S.N.B. A significant body of work in the past decade has helped shed light on the journey a nanoparticle takes to reach its destination; from the formation of protein corona in the bloodstream to overcoming subsequent physiological barriers at the organ, tissue, cellular and subcellular levels20. This improved insight has informed a newer generation of targeting technologies to enhance binding to specific tissues and cell populations; for example, by leveraging serum absorption of apolipoproteins to target the liver21; endogenous trafficking processes such as albumin hitch-hiking to increase accumulation in lymph nodes22; or other biologically inspired processes such as decorating nanoparticles with peptides that leverage active transport pathways akin to viruses for deeper penetration into tumour tissues23. To complement these bio-inspired approaches, unbiased models of nanoparticle drug delivery systems have also been established. For example, high-throughput combinatorial screening of large nanoparticle libraries via barcoding systems enables rapid and functional screening in vivo24, and universal formulation strategies have been devised to systematically alter nanoparticle delivery behaviour25. Thus, both rational design based on evolutionary winners and unbiased approaches leveraging high-throughput science and data analytics offer paths forward to improve nanoparticle delivery to sites of interest.

X.C. To improve nanoparticle delivery, we need a better understanding of the journey that nanoparticles take in the body after different routes of administration26. In the past, we have overemphasized the concept of leaky tumour vasculature and the so-called enhanced permeability and retention (EPR) effect. This often works quite nicely in preclinical xenograft models but not necessarily in the clinic. The nano-formula will circulate in the blood, extravasate to the interstitial space, penetrate the tumour microenvironment, bind to the target on the tumour cells, internalize into the intracellular compartment, release the cargo and, finally, act to kill or provide an insult to the cells of interest27. Every step of this process can be considered in the design of nano-formulas to improve delivery, reduce unwanted reticuloendothelial system uptake, and lead to safer and more efficacious treatment outcomes.

For example, patients can be stratified (for instance, by contrast-enhanced MRI using nanoparticles) according to their tumours’ propensity to passively accumulate nanomedicines. Furthermore, nanomedicines can be made into smart formulas that can transform in terms of size, charge, shape and so on by external (for example, temperature, magnetism, light and X-ray) or internal (for example, pH and enzyme) stimuli for more efficient tumour delivery. Nanomedicines can also be designed to target not only tumour cells but also various stromal components in the tumour microenvironment to modulate the immune system to recognize cancer cells, or to target and potentiate tumour-infiltrating myeloid cells28. For some applications, nanomedicines such as cancer nano-vaccines can be designed to be administered locally (for instance, via topical or intramuscular routes), which can drastically reduce off-target effects29. We can also further improve encapsulation and targeting strategies to maximize the amount of drug or drug combinations delivered to the tumour region.

M.A.D. The key to improving drug delivery is understanding the relationship between nanoparticle physico-chemical properties and the desired type of biological responses. Although a lot has already been achieved in this direction, many areas are still unclear owing to the structural complexity of nanoparticles and the limitations of current methods for nanoparticle physico-chemical characterization. The significant barrier in this area is that not all physico-chemical characterization methods and characterization frameworks apply to all nanoparticle types. Moreover, using a nanocarrier successfully to deliver one drug does not equate to success in the delivery of another drug. One dream for a formulation scientist is a computer algorithm that would take an active pharmaceutical ingredient (API) structure and information about the delivery target site, and select a nanoparticle carrier with optimal properties for achieving such delivery. Many bioinformatics attempts are currently underway but, to my knowledge, none of them so far is broadly applicable to all types of drugs and all types of nanocarriers.

On another important note, according to one recent study, the female menstrual cycle dramatically changes the delivery of nanoparticle-formulated drugs to tumours in various locations in the body30. Therefore, another approach for delivery improvement is advancing the fundamental understanding of how biological processes such as circadian rhythms and hormonal fluctuations influence nanoparticle pharmacokinetics, efficacy and safety.

T.L. We have to realize that several clinically used drug delivery platforms, particularly antibodies (that is, nature’s own targeting vectors) and PEGylated liposomes, already present with very good tumour-targeting capabilities. If tumour targeting with such formulations does not work out well in certain patients, then this is typically not due to the delivery system but, rather, to tumour-specific pathophysiological constraints, such as poor tumour perfusion and high stromal density.

There are multiple means to improve nanoparticle delivery to tumours. These include pharmacological and physical priming approaches, using systemically administered agents to induce vascular normalization or stromal remodelling, as well as locally applied co-treatments, such as hyperthermia, radiotherapy and ultrasound31. A downside of the latter is that they are locally constrained, and thus not very helpful in the case of metastasis. A downside of priming treatments in general is that it is nearly impossible to non-invasively, quantitatively and repetitively monitor tumour-directed drug delivery in patients. Consequently, we typically do not know how good or bad tumour-directed drug delivery actually is, and we will have to come up with ways to rationalize which pharmacological or physical co-treatment may work at which dose for which tumour in which patient.

An important misconception with regard to improving nanoparticle delivery to tumours relates to the added value of ‘active’ targeting. It seems that people fail to realize that active ligand-mediated targeting depends heavily on passive targeting principles, particularly on a long nanoparticle circulation time, efficient tumour perfusion, high tumour vascular permeability and proper tumour tissue penetration. None of these principles profit from the presence of a targeting ligand, and some of them may even suffer from it; for example, a shorter circulation time resulting from more rapid phagocytic capture, or less efficient tumour penetration because of the presence of the binding site barrier. What may change upon active targeting is that cancer cell uptake (versus otherwise predominant uptake in tumour-associated macrophages) and tumour retention increase a little, but certainly not massively. Active targeting should therefore not be used to enhance overall tumour-directed drug delivery but, rather, as a means to ensure uptake by specific cells in tumours. When considering the use of nanoparticles for anti-oncogene siRNA therapy, then active targeting may be a must, as siRNA really needs to be delivered into cancer cells. This is different for small-molecule drug delivery, for example liposomal doxorubicin, which upon predominant uptake in and processing by tumour-associated macrophages will become available to neighbouring cancer cells.

In the era of cancer (nano-)immunotherapy, we have furthermore realized that targeted delivery to tumours and tumour cells is not at all a must. Using nanoparticles to deliver immunomodulatory agents to antigen-presenting cells in the spleen or to myeloid progenitor cells in the bone marrow may be at least equally promising32,33. This is arguably easier to achieve, as it targets tissues with a natural propensity for nanoparticle accumulation, and therefore suffers much less from issues related to heterogeneity in nanoparticle tumour targeting.