Analysis of polymeric nanoparticle properties for siRNA/DNA delivery in a tumor xenograft tissue slice air–liquid interface model

Classical two‐dimensional (2D) cell culture as a drug or nanoparticle test system only poorly recapitulates in vivo conditions. Animal studies are costly, ethically controversial, and preclude large‐scale testing.

indispensable part of drug development and are necessary for the approval of new drugs. Preclinical studies usually consist of tests in vitro (e.g., classical cell culture) and in vivo (animal studies). [1] Normally, cell culture experiments represent the earliest and most important pharmacological tests in drug development, allowing the assessment of therapeutic drug effects and cytotoxicity in viable cells under physiological conditions of pH, osmotic pressure, and temperature. However, common two-dimensional (2D) cell cultures are not able to accurately represent three-dimensional (3D) tissue conditions, for example, in the case of drugs intended to treat solid tumors. This can be explained by the lack of 3D interactions between tumor cells and extracellular matrix (ECM) proteins and neighboring cells (e.g., stromal cells), as well as the absence of several important physiological processes, such as drug transport through cell layers. [2] In particular, the ECM can form a tight network, which can seriously hinder the penetration of nanoparticles into tumors or affect the sensitivity of tumor cells toward nanoparticles. [3] Furthermore, unlimited access to nutrients and lack of cell-cell interactions have crucial effects on cell proliferation, morphology, differentiation, protein expression, and responsiveness to drugs and their metabolism. [4] The drawbacks of 2D models require alternative models, such as 3D culture models, that are better able to mimic a natural tumor environment. [5] In vitro 3D cell culture models can address this issue and fill the gap between 2D cell culture and in vivo tumor models. [6] They range from cancer cell spheroids (including co-culture models involving multiple cell lines) to 3D matrix scaffolds (e.g., hydrogels), organoids, and biochips. [7,8] The most popular 3D system represents spheroids consisting of tumor cells alone or in combination with other cell types.
Spheroids are formed by the natural properties of many cells to aggregate and build 3D cell-cell interactions, which allows them to mimic some properties of tissue physiology. [9] However, these systems do not reflect an intact tissue architecture. Among the models mimicking the in vivo situation are ex vivo tissue slice models with the potential to provide an alternative to current in vivo models. Slices prepared from freshly excised tumor tissues from patients and cultured under air-liquid interface (ALI) conditions were reported to retain their viability for at least 1-2 weeks. Comparably small amounts of tumor tissue are sufficient for these experiments. [10,11] However, the availability of sufficient amounts of (tumor) tissue represents a limitation, which can be avoided by switching to tumor xenografts from animals, for example, mice. [12,13] Comparably thin tissue slices allow diffusion of oxygen, nutrients, and drugs. The fact that cell viability and intact tissue architecture are maintained for up to 14 days is sufficient for the monitoring and analysis of experimental interference, for example, by applying test drugs. [13,14] Thus, the use of such tissue slice models is also a promising approach for the reduction of in vivo studies, according to the "three R principle" (replacement, reduction, refinement of animal experiments), [15] which saves time and financial resources and allows the direct interrogation of drug effects on the target tissue. [16] More recently, we demonstrated that nanoparticles can penetrate tissue slices. When delivering small interfering RNAs (siRNAs) for the induction of RNA interference (RNAi)-mediated gene knockdown, relatively profound target gene downregulation was observed, indicating nanoparticle-mediated delivery into the deeper cell layers of the tissue. In line with this, fluorophore-labeled siRNAs showed penetration into the tissue, which seemed to be dependent on the nanoparticle properties, especially surface charge. [13] Comparable to low-molecular-weight drugs, this would thus allow the monitoring of biological nanoparticle properties and efficacies at their intended site of action under in vivo-like conditions. We previously established a large set of polymeric nanoparticles based on polyethylenimine (PEI) for the delivery of siRNA or plasmid DNA. In particular, this includes various linear or branched PEIs chemically modified by tyrosine grafting, which leads to the improvement of physicochemical and biological nanoparticle properties. [12,[17][18][19][20][21] Likewise, the combination of PEIbased nanoparticles with liposomes, which leads to the formation of lipopolyplexes (LPPs), results in altered and improved nanoparticle properties. [22,23] In this study, we comprehensively compared these nanoparticles in tissue slice ALI cultures, in direct comparison with their in vitro activities. We introduce and characterize tissue slice ALI cultures as a powerful system for analyzing nanoparticle properties and efficacies in a biologically relevant ex vivo model.

EXPERIMENTAL SECTION
Further information regarding materials and methods is given in the Supplementary material.  Supplemental Table S1.

Tumor tissue slice preparation, cultivation, and treatment
To generate subcutaneous tumor xenografts, 5 × The next day, the slices were transferred to a 96-well-plate containing 100 μl of pre-warmed medium.
Polymeric PEI-based nanoparticles were prepared by mixing of 1.5 μg siRNA or pDNA with 3.75 μg (11.25 μg in case of F25) of the respective polymer in trehalose buffer (10% w/v trehalose, 20 mM HEPES, pH 7.4). LPP were prepared as described previously by combining PEI-based polyplexes with liposomes. [22] The tissue slices were incubated for 6 h in the nanoparticle-containing solution prior to placing them back on the membrane culture inserts and cultivation for another 72 h, with the culture medium being replaced every day. After approximately 72 h, the slices were harvested for further analysis.

2.3
Tissue slice analyses by flow cytometry, RT-qPCR, and confocal microscopy  In the next step, tumor slices were prepared using a vibratome and pieces of ∼3 mm in diameter are stamped out using a biopsy punch. Then, tissue slices were placed onto membrane culture inserts, in a cell culture well filled with cultivation medium and kept in an incubator overnight. For treatment, slices were transferred to a 96-well-plate containing pre-warmed medium followed by the addition of nanoparticles. After transfection, slices were placed back on the membrane culture inserts, prior to cultivation and further experimental evaluation.

Optimized procedures for tissue slice preparation, cultivation, and treatment
Tumor tissue slices were prepared from the tumor xenograft tissues as shown in Figure 1. Compared with a previously published study, [12] the tumor tissue was sliced using a vibratome instead of a tissue chopper, yielding very precise and uniform tissue slices. For better comparison of individual tissue slice samples of equal size, pieces approximately 3 mm in diameter were stamped out of the sliced tissue using a biopsy punch. This resulted in uniformly sized tissue slice samples, which were sufficiently small for incubation and transfection in a 96-well format.
Compared to the previously employed pipetting of the test solution onto the tissue slices, the present procedure allows prolonged exposure times and well-defined exposure conditions of the tissue with nanoparticles. Very short interaction times would not mimic the in vivo situation. Indeed, direct comparison of tissue slices submersely transfected in solution versus slices exposed to the transfection medium on the insert revealed marked differences in biological efficacies, as determined by GAPDH knockdown (Supplemental Figure S1). Biopsy punching of tissue slices also allowed for the selection of appropriate tumor areas, while excluding visibly necrotic or damaged tissue areas, and obtained a larger set of samples (typically up to 50) with identical dimensions. The tissue slices were kept at the ALI, which was shown previously to preserve tissue integrity and viability for at least two weeks.

Nanoparticle penetration and gene knockdown in tissue slices
To investigate cellular uptake and tissue penetration, tissue slices from PC3 prostate carcinoma xenografts were treated with complexes based on linear PEI 10 kDa tyrosine-modified (LP10Y). Among the different PEI-based siRNA complexes, LP10Y/siRNA nanoparticles had been found previously to be particularly efficient in vitro. In combination with previous in vivo studies on LP10Y/siRNA complexes for therapeutic knockdown in preclinical mouse models, [19] LP10Y was an ideal candidate for experiments in our 3D tissue slice system.  Figure 2B) and the fluorescence intensity were dependent on the amount of LP10Y/siDY647 complex, with a two-fold increase in siRNA amounts leading to doubling of fluorescence ( Figure 2C). Complex penetration was further studied using confocal microscopy of cross-sections of tissue slices ( Figure 2D). The images revealed fluorescence in the cell layers on the tissue surface. Notably, however, the cell layers below were also stained, albeit with decreasing intensity. This was also observed when tumor xenograft tissue slices were treated with siDY647-containing complexes based on unmodified branched low-molecular-weight PEI F25-LMW or tyrosine-modified branched 10 kDa PEI (P10Y; Supplemental Figure S2). Thus, a gradient was observed, which could be explained by the rapid cellular uptake by outer cells and slower penetration into deeper tissue regions. This was expected and more closely resembled the in vivo situation with regard to penetration into intact tissue.
The delivery of specific siRNAs led to target gene knockdown. In the case of tissue slice cultures from PC3-EGFP tumor xenografts, this was readily visible when analyzing EGFP expression levels in the tissue slices by confocal microscopy ( Figure 2E). A marked reduction in EGFP fluorescence intensity was observed upon transfection with siEGFP-containing complexes, while treatment with equally complexed siCtrl resulted in no alterations compared to untreated cells.
The findings also indicated the absence of possible non-specific toxicity associated with the nanoparticle treatment of the tissue slices.
While the observed knockdown was mainly related to surface cells, and was expected, the preparation of full tissue slice lysates allowed monitoring of the whole tissue section. Using RT-qPCR, mRNA prepared from tissue slice lysates was analyzed for GAPDH mRNA levels as an endogenous target gene. Notably, a knockdown of GAPDH of approximately 50% knockdown was observed ( Figure 2F). Given the considerable thickness of the tissue slices (∼300 μm), which harbored many cell layers, the finding indicates a deeper tissue penetration that is required to explain this overall reduction in GAPDH mRNA. Direct comparison with the results from the fluorophore-labeled siRNA experiments (see Figure 2D) also suggests that even smaller quantities of siRNA (as indicated by the rather faint fluorescence in the inner parts of the tissue slice) are already sufficient for gene knockdown. This was also observed in tissue slice cultures derived from xenografts of other tumor cell lines, including HROC24 colon carcinoma and A549 lung adenocarcinoma cells ( Figure 2F). The collective findings indicate that LP10Y/siRNA nanoparticles can penetrate intact (tumor) tissue and exert efficient gene knockdown; this tissue slice setup is thus well suited for monitoring these effects. Essentially no knockdown efficacy was seen in the case of P2Y/siRNA and LP25/siRNA complexes ( Figure 3A). This is consistent with previous findings demonstrating the weaker complexation efficacy of tyrosine-modified, very small, 2.5 kDa branched PEI, [12] and very little overall complexation when using nonmodified linear 25 kDa PEI. [19] When switching to the 3D tissue slice model, the overall gene targeting efficacy was lower, as expected from impaired tissue penetration. The  Figure 3C) revealed some correlation for many polymeric nanoparticles tested. In several cases, this allowed the extrapolation from 2D cell culture results to 3D tissue knockdown. However, several remarkable exceptions were evident. The branched tyrosine-modified PEIs P5Y and P25Y showed poorer knockdown efficacy in 3D tissue slices than that expected from 2D cell culture. This was even more pronounced for PPI-Y/siRNA complexes, which can be explained by their large size (Supplemental Table S2) prohibiting efficient tissue penetration in 3D tissue architecture. In contrast, 2D cell culture tends to underestimate the biological efficacies of PEI F25-LMW and in particular the LPP in a 3D setting, demonstrating gene targeting that greatly exceeds the expectations of 2D cell culture. This is consistent with the pronounced efficacies previously observed in vivo [22] and also applies to LP10Y complexes, which are also above the line. [19] In particular in the case of LPPs, this may reflect lower positive surface charges (for zeta potentials, see Supplemental Table S2), facilitating tissue penetration, while retaining their very high cell uptake as described above.

Analysis of biological efficacies of various PEI-based, siRNA containing nanoparticles in direct comparison with 2D gene knockdown
These conclusions were also supported by Pearson's correlation analyses. As shown in Figure 3C However, the correlation was more pronounced in the case of the 3D system (R = −0.653) and was statistically significant (p = 0.0214; Supplemental Figure S3). The findings highlight the differences between 2D and 3D systems, and the advantages of the latter. Furthermore, the findings demonstrate that the nanoparticle size plays a larger role in intact 3D tissue than in 2D monolayer culture.

Assessment of nanoparticle cytotoxicity
In addition to the biological activity of nanoparticles, their biocompatibility is also of major relevance. This is particularly true for positively charged nanoparticles, which are somewhat prone to exerting cytotoxic effects. Cytotoxicity has also been observed for certain PEIs, especially those with higher molecular weights. Previously, we showed that the tyrosine modification of PEIs leads to higher biocompatibility. When comparing the larger set of nanoparticles in 2D cell culture in a standard LDH release assay, no acute cell damage was observed ( Figure 4A). This was also confirmed by microscopic inspection of the cells ( Figure 4A, right panel, for representative examples). As shown in Figure 4B, it was possible to perform this assay also in our 3D have previously been shown to exert only poor transfection efficacy for plasmid DNA, probably because of too high complex stability and thus insufficient intracellular DNA release from the complex ( [20] and unpublished data). Accordingly, these polymers were excluded here.
Instead, we added the disulfide (SS) bridged, tyrosine-modified, very low-molecular-weight PEI derivative SSP2Y, which shows particularly high transfection efficacy for plasmid DNA. [20] Again, the direct comparison between standard 2D cell culture ( Figure 4C) and the 3D tissue slice model ( Figure 4D) revealed comparable increases in the positive control vs. background values, albeit at a higher level in the 3D setting.
However, the 3D tissue slice model was again more capable of identifying differences in biocompatibility; this was evident from the slight cytotoxicity of the linear 25 kDa PEI (LP25) and PEI F25-LMW, which was not observed upon tyrosine modification ( Figure 4D).

Analysis of biological gain-of-function efficacies (plasmid DNA transfection)
Initially, using SSP2Y, reporter gene expression was observed when tissue slices from tumor xenografts based on PC3 wild-type cells were transfected with luciferase (pGL3) or EGFP encoding expression plasmids (pEGFP). This was true when analyzing tissue lysates (luciferase activity upon pGL3 transfection; Figure 5A) or digested tissue slices for EGFP-positive cells using flow cytometry ( Figure 5B).
EGFP reporter gene expression was also observed using confocal microscopy ( Figure 5C).
Analysis of the whole set of selected polymers for transfection efficacy in standard 2D cell culture revealed major differences ( Figure 5D). SSP2Y showed particularly high reporter gene expression. Fractionated PEI F25-LMW was more efficient than its "parent" 25 kDa PEI The order SSP2Y > LP25 > F25 > P25 was also observed in the 3D tissue slice culture model, with the advantage of SSP2Y over LP25 and F25 being even more profound ( Figure 5E). A scatter plot to directly compare 2D and 3D transfection efficacies also revealed that some complexes based on linear tyrosine-modified PEIs showed somewhat higher activity in 3D, as expected from 2D standard cell culture, albeit at a comparably low level ( Figure 5F). In contrast, the expression levels of the unmodified linear or branched PEIs (LP25, P25, and PEI F25-LMW) in the 3D model were below the expectations of 2D cell culture, indicating poor tissue penetration or cell uptake in the 3D architecture.
Notably, this was particularly true for the 25 kDa PEI, often considered as the gold standard in PEI transfection.
Taken together, these results indicate that the 3D tissue slice model allows a finer distinction between nanoparticles with regard to biological activities and possible cytotoxic effects compared to classical 2D cell culture. representing the initial histological architecture and genetic profile, even after several rounds of propagation in mice. [25] However, one limitation of this approach is the absence of immune cells when using tumor material from immunodeficient mice. Thus, the immunomodulatory effects of nanoparticles will require tissue material from an immunocompetent background. However, in these tissues, immune cells, such as T lymphocytes, can retain their viability in ALI tissue slice culture for several days, [26] which is sufficient for assessing Given the comparably large size of the nanoparticles, their deeper penetration into intact tissue is particularly notable. It should be noted, however, that mechanical tissue properties and in particular their solid or fluid behavior have been shown to influence the spread of malignant tumors. [27] Similarly, the mechanical properties of the stroma, such as its architectural network and tissue matrix stiffness, affect cancer cell migration. [28] It can be assumed that this may also apply to nanoparticle penetration. This highlights the necessity to conduct experiments in an intact tissue environment. However, the tissue penetration of nanoparticles is affected by their size, as evident in the present study by the direct comparison of the different nanoparticles. This goes well beyond 2D cell culture, where previous studies from other groups and our lab results have shown that nanoparticles, even in the range of 200-800 nm, are still able to enter target cells via endocytosis. [21] Nanoparticle zeta potential has also been identified as a major determinant of tissue penetration, [13] which can be readily explained for example by the interaction of positively charged nanoparticles with negatively charged ECM components, such as heparansulfate proteoglycans. While cellular uptake in classical 2D cell culture may actually benefit from this interaction, it will hinder nanoparticle penetration into deeper cell layers of the tissue, leading to poorer biological activity.

DISCUSSION
Thus, the identification of optimal nanoparticles based on their cellular uptake in 2D culture may lead to the selection of wrong candidates with suboptimal properties ex vivo and in vivo. In contrast, nanoparticles were observed here, whose performance in tissue slices was higher than expected from 2D cell culture and which have already shown high activity in vivo. Thus, despite the absence of certain in vivo properties, such as intact tissue blood flow and perfusion, tissue slice models will be able to more accurately define optimal nanoparticle candidates for subsequent in vivo studies.