collagen, alginate and 3D-printed hydrogels can be used to develop matrix-embedded models, including architecturally complex tumouroids and immunologically active tumouroids. c, High-throughput image analysis workflows can be applied to evaluate and optimize photodynamic therapy dosimetry and combinations, including LIVE/DEAD imaging, comprehensive high-throughput image analyses for architecturally complex tumouroids, volumetric imaging of matrix-embedded tumouroids, assessment of photosensitizer microdistribution and imaging of optical redox ratios. Part c adapted with permission from ref. 223, International Society for Optics and Photonics; adapted from ref. 101, Springer Nature Limited; adapted with permission from ref. 224, American Chemical Society; and adapted with permission from ref. 108, Elsevier.

Effective PDT also requires the coordination of photosensitizers, light sources and oxygen at the target tissue. The choice of photonics device depends on the disease location, treatment schedule, type of photosensitizer and many other variables. To improve the consistency and safety of PDT, implantable photonic devices are being developed to treat large and disseminated malignancies, whereas image-guided strategies can inform the timing and dosimetry of PDT. However, the limited penetration depth of visible light in tissue confines PDT to superficial tumours. Deep-seated lesions thus have to be treated by multiple interstitial illuminators and intraoperative illumination. Alternatively, materials can be applied that can be excited by other type of energy sources, such as, ionizing radiation, ultrasound, chemiluminescence or bioluminescence, and Cerenkov radiation. If energy can be delivered in a minimally invasive manner to activate low-cost, off-label, clinically approved photosensitizers using patient-customized applicators and affordable, transportable light delivery systems, PDT also becomes well-suited as a low-cost treatment modality in resource-limited settings (Box 1).

Box 1 |. Photodynamic therapy in resource-limited settings.

By 2030, 75% of global cancer deaths are expected to occur in low-income and middle-income countries²²⁵. Therefore, user-friendly technologies for cancer treatment are required that are portable, allow battery-powered operation, ensure cost-effectiveness and enable telemedicine integration. Photodynamic therapy (PDT) is well-suited to meeting these requirements, through the use of low-cost light sources and handheld fluorescence imaging tools integrated with technologies such as smartphones and artificial intelligence, and well-tolerated topical or oral photosensitizer formulations.

For example, the PDT Brazil program introduced PDT for nonmelanoma skin cancer treatment in over 100 clinics across Brazil, Bolivia, Chile, Ecuador, El Salvador, Colombia, Cuba, Mexico, Peru and Venezuela^226,227. Through industry partnership, an affordable, simple-to-use light-emitting diode (LED)-based irradiation device (LINCE device, MM Optics LTDA) was disseminated to participating clinics. This portable device includes handheld probes for treatment (630-nm LED array) and fluorescence-based monitoring of protoporphyrin IX accumulation and photobleaching. Using this approach, an 80% cure rate of lesions could be achieved for patients treated since 2012²²⁷.

Over 135,000 new cases of oral cancer occur annually in India²²⁸. A low-cost, portable, battery-operated system was developed for image-guided PDT of oral cancer^229–232. This platform is based on a cost-effective fibre-coupled LED for PDT, coupled with smartphone-based image guidance using an LED array over the camera sensor²²⁹ or an attached intraoral probe²³⁰. Ergonomic light delivery is achieved using 3D-printed intraoral applicators, which also standardize beam size and dosimetry²³¹. Using this PDT device, complete and durable tumour response was reported after a single PDT session in 22 out of 30 patients (NCT03638622), with no serious adverse events, scarring, fibrosis or loss of function²³².

Understanding photosensitizer and photonic device design considerations is key to the optimization and clinical translation of PDT for advanced cancers (Box 2). In this Review, we discuss photodynamics, in particular PDP, and the molecular engineering of photosensitizers, with an emphasis on design considerations that cover compartmentalization, physical entrapment and chemical linkage, photophysical and photochemical alteration, degradation, and controlled drug release. We further highlight PDP as a powerful clinical tool for tumour modulation at the molecular, cellular, tissue microenvironmental and immunological level. We also examine engineering strategies for photonic devices and alternative light sources, as well as image-guided and radiomics methods for the personalization of PDT, and provide a roadmap for their integration into clinical protocols. Finally, we investigate materials chemistry, image analysis and smartphone interface approaches, which are particularly relevant in resource-limited settings.

Box 2 |. Regulatory and translational considerations.

Photodynamic therapy (PDT) is typically regulated as a combination product that includes a pharmaceutical and a medical device used for light activation. In the USA, this regulatory pathway involves pre-market approval and is executed by the US Food and Drug Administration (FDA). The dosimetry for PDT and photodynamic priming (PDP) is complex and translationally challenging, involving the photosensitizer agent, potential excipients (for example, nanoscale carriers) and light irradiation parameters. Although photosensitizers are generally considered non-cytotoxic and safe in the absence of light, explicit toxicological evaluation of photosensitizer doses and the associated concentration of excipients are required. With respect to light dosimetry, clinical dosimetry values must explicitly specify the wavelength of the light source, the total light dose in J cm⁻² and the irradiance in mW cm⁻². Therefore, preclinical research should identify and report these parameters.

Importantly, intraluminal, interstitial and/or endovascular light irradiation need to be safe. In particular, multimodal and multiwavelength light sources can provide real-time feedback on treatment efficiency (for example, drug localization and implicit dosimetry). The diverse therapeutic and biomodulatory effects of PDT and PDP may be achieved using alternative activation mechanisms, such as ultrasound, radiotherapy and radioluminescence. However, the mechanisms of sonodynamic therapy, radiodynamic therapy and X-ray-activated PDT or PDP remain unclear and require further elucidation.

Three-dimensional models can enable translationally relevant evaluation of PDT, PDP and combination therapies. For example, desmoplastic models can be applied to investigate PDP, and (vascularized) flow models can be used to study resistance, drug delivery and enhanced permeability and retention.

Patients with locally advanced or recurrent diseases may be selected for PDT safety and feasibility studies³. However, such clinical studies are not designed to investigate the full potential of PDT and its long-term effects, in particular with regard to PDP. Preclinical studies using large animal models may also have an important role in evaluating the safety, efficacy and long-term effects of PDT-based therapies, and can be applied to train surgical personnel in the application of PDT and PDP.

Photodynamic priming

Tumour ablation is the first and highest priority for PDT. However, as light traverses through the tissue, the irradiances and fluences of light drop below the threshold required to ablate tumour tissue. In areas that extend beyond the direct ablative zone of PDT, sub-cytotoxic PDT regimens inadvertently exhibit a range of tumour-modulating phenomena, which are referred to as PDP⁶. PDP effects prime the tumour microenvironment (vasculature, stroma, immune landscape and oxygenation) in a way that may synergize with adjuvant modalities, including immune checkpoint therapy, sensitization to radiotherapy, modulation of tumour tissue and stroma for improved drug delivery, and photochemical internalization for endosomal escape. Such subablative priming mechanistically sensitizes tumour tissue that survives beyond the PDT ablative zone (Fig. 2).

PDT can serve as an adjuvant to a number of modalities, including chemotherapy, receptor tyrosine kinase inhibitors, biologics, immune checkpoint therapies and radiation therapy. For example, PDT can be applied as an adjuvant for chemotherapeutics, combining the tumour-ablative effects of PDT with the systemic cytotoxic and antiproliferative effects of chemotherapy¹¹. Such PDT–chemo combinations are likely to benefit from PDP, which can increase the chemo-sensitivity of tissue that survives beyond the direct ablative zone of PDT. In addition, PDP can be applied to maintain the integrity and perfusion of tumour vasculature to enhance subsequent systematically administered therapies. However, the effects of PDP on tumour tissue that had received subablative regimens of PDT need to be thoroughly investigated to design the sequence of light activation and administration of concomitant secondary therapies, such as chemotherapy.

Upon light activation, the photosensitizer generates ROS that can oxidize or destroy biomolecules, such as anti-apoptotic proteins (B-cell lymphoma 2 (BCL-2) and B-cell lymphoma-extra large (BCL-xL))^12,13 and drug efflux transporters^7,14 responsible for the development of multidrug resistance. PDP can also increase vascular permeability and alter the tumour extracellular matrix, improving drug transport to the tumour site and sensitizing cancer cells to chemotherapeutic agents. Moreover, oxidative stress and biomolecule oxidation triggered by PDP can lead to immunogenic cell death responses, promoting the release of tumour-associated antigens and anti-tumour immune responses.

Priming the tumour vasculature

Vascular responses to PDT (such as, permeability and occlusion) are elicited by short drug–light intervals (that is, the time interval between photosensitizer administration and subsequent light activation), which ensure that photosensitizers are in circulation during activation. Vascular responses are also dictated by the PDT dose, with low doses typically inducing vascular permeability, and high doses typically inducing vascular occlusion and ablation. For example, PDT using a 15-min drug–light interval can impede blood flow 6 h after treatment, leading to a 36% reduction in tumour regrowth rates in the RIF-1 mouse fibrosarcoma model¹⁵. By contrast, a drug–light interval of 3 h does not result in a therapeutic effect. Similarly, vascular perfusion, tumour vessel density and growth of KLN205 mouse lung tumours and LM8 mouse osteosarcoma tumours can be reduced by applying PDT in a 15-min drug–light interval, whereas treatment with a 3-h drug–light interval is less efficacious¹⁶. Vascular responses are also specific to the photosensitizer. Using a photoimmunoconjugate, a drug–light interval of 192 h delays tumour growth and increases survival in an M-1 myosarcoma mouse model with respect to animals undergoing light activation immediately after photoimmunoconjugate administration¹⁷.

Vascular effects of PDT have first been described using fluorescent spheres with diameters ≥100 nm, which show increased tumour extravasation in mouse models of colon cancer following 2-(1-hexyloxyethyl)-2-devinyl pyropheophorbide-a PDT. This phenomenon has been utilized to improve the efficacy of liposomal doxorubicin¹⁸. In general, liposomes can be applied for PDT-controlled chemotherapy; for example, chemotherapeutics, such as doxorubicin and irinotecan, can be delivered by porphyrin–phospholipid liposomes to increase their local concentration in the tumour and improve their efficacy following light activation in preclinical animal models^19–21. This ‘chemophototherapy’ approach initially reduces blood flow in the tumour; this effect is followed by a transient increase in blood flow and blood flow stasis^19,20,22. The acute drop in blood flow may be associated with immediate vascular permeability, followed by narrowing of the blood vessels and an increase in flow during thrombosis as well as subsequent vascular occlusion. The efficacy of chemophototherapy is influenced by the drug–light interval; shorter drug–light intervals of 0.5–1 h are typically more efficacious than longer drug–light intervals^20,21 owing to the vascular presence of porphyrin–phospholipid liposomes early after administration. However, these effects may differ across cancer models²⁰, and the native vascularization and vessel diameters may play a role in vascular permeabilization by PDT and PDP²³. To fully take advantage of vascular permeabilization by porphyrin–phospholipid liposomes, a short drug–light interval can initially be applied to induce vascular permeability, followed by a longer drug–light interval to give porphyrin–phospholipid liposomes sufficient time to accumulate in tumours. This two-step treatment improves tumour growth inhibition and survival in a pancreatic cancer xenograft mouse model²⁴.

Arginylglycylasparatic acid (RGD)–ferritin conjugates loaded with photosensitizers target the vasculature, improving vascular permeability and albumin extravasation, thereby augmenting the efficacy of liposomal doxorubicin in a mouse colon cancer model¹⁸. The vasculature can also be targeted with cationic liposomes^25,26; for example, cationic liposomes delivering a chemotherapeutic may increase vascular permeability owing to their specific adhesion to neoangiogenic vessels, resulting in necrosis of the microvascular endothelium²⁷. Combination of low-dose PDT (20 J cm⁻²) and doxorubicin using cationic doxorubicin-loaded porphyrin–phospholipid liposomes achieves synergistic inhibition of tumour growth in a MIA PaCa-2 mouse model²². Interestingly, the delivery of high doxorubicin payloads to tumour vascular endothelial cells does not reduce tumour growth, whereas treatment with cationic porphyrin–phospholipid liposomes devoid of doxorubicin is effective in reducing tumour growth in the MIA PaCa-2 mouse model, suggesting that vascular permeability is not the sole contributor to the enhanced doxorubicin efficacies observed after PDT. Therefore, non-ablative PDP modulation of tumour tissue, reduced stromal cell densities and extracellular matrix remodelling may also impact the efficacy of chemotherapy.

Overcoming drug delivery barriers

Low-dose PDT can improve the delivery and distribution of liposomal chemotherapies in colorectal tumours¹⁸ and lung tumours²⁸ in rodents through vascular permeabilization. PDP modulation of vasculature is also relevant in combination therapies^29,30. Here, the recruitment of leukocytes is essential to the endothelial permeabilizing effects of photochemistry³¹. PDP can also shift human brain microvascular endothelial cell junctions (vascular endothelial-cadherin and zonula occludens-1) from a continuous phenotype to a more punctate and perpendicular phenotype in vitro, resulting in increased endothelial permeability to nanoscale liposomes and macromolecules³². In addition to vascular modulation, PDP can decrease tumour interstitial fluid pressure in vivo³³, soften the extracellular matrix and increase nanoparticle penetration in 3D co-culture models³⁴, overcome tumour desmoplasia by modulating tumour collagen content and the extracellular matrix in vivo⁸, enable chemotherapy dose-reduction in mouse models³⁵, and enhance the cytotoxicity of radiation therapy in 3D tumour nodules³⁶. Improved permeability of the cancer microenvironment likely stems from direct oxidation of extracellular matrix proteins, such as collagen and fibronectin³⁷. PDP can also reduce tumour collagen deposition and alignment in a pancreatic ductal adenocarcinoma mouse model⁹. Tumour collagen content and extracellular matrix density can further be reduced by PDP in mouse prostate tumour models, improving nanoparticle penetration⁸.

Modulation of multidrug resistance

Efforts to overcome cancer multidrug resistance using small-molecule inhibitors of drug efflux transporters have largely failed in the clinic owing to systemic toxicities. PDP can also inhibit ATP-binding cassette drug efflux transporters (such as, ATP-binding cassette sub-family G member 2 (ABCG2) and ATP-binding cassette sub-family B member 1 (ABCB1)); for example, using a verteporfin photosensitizer and a 690-nm laser, PDP can reduce ABCG2 expression in pancreatic cancer cells and increase irinotecan accumulation in tumours more than tenfold in a MIA PaCa-2 human pancreatic carcinoma orthotopic xenograft mouse model¹⁴. Verteporfin can attach to and induce photodamage in ABCG2 and ABCB1³⁸; for example, low light doses (690 nm, 0.05 J cm⁻²) and a nanomolar concentration of verteporfin is sufficient to inhibit the ATPase activity and induce protein crosslinking of ABCB1 and ABCG2⁷.

Increasing tumour permeability without promoting migration

Increasing tumour permeability to chemotherapeutics by depleting or enzymatically degrading cancer stroma has not improved survival in patients with pancreatic cancer^39,40. PDP can increase pancreatic tumour permeability to drugs, and surviving cancer cells may be less capable of escaping the primary tumour and form distant metastases. Of note, invasive pancreatic cancer cells with the potential for epithelial–mesenchymal transition exhibit higher sensitivity to PDT than cells with a distinct epithelial phenotype in the same 3D in vitro culture⁴¹. For example, porfimer sodium-PDP can suppress migration and inhibit the invasive process of glioblastoma cells in vitro⁴². Similarly, PDP using 5-aminolevulinic acid suppresses migration and invasion of head and neck cancer cells in vitro, along with inhibition of the focal adhesion kinase-Src-kinase-extracellular-signal-regulated kinase (FAK-Src-ERK) signalling pathway⁴³. Furthermore, proteomic analyses of PDT-treated cancer cells revealed that integrins are a major target of photooxidation events^44,45, which can prevent cellular interactions with the extracellular matrix and potentially explain the reduced migration potential of cancer cells following PDT^46–50. Epidermal growth factor receptor (EGFR)-targeted PDP combined with prostaglandin EP4 receptor inhibition attenuates cancer-promoting cell signalling and behaviours linked to metastasis in ovarian cancer cells in vitro⁵¹. PDP can also reduce the tumoural expression of CD44 and CXCR4 stemness markers in a pancreatic cancer mouse model, overcoming the development of resistance after multicycle chemotherapy⁶. Moreover, combining PDP with vitamin D receptor activation in fibroblasts suppresses pro-tumorigenic CXCL12/CXCR7 crosstalk and further improves chemotherapy efficacy in vivo³⁵. Interestingly, low doses of PDT, which are sub-lethal to pancreatic tumour nodules in 3D homoculture, become lethal if the same cells are co-cultured with fibroblasts, suggesting that crosstalk with stromal fibroblasts may enhance tumour response to PDT⁵².

Therefore, not only can PDP increase the permeability of the tumour microenvironment, but it can also prevent cancer cells from interacting with the extracellular matrix, resulting in reduced migration, epithelial–mesenchymal transition and stemness³⁷. Investigating the effect of PDP on the development of metastatic disease is thus crucial to provide information on the safety of PDP.

Photochemical release of endolysosomally sequestered agents

Beyond the initial ablative PDT zone, surviving cancer cells may experience PDP through damage of intracellular organelles, including the membrane of endolysosomes. Most water-soluble agents, proteins and nanoparticles, including therapeutic and immunologic agents, are internalized by cancer cells through endocytosis and are frequently sequestered in the endolysosomes. Photochemical internalization relies on a photodynamic trigger to enable the endolysosomal escape of cancer therapeutic agents⁵³, including chemotherapeutics, such as bleomycin (NCT00993512) and gemcitabine (NCT04099888), and antigens for vaccine delivery (NCT02947854), by confining photosensitizers to the endolysosomal membrane. Typically, lysosome-specific amphiphilic photosensitizers, such as fimaporfin, are used for photochemical internalization; however, hydrophobic photosensitizers may also show lysosomal localization. Therefore, photochemical internalization may be a common mechanism by which cancer cells are sensitized to exogenous adjuvant agents. Photochemical internalization is thus a type of PDP enabling the priming of tumour cells that are located beyond the direct ablative zone to respond to agents sequestered and inactivated in the endolysosomes.

Clinical implementation

PDT is clinically used for the treatment of prostate cancer, brain cancer, bladder cancer and non-cancer applications, such as actinic keratosis and macular degeneration³. In addition, PDP effects on tumour tissue beyond the PDT ablative zones can be utilized to augment the effects of standard treatment modalities. Computed tomography (CT) is routinely used to detect PDT-induced necrosis in cancer tissues, and CT texture analysis can be applied to identify PDP effects⁵⁴. For example, early-phase clinical studies showed that interstitial endoscopic ultrasound-guided PDT using Visudyne⁵⁵ or porfimer sodium⁵⁶ can induce necrosis up to 15.7 ± 5.5 mm in diameter and 18 ± 22% necrosis, respectively, in unresectable pancreatic tumours. Beyond this zone of necrosis, a range of PDP effects may be induced.

To take advantage of the effects of PDP in clinical practice, PDP may be leveraged to eradicate residual disease and reduce recurrence by enhancing anti-tumour immunity. The enhancement of the immune response following PDT or PDP can go beyond the local cytotoxic zone, leading to sustained immune-mediated surveillance and suppression of neoplastic cell growth in other parts of the body^2,57,58. For example, PDP using a triple-receptor-targeted formulation can enhance immune responses by upregulating the membrane exposure and through release of damage-associated molecular patterns, inducing T cell activation in a 3D tumour model¹⁰. The mechanisms of PDP-induced anti-tumour immune enhancement remain to be fully investigated; however, PDP may be a promising addition to immunotherapies.

PDP can also be applied to improve the delivery and selectivity of chemotherapy to tumours. For example, low-dose PDT can increase the enhanced permeability and retention effect in breast tumour⁵⁹ and pancreatic cancer⁶ mouse models, thereby increasing the efficacies of adjuvant chemotherapy. However, the impact of PDP depends on the light dose⁶⁰ and drug–light interval²⁰. Nanoscale liposomes allow the co-delivery of both photosensitizers and adjuvant therapeutics within a single formulation^9,61,62, and they can be further engineered to release their content only in light-exposed areas, thereby reducing off-target drug toxicity^63,64. Therefore, cellular priming, stroma remodelling and immunomodulatory effects of PDP may be utilized to improve cancer treatment.

Molecular engineering of photosensitizers

Pharmacokinetic and pharmacodynamic behaviour

The pharmacokinetic and pharmacodynamic profiles of photosensitizers can be improved by increasing the water solubility of hydrophobic and amphiphilic photosensitizers or by reducing the clearance rate of hydrophilic photosensitizers (Table 1). Most photosensitizers are hydrophobic or amphiphilic to enable plasma membrane penetrance and intracellular photodamage, and thus require solubilization. These photosensitizers are typically delivered using a nanoscale construct (also called a photonanomedicine); alternatively, photosensitizing nanoparticles can be applied⁶⁵. For example, Visudyne is a liposomal formulation of verteporfin⁴. In addition, macromolecules can be conjugated to photosensitizers to impart molecular specificity^17,66.

Table 1 |.

Examples of photosensitizer carriers and enhancers

Chemistry	Degradation mechanisms	Release	Application	Advantages	Disadvantages	Refs.
Photoimmunoconjugates and targeted conjugates
Photoimmuno-conjugates with stochastically conjugated photosensitizers	Intracellular reduction, disassembly and proteolysis	Photosensitizers are released after proteolysis	Approved: photoimmunotherapy of head and neck cancer (Ministry of Health, Labour and Welfare in Japan) Clinical trials: phase III clinical trial of photoimmunotherapy of head and neck cancer (NCT03769506) Pre-clinical: photoimmunotherapy of solid and disseminated tumours (for example, head and neck cancer, ovarian cancer, oral pre-cancers, myosarcoma and mesothelioma)	Specificity improves efficacy and makes high, curative doses safe Long circulation times	Receptor heterogeneity can result in incomplete responses Inhomogeneous tumour distribution owing to binding site barrier effect Photosensitizer-to-antibody ratio limited by reactivity in antigen-binding regions	17,75,195,196
Photoimmuno-conjugates with site-specifically conjugated photosensitizers	Intracellular reduction, disassembly and proteolysis	Photosensitizers are released after proteolysis and/or as a result of the reduction of thioether linkers	Pre-clinical: Photoimmunotherapy of solid and disseminated tumours	If a polyvalent carrier is used, the photosensitizer-to-antibody ratio can exceed the stochastic labelling method No impedance of antigen binding	If no polyvalent carrier is used, the photosensitizer-to-antibody ratio is limited by the number of site-specific modification sites	197,198
Nanobodies	Intracellular reduction, disassembly and proteolysis	Photosensitizers are released after proteolysis	Pre-clinical: Nanobody-targeted photodynamic therapy of resistant tumours, triggering immune responses	Efficient tumour penetration and homogenous distribution Rapid clearance from healthy tissue minimizes drug-to-light interval	Rapid clearance may limit bulk accumulation in solid tumours	199,200
Lipid and polymer nanoformulations
Liposomes	Light-triggered degradation, disassembly in endolysosomes or tumour parenchyma	Passive photosensitizer transfer to lipophilic molecules in serum or to target cell membrane	Approved: PDT for age-related macular degeneration (Visudyne; FDA) Clinical trials: PDT and non-PDT (dark effects) using Visudyne for solid tumours, including glioblastoma (NCT04590664, phase I/II), pancreatic cancer (NCT03033225, phase II) and prostate cancer (NCT03033225, phase I/II) Pre-clinical: PDT-based combinations (for example, chemotherapy, immunotherapy, targeted inhibitors and biologics)	Improves solubility, photoactivity, PK/PD and tumour selectivity of hydrophobic, amphiphilic and hydrophilic photosensitizers Light-triggered drug release Enables multiagent co-delivery and potentiates efficacy	Premature photosensitizer leakage and off-target accumulation Can exhibit instability in serum Short circulation half-lives without polyethylene glycol coating	61,67,201,202
Porphysomes	Light-triggered degradation, disassembly in endolysosomes or tumour parenchyma	Encapsulated agents (if included) are released from the carrier upon light activation	Pre-clinical: photothermal therapy, PDT, light-triggered combination therapies, photoacoustic and multimodal imaging	No photosensitizer leakage owing to efficient self-assembly Simplified system, in which the carrier also serves as the photosensitizer Improved photosensitizer pharmacokinetics with no requirement for a nanoscale vehicle Light-triggered drug release Multiagent co-delivery and potentiated efficacy	PDT efficacy is contingent on partial or full dissociation of the construct. Otherwise photothermal effects dominate	63,203–205
Liposomes with lipid-anchored photosensitizers	Light-triggered degradation, disassembly in endolysosomes or tumour parenchyma	Encapsulated agents are released from the carrier upon light activation	Pre-clinical: PDT, light-triggered combination therapies and lysosome-targeted photodamage	PDT efficacy not contingent on carrier dissociation No photosensitizer leakage owing to stable lipid anchoring into liposomes Improved PK/PD of photosensitizer and encapsulated agents Light-triggered drug release of large and small therapeutic agents	Photodamage is restricted to lysosomes, so treatment is less cytotoxic than mitochondrial photodamage	9,19,67,82,202,206,207
Micelles	Hydrolysis	Encapsulated agents are released from the carrier upon light activation	Pre-clinical: PDT-based combinations and image-guided PDT	Small size favours tumour penetration High loading capacity for hydrophobic photosensitizers compared to liposomes	Short circulation half-lives Low serum stability Loading restricted to hydrophobic photosensitizers	208–212
AIE photosensitizer formulations	(Contingent on formulation)	Formulations release only the encapsulated agent while maintaining the photosensitizer in an aggregated, active state	Pre-clinical: PDT-based combinations	Formulations maintain the aggregated state of AIE photosensitizers, minimize molecular rotations and maintain their phototoxicity Degree of AIE photosensitizer loading provides control over the degree of phototoxicity	Destabilization may lead to de-aggregation and inactivation of AIE photosensitizers	96,97,213,214
Nanoscale metal–organic frameworks
Hafnium – photosensitizer frameworks	Low-pH-induced disassembly	Secondary drug release triggered by photoactivation and/or low pH in tumour microenvironment and endolysosomes	Pre-clinical: PDT-based combinations	Improved photosensitizer PK/PD with no requirement for a nanoscale vehicle	Photochemistry limited by static quenching of tightly packed photosensitizers Low stability Short circulation times	73,215
Iron – photosensitizer frameworks	Low-pH-induced disassembly	Secondary drug release triggered by photoactivation and/or low pH in tumour microenvironment and endolysosomes	Pre-clinical: PDT-based combinations (for example, chemotherapy and immunotherapy)	Hypoxia activity Improved photosensitizer PK/PD with no requirement for a nanoscale vehicle	Photochemistry limited by static quenching of tightly packed photosensitizers Low stability Short circulation times	216,217
Copper – photosensitizer frameworks	Low-pH-induced disassembly	Secondary drug release triggered by photoactivation, redox chemistry and/or low pH in tumour microenvironment and endolysosomes	Pre-clinical: PDT-based combinations (for example, chemotherapy and immunotherapy)	Enhanced PDT efficacy through glutathione depletion Improved photosensitizer PK/PD with no requirement for a nanoscale vehicle	Photochemistry limited by static quenching of tightly packed photosensitizers Low stability Short circulation times	218,219

AIE, aggregation-induced emission; FDA, US Food and Drug Administration; PDT, photodynamic therapy; PK/PD, pharmacokinetic/pharmacodynamic.

Hydrophobic photosensitizers typically partition within lipophilic regions of a lipid-based delivery vehicle; therefore, the pharmacokinetic and pharmacodynamic profiles of the vehicle should be considered, for example, for porphyrin–phospholipid liposomes delivering doxorubicin²⁰. Prior to photoactivation, these doxorubicin-loaded porphyrin–phospholipid liposomes exhibit almost identical pharmacokinetic and pharmacodynamic profiles to the FDA-approved doxorubicin liposomal formulation¹⁹. Importantly, the hydrophobic porphyrin–phospholipid photosensitizer remains stably entrapped during circulation. Pharmacokinetic and pharmacodynamic modelling also revealed that phototriggered release of doxorubicin in the vasculature is responsible for the marked improvement in anti-tumour efficacy²⁰. Polyethylene glycol coating of liposomes can further prevent leakage of unmodified benzoporphyrin derivative (BPD) photosensitizer molecules into serum; however, lipid anchoring of BPD to liposomes is needed to prevent non-specific switching and transfer of the photosensitizer to cancer cell membranes⁶⁷. Interestingly, lipid anchoring of BPD to liposomes re-routes the photosensitizer to lysosomes from the mitochondria and endoplasmic reticulum⁶⁸. Combined with liposomes containing native BPD, multiorganelle-targeted PDT potentiates apoptosis in 3D cancer models⁶⁷.

Self-assembled nanovesicles comprised of porphyrin–phospholipid liposomes⁶³ are potent photothermal and photoacoustic contrast agents in their intact quenched state; however, their disassembly can be triggered by cancer receptor targeting, which causes de-quenching of the porphyrin–phospholipids, thereby enabling PDT and fluorescence imaging⁶⁹. In addition, various metals can be chelated within porphysomes to allow imaging, such as magnetic resonance imaging (MRI)⁷⁰ and positron-emission tomography imaging⁷¹. Similarly, photosensitive metal–organic frameworks can be applied that combine PDT, chemotherapy and immune checkpoint blockade⁷². Unifying the pharmacokinetic and pharmacodynamic profiles of multiple treatment regimens would allow further control of combination therapies and simplify treatment protocols^{9,20,61,62,73}.

The membrane configuration of lipid-anchored photosensitizers (membrane protruding versus membrane inserting) also has a key role in these photosensitizers’ pharmacokinetics, pharmacodynamics, photochemistry and therapeutic efficacy; for example, a membrane-protruding lipid conjugate of the photosensitizer IRDye700DX reduces tumour delivery of liposomes 7.2-fold as compared with a membrane-inserting lipid conjugate of the photosensitizer verteporfin⁷⁴. In addition, the phototherapeutic efficacy of the membrane-protruding configuration is decreased tenfold, despite notably higher ROS production. Therefore, the surface configuration of photosensitizers tethered to liposomes is a crucial and often overlooked parameter in modulating the pharmacokinetics and pharmacodynamics of photosensitizers.

Molecular specificity

Molecular specificity of photosensitizers can be achieved by tethering photosensitizers to macromolecular targeting moieties that exhibit affinities towards tumour-associated receptors and biomolecules (Table 1). Molecular specificity provides tumour-targeted photodamage, thereby minimizing healthy tissue toxicity and allowing the safe use of escalated doses of PDT for a complete curative regimen⁵. Antibodies and engineered antibody formats are often applied as macromolecular photosensitizer conjugates, an approach referred to as antibody-targeted PDT or photoimmunotherapy¹⁷. For example, photoimmunotherapy using cetuximab saratolacan sodium⁵ has been approved for head and neck cancer in Japan, and it is currently in phase III clinical trials in the US (NCT03769506).

Targeting moieties, such as antibodies, antibody fragments and nanobodies, can be tethered either directly to photosensitizers through covalent conjugation^75–78 or to photosensitizer-carrying macromolecular structures, including dendrons⁷⁹, liposomes^80–82 and polymeric nanoparticles⁸³, amongst others^84,85. Tethering antibody–photosensitizer conjugates to liposomes carrying chemotherapeutics^62,86 or to polymeric nanoparticles⁸⁷ further improves tumour-selective photosensitizer uptake and anti-tumour responses in vivo. In addition, a virus-like particle–phthalocyanine photosensitizer conjugate (AU-011) can target cell-surface heparan sulfate proteoglycans for PDT of uveal melanoma (NCT03052127)⁸⁸. Molecular specificity, and subsequent cancer cell internalization, of targeted photosensitizer constructs may also be important for improving anti-tumour efficacy with respect to bulk delivery to the tumour interstitium, even if bulk delivery is selective⁸⁹.

Stimuli responsiveness and image guidance

Stimuli-responsive platforms enable spatiotemporal control over treatment induction, for example, by controlling the release of drugs⁹⁰, such as chemotherapeutics¹⁹, small-molecule inhibitors⁶¹ and immune checkpoint inhibitors⁹¹, by light (Table 1). Such phototriggered treatment induction typically improves anti-tumour responses compared with cocktails of PDT and secondary or tertiary therapeutics^9,61. Photochemical internalization also utilizes photodynamics to spatiotemporally control the intracellular release of chemotherapeutics from endolysosomes by light. For example, this approach can be applied for the endosomal escape of lipid nanoparticles carrying single-guide RNA for CRISPR–Cas9 gene editing⁹², RNAi to inhibit angiogenesis⁹³, and small interfering RNA (siRNA), such as in the clinically approved siRNA-carrying lipid nanoparticle patisiran⁹⁴. Endosomal escape of these therapeutic payloads enables them to reach their intracellular targets, which can in turn complement the therapeutic efficacy of PDT that is applied for endosomal escape.

Instead of using macromolecules to solubilize hydrophobic and amphiphilic photosensitizers, aggregation-induced emission (AIE) photosensitizers can be applied; in this case, phototoxicity is contingent on aggregation, which restricts intramolecular motions⁹⁵. A number of macromolecular structures can preserve the stacking and aggregation of AIE photosensitizers, thereby maximizing phototoxicity. For example, nanoscale lipid and polymer carriers improve the efficacy of AIE photosensitizers⁹⁶. In addition, site-directed aggregation and phototoxicity can be engineered in mitochondria-directed AIE photosensitizers conjugated with pyridinium moieties⁹⁷. Once the conjugates localize and aggregate within the mitochondria, they become phototoxic in vitro and in vivo. Similarly, AIE photosensitizers directed to intracellular lipid droplets, the plasma membrane or the endoplasmic reticulum can exhibit site-specific aggregation and activation of photochemical properties⁹⁸.

Models for optimizing photodynamic therapy

Engineered models are key to preclinically evaluate PDT approaches. For example, 2D cell culture models are valuable for studying treatment effects; however, they lack translational power. By contrast, in vivo models have higher translational value, partly recapitulating the complexity of human cancer tissues within a living organism. However, to minimize the use of animals in research and better recapitulate human conditions, compared with 2D culture models⁹⁹, 3D human cell culture models can be engineered that mimic the complexity of cancer tissues and offer the investigational accessibility and throughput of in vitro models^100–102 (Fig. 3). Three-dimensional culture models can be non-adherent spheroids, composed of cancer cells and supplemented with stromal cells; matrix-adherent or matrix-embedded microtumours, composed of established cancer and stroma cell lines; or matrix-adherent and matrix-embedded cancer organoids, which are obtained from dissociated surgically resected tumour tissues.

Imaging-based assays

The heterogeneity of models and the absence of functional assays are major challenges in the utilization of 3D cancer models for evaluating treatment effects^103,104. Predominantly measured by probing cellular metabolic activity, treatment effects can also materialize by changes in tumour size, tumour growth rates, viability and redox state, among other factors. Imaging-based assays, which are capable of multiparametric analysis of hundreds to thousands of cancer organoids, have been developed to embrace the heterogeneity of these models as a strength, rather than as a weakness. For example, quantitatively analysing images obtained by darkfield microscopy enables the identification of distinct alterations in the size distribution of ovarian cancer organoids following carboplatin chemotherapy and PDT¹⁰⁵. This image analysis procedure can also be combined with a LIVE/DEAD cell staining procedure to extract cancer organoid volumes and viability indices for each individual organoid imaged by low-magnification microscopy¹⁰⁰. This protocol can be adapted to asymmetrical cancer organoid cultures, providing the readouts in area rather than in volume¹⁰¹.

Quantitative imaging approaches can thus be implemented to identify specific treatment responses in 3D cell cultures. For example, high-throughput image analysis revealed that chemotherapeutics reduce the viability of architecturally complex heterotypic organoids, that PDT decreases the relative live areas or volumes of microtumours by causing peripheral necrosis, and that radiotherapy inhibits cancer organoid growth without affecting viability¹⁰¹. Image analysis can also inform on the uptake and localization of photosensitizers^106,107 as well as on perturbations in redox state using NAD(P)H and FAD autofluorescence^108–111. Moreover, imaging approaches can be adapted for any biomarker expression analysis by immunofluorescence.

Spheroid and microtumour models

Spheroid cancer models are cell aggregates cultured on U-bottom multiplate wells with a non-adherent coating. Spheroids can be composed of cancer cell lines and may be supplemented with endothelial cells, fibroblasts or immune cells to more closely mimic the cancer microenvironment. As spheroid models are inexpensive and homogeneous, they constitute the most widely used 3D models in PDT research. For example, spheroids composed of both cancer and stroma cells have been used to demonstrate that antibody-functionalized liposomes can facilitate selective PDT of cancer cells⁸², that PDT–radiotherapy combination therapy results in smaller and more necrotic pancreatic cancer spheroids than each individual therapy³⁶, and that such spheroids are useful to evaluate the uptake and penetration depth of various nanocarriers for photosensitizers¹¹². However, not all cell types can grow in non-adherent environments. In addition, the formation of non-spherical concave structures rather than true spheroids may not represent cancer tissues as accurately as intended.

Alternatively, cancer cells can be grown in extracellular matrix scaffolds to form spherical microtumours at the gel–medium interface, thereby simulating the mechanobiological environment of a tumour. For example, ovarian microtumours have been used to demonstrate synergistic, yet sequence-dependent, enhancement of carboplatin chemotherapy by PDT¹¹³. Similarly, short-term and long-term effects of oxaliplatin–PDT combinations have been studied in pancreatic microtumours¹¹⁴. More complexity can be added by introducing multiple cell types; for example, in pancreatic microtumours grown on extracellular matrix hydrogels, the addition of cancer-associated fibroblasts substantially alters the morphology and redox state of microtumours as well as their susceptibility to oxaliplatin and PDT¹⁰⁸. However, a limitation of both spheroid and hydrogel-based models is the lack of fluid flow, which can be introduced through microfluidics. Under fluid flow, human ovarian cancer microtumours undergo epithelial-to-mesenchymal transition¹¹⁵, which induces carboplatin resistance but improves the cellular delivery of photosensitizers formulations and increases photoimmunotherapy efficacy^116,117.

Patient-derived organoids

Cells can be extracted from patient biopsies and grown into cancer organoids to identify the most effective treatment regimens for a specific patient¹¹⁸. This requires the dissociation of tissues into single-cell suspensions, which can then be grown on or in extracellular matrix scaffolds¹¹⁸. For example, patient-derived head and neck cancer organoids were utilized to examine the selectivity of immunotargeted PDT by comparing phototoxicity in the cancer organoids with phototoxicity in healthy organoids developed from nearby normal tissue tissue from the same patients¹¹⁹. Similarly, patient-derived cholangiocarcinoma organoids demonstrated higher protoporphyrin IX biosynthesis following 5-aminolevulinic acid treatment than non-cancerous intrahepatic and extrahepatic bile duct organoids derived from the same patients¹²⁰. Therefore, establishing selectivity of photosensitization approaches in patient-derived organoids may inform the suitability of patients to undergo targeted PDT or fluorescence-guided surgery.

Taken together, evidence from studies indicates that 3D cancer models recapitulate PDT dose responses better than 2D cell cultures and that the responses of the former are similar to the in vivo responses. Importantly, rather than information-poor biochemical assays, imaging-based assays can be applied to 3D culture models to track tumour growth and treatment-induced shrinkage, which is similar to how ultrasound, CT and positron-emission tomography are used to evaluate treatment responses in patients. Further utilization of 3D cancer models will greatly benefit from rapid optical imaging systems, optical clearing methods¹²¹ combined with adaptive optics¹²² or light-sheet microscopy¹²³, and artificial intelligence-based image analysis¹²⁴.

Photonic devices and image-guided strategies

Light sources

The light threshold to induce PDT phototoxicity in tissues and cells is in the range of 10⁹–10¹⁰ photons absorbed per cell^125–127. Although the delivered radiant energy can be as low as 1–10 J cm⁻² in vitro, depending on the photosensitizer and its concentration, in vivo, macroscopic PDT typically requires more than 50 J cm⁻² for bulk tissue damage^128,129. PDP is elicited at fluences that are lower than these values^6,54,82. Large class IV laser systems were originally used as light sources for experimental PDT, whereas most high-power needs are now supplied by diode laser modules or light-emitting diode (LED) lamps¹³⁰. In particular, if focal light delivery to isolated locations is required, multichannel diode lasers can be delivered by fibres, for example, for intraocular, interstitial, intraoperative intraluminal and diffuse illumination^131–133. These fibres are terminated by specialized diffusers or applicators that are specific to the tissue or organ^134–136. Interstitial and superficial light delivery enable bulk tumour irradiation, in particular for inner organ cancer therapy. Clinical light delivery systems have also been developed for different cancers (such as, peritoneal metastasis and brain cancer) and for online treatment monitoring and analytics^62,137–139. In ophthalmology, light delivery systems are also cloud-connected and tablet-controlled to improve usability and clinical adoption¹⁴⁰. External dermatological PDT is typically performed with LED-based lamps or LED-based blankets that cover the tissue^141–144. Deep-tissue PDT can be achieved with LED implants^145–147 or by chemiluminescence, bioluminescence, X-ray, ultrasound or Cerenkov radiation.

Image guidance

Image guidance in PDT has mainly focused on light delivery verification and localization to confirm that a sufficient amount of light has been delivered. However, geometric guidance can also be achieved by CT¹⁴⁸, ultrasound¹⁴⁹, MRI^150,151 and surface and cavity capture^152–154. Moreover, photosensitizers can be indirectly guided through quantifying fluorescence pre-treatment⁶² or photobleaching to assess absorbance during treatment^62,155. In addition, ¹O₂ guidance can be applied to quantify photochemical delivery during treatment^156,157. Although these image guidance approaches require additional instrumentation or procedures, they can ensure that there is no under-treatment or over-treatment. Clinical light delivery systems typically integrate treatment monitoring functions based on tissue absorption or fluorescence to inform dosimetry and report on efficacy^62,137–139. In particular, over-treatment can have devastating effects in specific anatomical locations, such as oesophageal structures^158,159 and the bladder. Under-treatment may often occur in PDT, requiring repeated and customized treatments for full efficacy⁶². Of note, under-treatment essentially is the core concept of PDP, which subtly causes alterations in permeability and tissue response.

Surrogate and minimal approaches to dosimetry

Experimental PDT typically takes an extreme dosimetry approach; here, a number of parameters (such as, photosensitizer delivery, photobleaching, tissue optical properties and ¹O₂ levels) are measured to understand the response and determine dominant factors in treatment control. However, clinical PDT aims for delivery with low dosimetry to avoid over-treatment. Clinical light delivery systems are typically minimally invasive single-probe systems that provide treatment monitoring for usability, repeatability and efficacy. Surrogate endpoints, as typically used in pre-CT and post-CT scans with contrast or pre-skin and post-skin measurements of photosensitizers, can be useful to quantify the delivery and efficacy of PDT and PDP⁵⁴. In addition, laser, optical imaging and photonanomedicines may improve treatment planning, dosimetry and treatment response monitoring.

Alternative activation sources

Ultrasound

Ultrasound as a diagnostic tool benefits from deep tissue penetration (for example, low frequency ≈1 MHz pressure pulses can reach up to a depth of 10 cm) to probe deep-seated tumours. In particular, sonodynamic therapy relies on low-frequency ultrasound to activate an acoustically sensitive agent, that is, a sonosensitizer or photosensitizer. However, the mechanisms underlying sonodynamic therapy remain to be fully unravelled. In sonodynamic therapy, ultrasound itself can induce a therapeutic effect. Although typically performed with minimally hyperthermic ultrasound doses, sonomechanical tissue damage can occur¹⁶⁰. However, sonodynamic therapy primarily relies on mechanisms that generate ROS,- in particular, ¹O₂. Two mechanisms have been proposed to explain the origin of ROS production in sonodynamic therapy: sonoluminescence and pyrolysis. Sonoluminescence is an emission of photons generated during cavitation, peaking around 400–450 nm, a range that overlaps well with the absorption peak of commonly used photosensitizers. Sonoluminescence may excite photosensitizers to induce PDT in tumour regions that may be typically inaccessible to direct light activation. However, although an overlap between the sonoluminescence spectrum and the absorption of porphyrin sonosensitizers has been reported, photoinactive compounds can also produce ¹O₂ upon ultrasound excitation¹⁶¹. By contrast, pyrolysis may be triggered by inertial cavitation, which causes high local temperatures and pressures. Although pyrolysis is independent of the presence of a sonosensitizer, the proximity of a drug may amplify the efficacy of pyrolysis observed with ultrasound alone^162,163. Therefore, sonodynamic therapy may be able to induce a therapeutic effect even in tumours localized at a depth of 10 cm within tissue; however, the underlying mechanisms remain to be fully investigated.

X-rays

X-rays are commonly used in radiology and radiotherapy for diagnosis and treatment, respectively. X-rays readily penetrate soft tissue and may thus provide a promising tool to activate PDT regardless of tumour depth^164–166. In particular, radiodynamic therapy relies on the direct X-ray excitation of a photosensitizer, which in this context is referred to as radiosensitizer. Alternatively, indirect excitation of photosensitizers can be achieved using nanotransducers that locally down-convert high-energy X-rays into light to excite nearby photosensitizers in an approach referred to as XPDT.

Radiodynamic therapy.

Several photosensitizers can be excited by X-rays to produce ROS, including ¹O₂¹⁶⁷. X-rays may excite photosensitizers through direct electronic excitation or through indirect excitation by what is speculated to be Cerenkov radiation. Cerenkov radiation causes optical emission if charged particles, such as electrons or positrons, travel faster than the phase velocity of light in a given dielectric medium. The spectrum of Cerenkov radiation is mainly in the ultraviolet–visible range, which overlaps with the strong absorption (Soret) bands of commonly used photosensitizers. During X-ray radiotherapy, secondary electrons are created by the interaction of ionizing photons with tissue. These electrons can generate Cerenkov emission if their energy is higher than approximately 250 keV^168,169. Cerenkov radiation can be measured in tissues that are exposed to a 6–18 MeV photon or electron irradiation, causing the excitation of protoporphyrin IX embedded in biological phantom structures¹⁶⁸. For example, 5-aminolevulinic acid, the precursor of endogenous protoporphyrin IX, shows radiosensitizing properties upon 15 MV photon irradiation in a subcutaneous model of lung cancer (KP1 cells implanted in C57BL/6 mice)¹⁷⁰. However, although protoporphyrin IX may produce ROS, including ¹O₂, upon X-ray irradiation below 250 keV^{167,171–173}, such energies are insufficient to induce Cerenkov radiation, emphasizing the fact that Cerenkov radiation cannot be the only mechanism at play here. 5-Aminolevulinic acid has also been shown to possess radiosensitizing properties in colorectal cancer models in vitro and in vivo, using an X-ray generator operating at 150 kVp, which is also insufficient to trigger Cerenkov emission¹⁷⁴. However, other photosensitizers, such as a haematoporphyrin derivative, have been reported to be directly excited by X-rays; it nevertheless remains unclear what this depends on, except for the concentration of the photosensitizer¹⁶⁷.

Therefore, the role of Cerenkov radiation in the radiodynamic process and the radiosensitizing properties of specific photosensitizers remain to be fully understood. Because of the origin of the Cerenkov radiation and the associated energy threshold, the role of X-ray energy in radiotherapy should be reported and investigated. Nevertheless, radiodynamic therapy has clinical potential, as illustrated by a clinical trial for first-time relapse of malignant glioma (NCT05590689). This phase I/II clinical trial was designed to investigate the tolerance of repeated administration of 5-aminolevulinic acid, usually applied as a single dose, in combination with radiotherapy. In addition, the efficacy of additional 5-aminolevulinic acid administration is being investigated.

Nanoscintillators.

XPDT relies on nanotransducers called nanoscintillators. These nanoparticles locally down-convert X-rays into radioluminescence, which can excite nearby photosensitizers^164,175. Nanoscintillators are usually made of insulating materials doped with rare-earth elements. Various nanoscintillators have been preclinically investigated^165,166 at doses of radiation (≈5 Gy) that are lower than common radiotherapy doses (up to 80 Gy), applying computer simulations to estimate the maximal radioluminescence-induced PDT effect^176,177. Interestingly, the radioluminescence emitted by nanoscintillators upon radiotherapy is orders of magnitude lower than the light dose required to induce such an effect. Therefore, the efficacy of XPDT cannot be solely explained by radioluminescence, and other parameters are likely to be at play. In particular, the existence of non-resonant Förster energy transfer between a nanoscintillator and photosensitizer may enhance treatment efficacy¹⁷⁸. In addition, radiotherapy and PDT have different cellular targets, activating different pathways, and may thus synergize to improve outcomes³⁶. Another (synergistic) contribution comes from the heavy elements present in nanoscintillators, which can locally enhance the radiation dose owing to the energy deposited by Auger and photoelectrons emitted during photoelectric interactions between X-rays and nanoscintillators¹⁷⁹. Moreover, Cerenkov emission may be considered if radiotherapy is delivered by high-energy X-rays or particle beams.

The design of XPDT protocols should consider the energy of X-rays. Unlike Cerenkov radiation, which is maximized at high-energy X-rays, the absorption cross-section of nanoscintillators is maximized for X-ray energies below 250 keV, for which the photoelectric effect dominates^176,177. Preclinical studies are typically performed using small animal irradiators, which operate in the keV energy range; by contrast, clinical irradiators operate in the MeV range. If the efficacy drops at high energy, strategies such as brachytherapy may have to be considered in the clinic. Furthermore, the bioavailability and toxicity of nanotransducers containing heavy elements need to be considered. Encouragingly, Gd-based and Hf-based nanoparticles were shown to be safe in in phase I clinical trials^180,181.

Self-illuminating compounds

Bioluminescence originates from the oxidation of luciferins (for example, D-luciferin and coelenterazine)¹⁸² catalysed by luciferase enzymes. Bioluminescence may be able to locally excite photosensitizers to induce PDT, regardless of tumour location. However, the intensity of bioluminescence is orders of magnitude lower than the intensity of the light applied to activate PDT¹⁸³. Therefore, bioluminescence-induced PDT likely originates from a combination of radiative and non-radiative energy transfer. A bioluminescence energy of around 100 μJ may induce a photochemical effect superior to a laser delivering an energy of ~100 mJ¹⁸³. Bioluminescence-induced PDT has been shown in vitro in human glioblastoma cells (U87-MG)¹⁸⁴, mouse Lewis lung cancer (LCC) cells¹⁸⁵, mouse melanoma cells (B16F10)¹⁸³ and mouse colon cancer cells (CT26)¹⁸³, as well as in vivo for LCC, B16F10 and CT26 tumours¹⁸³. However, cancer cell-specific expression of luciferase will need to be demonstrated for clinical translation, as shown for intratumourally administered D-luciferase conjugated to QD-655 quantum dots¹⁸³. Intratumoural photosensitizer delivery by minimally invasive techniques may increase PDT efficacy¹⁸⁶, in particular if combined with implantable biomaterials and self-illuminating light sources.

Radionuclides can also excite photosensitizers regardless of the tumour depth and without the need for an external stimulus^{169,185,187,188}. Although not yet fully elucidated^189,190, the synergy of radionuclides injected together with a photosensitizer may initiate from the activation of the photosensitizer by Cerenkov radiation, created by the electron or positron emitted during radioactive decays. Although this hypothesis is challenged by the dim intensity of Cerenkov radiation, the activity of radionuclides may correlate with the activation of meso-tetrakis(4-carboxyphenyl)porphyrin¹⁹¹, mesotetraphenylporphin¹⁹² and TiO₂ nanoparticles^185,187, acting as photosensitizers. Here, radioactive elements were applied that are commonly used as radiotracers in nuclear imaging.

In addition to investigating the mechanisms of action leading to photosensitizer activation (Cerenkov radiation or other mechanisms), the role of the specific emitters (electron, positron, gamma), the energies of emitted particles, and the half-life of the radioactive elements remain to be fully investigated. For example, ⁶⁸Ga was found to be more efficient than ¹⁸F at producing ROS after activating TiO₂ nanoparticles both in vitro and in vivo in mouse breast cancer cells¹⁸⁷. In addition, it has been theoretically shown that the lifetime of the emitter is a critical parameter to consider when selecting the radionuclide: the longer the lifetime, the more efficient the Cerenkov emitter is for PDT¹⁶⁹. Regardless of the origin of the effect, radionuclides have been repeatedly shown to activate photosensitizers without an external stimulus.

Outlook

PDT and, in particular, photodynamics-based precision therapy will benefit from the development of new sensitizing agents, a thorough mechanistic understanding of photodynamic processes, and the integration and emergence of new nanotechnology and imaging tools, as well as microtumour models. Importantly, the low and nonoverlapping toxicity of PDT (when combined with chemotherapeutics) and the absence of adverse effects, such as fibrosis, make this approach promising for the treatment of diseases. However, many nanotechnology and image-guided strategies for PDT and PDP have not yet been implemented in the clinic, and may require further characterization and optimization in microtumours and in animal models.

In addition, controlled and robust clinical trials will be required to move more PDT approaches to the clinic. Typically, PDT is tested as a monotherapy if other approaches (such as chemotherapy and radiation) have failed or are not suitable, and it often results in tumour necrosis, reduction in tumour volume, and increased survival^55,153,193. However, these promising results are often only observed in a small sample size (fewer than 100 participants). Larger controlled clinical trials of PDT should thus be guided and informed by an organized framework, equivalent to the Radiation Therapy Oncology Group, to clarify its clinical implementation, establish guidelines for dosimetry and treatment design, and allow for appropriate comparisons with other therapies.

Although the regulatory process of photodynamic approaches is complex, as it involves the regulation of both a photosensitizer drug and a laser device, several PDT approaches have already been approved for the clinic³. For example, photodynamic treatment became the first line of therapy for age-related macular degeneration, a leading cause of blindness in the elderly. Importantly, industry involvement is key to commercialization (Box 2).

PDT is particularly suitable for the treatment of deep-seated tumours, and it could also be applied to extend the treatment potential of other therapies. For example, immunotherapy is not applicable to all cancer types or all patients, owing to dose-limiting immune-related toxicities and the inability of immune cells to reach tumours, particularly stroma-rich tumours, such as in pancreatic cancer. PDT can transiently permeabilize tumours (stroma, cells and vasculature) to increase T cell trafficking to tumours and blood. In addition, optical systems and devices as well as artificial intelligence approaches in imaging may be utilized to identify immune cells and microenvironmental changes. The ability of PDT to modulate tumour and vascular permeability could further control patient-specific dosing and timing of checkpoint inhibitors and cell therapies, improving their therapeutic index. Effective dosimetry by incorporating radiomics features of tumours and imaging would enable precise patient-specific administration of PDT therapies. Therefore, PDT should not only be considered a local therapy but as part of an arsenal of treatment combinations.

Moreover, photodynamically active nanomaterials should be explored that can switch photodynamic processes on and off to offer quantitative control of photodynamic effects. In addition to delivering multiple therapeutic agents, whose delivery and efficacy are augmented by photodynamic action, control of delivery by nanomaterials will allow more precise dosimetry and spatial control to limit undesired tissue damage. Such an approach would remove uncertainties related to the different pharmacokinetics of drugs when used in combination. Combining nanomaterials with on-line hyperspectral imaging opens up possibilities of precision medicine, for example, by applying different levels of light to different parts of the body or tissue. Finally, PDT may also be applied beyond cancer; for example, for age-related macular degeneration (Visudyne)⁴, and infectious diseases, such as SARS-CoV-2¹⁹⁴.

Key points.

• Photodynamic therapy (PDT), including photodynamic priming, allows the localized destruction of diseased cells and tissues by light, benefitting from minimal side effects, minimally invasive administration and the potential for combination therapies.

• Nanomaterials can be designed to enable controlled delivery of combination therapies, improve the pharmacokinetics of photosensitizers and enable optical imaging-guided PDT.

• PDT can be applied for tumour ablation, modulation of the tumour environment and boosting of the immune responses against local and metastatic disease.

• Robust clinical trials, industry involvement and collaboration between different research fields (such as, physics, mathematics, chemistry and engineering) are essential for the clinical translation of PDT-based therapies.