Localization of drug biodistribution in a 3D-bioengineered subcutaneous neovascularized microenvironment.

Local immunomodulation has shown the potential to control the immune response in a site-specific manner for wound healing, cancer, allergy, and cell transplantation, thus abrogating adverse effects associated with systemic administration of immunotherapeutics. Localized immunomodulation requires confining the biodistribution of immunotherapeutics on-site for maximal immune control and minimal systemic drug exposure. To this end, we developed a 3D-printed subcutaneous implant termed 'NICHE', consisting of a bioengineered vascularized microenvironment enabled by sustained drug delivery on-site. The NICHE was designed as a platform technology for investigating local immunomodulation in the context of cell therapeutics and cancer vaccines. Here we studied the ability of the NICHE to localize the PK and biodistribution of different model immunomodulatory agents in vivo. For this, we first performed a mechanistic evaluation of the microenvironment generated within and surrounding the NICHE, with emphasis on the parameters related to molecular transport. Second, we longitudinally studied the biodistribution of ovalbumin, cytotoxic T lymphocyte-associated antigen-4-Ig (CTLA4Ig), and IgG delivered locally via NICHE over 30 days. Third, we used our findings to develop a physiologically-based pharmacokinetic (PBPK) model. Despite dense and mature vascularization within and surrounding the NICHE, we showed sustained orders of magnitude higher molecular drug concentrations within its microenvironment as compared to systemic circulation and major organs. Further, the PBPK model was able to recapitulate the biodistribution of the 3 molecules with high accuracy (r ​> ​0.98). Overall, the NICHE and the PBPK model represent an adaptable platform for the investigation of local immunomodulation strategies for a wide range of biomedical applications.

Introduction

Modulation of the immune system is widely employed in various medical applications, including the management of foreign body response, enhancement of tissue repair and regeneration, and abrogation of transplant rejection. Conventional immunomodulatory strategies hinge on lifelong systemic administration of therapeutics, resulting in whole-body drug exposure, which is often associated with side effects and increased risks of different diseases and opportunistic infections [[1], [2], [3], [4]]. These adverse effects and risks may cause poor quality of life and affect therapeutic adherence, which ultimately hinders efficacy [5,6].

To address these challenges, localized immunomodulating strategies are emerging as attractive alternatives for the treatment of various conditions [[7], [8], [9], [10]]. In this context, a wide array of technologies have been explored for cell transplantation [7,11], wound healing [[12], [13], [14]], cancer vaccines [[15], [16], [17], [18]] and allergy treatment [19,20], among others. Nanoparticle constructs have been developed to achieve drug localization from systemic administration and reduce systemic drug exposure via active molecular targeting [[21], [22], [23]]. However, in most cases, less than 5% of the drug reaches the intended target due to the challenges of breaching through biological barriers [24]. The use of micro-nanocarriers, such as microparticles, micelles or liposomes, paired with localized delivery, has shown enhancement of drug localization and therapeutic index, whilst reducing their circulating levels [[25], [26], [27]]. However, drug release from these platforms is usually exhausted within days [28,29], and maintenance of long-term treatment would require frequent administrations, which limits their applicability. Immunomodulatory hydrogel and implantable scaffolds have also been developed and functionalized with immunomodulating molecules [14,30,31] to control the immune system in various settings [32]. While these provide longer-acting drug release, no mechanism is available to extend release from these systems upon drug exhaustion, requiring replacement or de novo administration. Similar considerations apply for non-refillable reservoir-based implantable devices, which can extend delivery of therapeutic on site via nanofluidic membranes for months [33,34], yet are unable to fulfill the delivery requirements for chronic applications. Overall, these strategies show favorable outcomes in the short term. However, they are largely inadequate in providing a long-lasting immunomodulatory effect due to drug depletion, pointing towards the need for replenishable platforms [35,36]. Further, and equally significant, spatiotemporal assessment of the local biodistribution of immunomodulatory drug from these systems and their resulting immunomodulatory effect in situ is challenging. For these reasons, these technologies are unsuitable for the investigation of novel local immunomodulatory approaches.

To this end, we developed a 3D-printed implantable system termed ‘NICHE’. Upon engraftment, the NICHE hosts a spatially-defined bioengineered vascularized microenvironment, which can be manipulated through sustained and transcutaneously replenishable delivery of molecules on-site. The NICHE was conceived as a platform technology for investigating local immunomodulation in the context of cell therapeutics [[37], [38], [39]]and cancer vaccines [16]. The NICHE consists of 2 compartments: a vascularized tissue reservoir and a surrounding refillable drug reservoir [40]. The compartments are separated by a nanoporous membrane that allows controlled elution of immunomodulatory agents from the drug reservoir to the cell reservoir. The latter is enclosed by 2 nylon meshes that allow for tissue and capillaries ingrowth. We previously showed this local delivery platform to maintain viability, function and retention of allogeneic cell grafts (Leydig cells and pancreatic islets) in immunocompetent rats without the need for systemic immunesuppression [37,38].

While we aim to employ NICHE to house transplanted insulin producing cells, such as pancreatic islets or stem cells derived beta cells for the treatment of type 1 diabetes, NICHE-like platforms are applicable to a broad spectrum of local immunomodulatory applications. Hence, here we sought to investigate the ability of the NICHE technology to effectively localize the biodistribution of immunomodulating molecules and minimize systemic drug exposure. For this, we employed 3 model molecules with different molecular weights and conformations as representatives of a wide spectrum of immunomodulatory agents: CTLA4Ig (92 ​kDa), a clinically approved immunosuppressant, human IgG (150 ​kDa) as a model for monoclonal antibody-based immunosuppressants and ovalbumin (42 ​kDa) as a model for fusion protein antibodies with lower molecular weights. The analysis was performed in rats with NICHE implanted in the subcutaneous space, which is regarded as a promising site for cell delivery due to accessibility and minimal invasiveness. Molecules concentration was monitored longitudinally within NICHE microenvironment and surrounding tissues, in the plasma and major organs over 30 days. Characterization of tissue properties within NICHE and its surroundings, inclusive of cell, blood vessel, and collagen density was performed with emphasis on molecular transport. In this context, using fluorescence recovery after photobleaching (FRAP), we measured the effective diffusivity of the 3 molecules within relevant tissues. Finally, the data was used to develop a broadly applicable physiologically-based pharmacokinetic (PBPK) model to describe the biodistribution of molecules locally eluted via NICHE. Overall, this work provides an engineered microenvironment technology platform and framework for the development and investigation of novel local immunomodulation strategies.

Materials and methods

NICHE fabrication

NICHE was fabricated as previously described [37,38]. Briefly, two polyethersulfone (PES) nanoporous membranes (30 ​nm, Sterlitech) and two sets of nylon meshes (Elko Filtering) were affixed on the 3D printed polyamide structure (30.4 ​mm ​× ​15.4 ​mm x 3.8 ​mm) with implantable grade silicone adhesive (MED3-4213, Nusil). The same adhesive was used to fabricate the silicone ports. All parts were autoclaved prior to sterile assembly under a laminar flow hood. Assembled NICHEs were sterilized at the Houston Methodist Research Institute (HMRI) Current Good Manufacturing Practice core with ethylene oxide gas.

Scanning electron microscopy (SEM) images

Nylon meshes and PES membranes were imaged with the Nova NanoSEM 230 ​at the HMRI SEM core. The samples were sputtered with 7 ​nm iridium and imaged using a 5 ​kV electron beam under high vacuum (10−3 ​Pa).

Animal experiments

In this study, eight-week-old male F344 rats (Charles River, Houston, TX, USA) were used. All animals were maintained and employed in conformity with guidelines established by the American Association for Laboratory Animal Science. Rats were kept in the HMRI animal facility and studies were conducted at the HMRI Comparative Medicine Program (CMP) according to provisions of the Animal Welfare Act, PHS Animal Welfare Policy, and the principles of the NIH Guide for the Care and Use of Laboratory Animals. The humane use of animals in research and all procedures detailed in the IACUC protocol number IS00005894 were approved by the Institutional Animal Care and Use Committee at HMRI. Animals were housed under standard conditions and had ad libitum access to water and a standard laboratory diet.

NICHE deployment

Sterile NICHE cell reservoir was loaded with mesenchymal stem cells (MSCs) to drive vascularization and tissue formation within the reservoir as described in our previous publications [37,38]. Briefly, a pluronic F-127 (Sigma) hydrogel (20% PF-127 in DMEM) was used to suspend F344 rat bone marrow MSCs (Cyagen). The mix was then injected into the cell reservoir of NICHEs (5 ​× ​105 MSCs per NICHE) through the central silicone port. MSC-loaded NICHEs were implanted in a subcutaneous pocket created via a 2 ​cm incision in the rat dorsum, as previously described [37,38].

Histology and immunohistochemistry analysis

For histological analysis, 2, 4, and 6 weeks post-implantation, n ​= ​4 NICHEs were explanted as a whole with the fibrotic capsule through terminal surgery. Tissues were fixed in 10% formalin for 5 days followed by dehydration in standard ethanol. NICHE polyamide structure was removed from the fixed tissue to facilitate sectioning. Dehydrated tissues were washed in xylene prior to paraffin embedding.

Hematoxylin-eosin (H&E) and Masson's Trichrome (MT) staining were performed on 5 ​μm sections at the HMRI Research Pathology Core. Sections scans were obtained with EVOS imaging system (Life Technologies), while magnified fields of view (FOV) were acquired with Olympus B×71 microscope equipped with Olympus DP72 camera and 40×/0.75 objective. Fibrotic capsule and skin thickness measurements were performed on MT-stained slides using ImageJ (n ​= ​5 measurements per sample, n ​= ​3 samples).

For immunofluorescence staining, 5 ​μm sections were subjected to heat-induced epitope retrieval in rodent decloaker buffer (Biocare Medical) at 121 ​°C for 20 ​min followed by blocking in 5% normal goat serum for 1 ​h at room temperature. Primary antibodies were incubated for 16 ​h at 4 ​°C in renaissance antibody diluent (Biocare Medical) and secondary antibodies for 1 ​h at room temperature in blocking buffer. Prolong Diamond mounting media with DAPI was added to preserve fluorescence (Invitrogen). Slides were visualized using FluoView 3000 confocal microscope (Olympus). Antibodies used were: alpha-smooth muscle actin (a-SMA; ab56894, Abcam), AlexaFluor 555 goat anti-rabbit (A21428, Invitrogen).

Cell, capillaries and collagen density analysis

The slides were visually divided in 4 macro-areas: Subcutaneous tissue on top of NICHE (SubQ), fibrotic capsule surrounding NICHE (Fib. Cap.), cell reservoir tissue close to the membranes (side) and tissue in the center of the cell reservoir (center).

DAPI and α-SMA stained sections were used for cell and capillary density analysis. Five FOV per macro-area were randomly acquired (n ​= ​4 sections). Olympus cellSens software was used to automatically count DAPI-positive nuclei and to measure α-SMA-positive area. The measurements were used to calculate cell density and capillary area fraction respectively, according to the formulas below.

MT-stained sections were used for collagen density analysis. Five FOV of each macro-area were randomly acquired (n ​= ​4 sections). Collagen area was measured through color deconvolution followed by color thresholding using a custom Matlab script (Mathworks). Collagen density was calculated according to the formula below.

Molecules fluorescent labeling

CTLA4Ig (Orencia, Bristol-Meyers Squibb), human IgG (Sigma), and albumin from chicken egg white (Ovalbumin, Sigma) were reconstituted and conjugated to Alexa Fluor 647 NHS ester, Alexa Fluor 488 NHS ester and Alexa Fluor 568 NHS ester (Invitrogen), respectively, following the protocol provided by the manufacturer. Stock solutions for each molecule (10 ​mg/ml) were prepared by mixing unlabeled and Alexa Fluor-conjugated molecules at a 19:1 ratio in water for injection (USP sterile grade, RMBIO). A cocktail solution (COMBO) containing 10 ​mg/ml of CTLA4Ig, Human IgG and Ovalbumin was also prepared by mixing unlabeled and Alexa Fluor-conjugated molecules at a 19:1 ratio in water for injection (USP sterile grade, RMBIO). CTLA4Ig was reconstituted and conjugated to Fluorescein-5-Isothiocyanate (FITC, Invitrogen) according to manufacturer instructions.

Tissue diffusion coefficient measurement

Vascularized NICHEs were explanted from n ​= ​3 rats 6 weeks post-implantation. The subcutaneous tissue above the device was collected and the fibrotic capsule was dissected from the nylon mesh. Devices were sectioned in half across their width (cutting plane parallel to the ports face of the device) and the tissue contained in the cell reservoir was carefully excised. Each tissue was divided in 3 parts and placed in media (DMEM/F-12) solutions at a 2 ​mg/ml concentration (n ​= ​3 set of tissues per solution) containing either CTLA4Ig-FITC, IgG-FITC (Sigma), or ovalbumin-FITC (Invitrogen), respectively. Control tissues were collected and incubated in media. After incubation overnight at 4 ​°C, the tissues were removed from the media solutions and placed in microscope chamber slides (Nunc Lab-Tek II, Thermofisher). The tissue diffusion coefficients of CTLA4Ig, IgG and ovalbumin were measured through FRAP method [41,42]. The cell reservoir tissues were virtually divided in 2 macro-areas: cell reservoir tissue in proximity to the drug-releasing membranes (side), and tissue in the center of the cell reservoir (center). An Olympus Fluoview 3000 microscope equipped with a 488 ​nm laser and a 10× objective was used to take measurements in each sample. FRAP measurements were performed as follows: 1) 3 pre-bleaching images (512 ​× ​512 pixels) of a selected FOV were collected at low laser intensity (1%); 2) a circular area (50 pixels or 124 ​μm radius) was bleached by exposing it at 100% laser intensity for 100 ​s; 3) fluorescence recovery was detected by imaging the bleached area at low intensity for 30 ​min (12 images/minute).

The images were analyzed using a custom Matlab script. Briefly, the normalized fluorescent intensity measured in the bleached area over time was fitted to the following equation [43,44]:

The bleaching parameter, b, and the characteristic diffusion time, τd, were fitted. τd was used along with the radius of the bleached area, ω, to calculate the diffusion coefficient, D, according to:

Stability assessment

CTLA4Ig, IgG and Ovalbumin are dissolved in water for injection to a concentration of 2 ​mg/ml. Each solution is divided in n ​= ​4 1.1 ​ml aliquots. One aliquot per solution is frozen at −80 ​°C while the others are incubated at 37 ​°C. After 4 and 8 weeks one aliquot per solution is removed from incubation and stored at −80 ​°C. Frozen aliquots are thawed and the molecules stability is assessed through gel permeation chromatography (GPC) with a Viscotek GPCmax (Malvern) equipped with a column for protein samples (P3000, Malvern). Phosphate saline buffer (PBS) with 0.02% sodium azide is used as GPC buffer.

In vivo biodistribution experiment

NICHE-MSCs were implanted in n ​= ​32 rats. At week 6 post-implantation, the animals were randomly placed in 2 groups: a loaded group (n ​= ​28) where NICHEs drug reservoirs were loaded with the COMBO solution, and a control group where the devices were not loaded. The loading procedure was carried out as described in our previous publications [38,45]. At day 1, 3, 6, 10, 15, 22 and 30 post drug loading, a group of rats was euthanized (n ​= ​4 per time-point) through CO2 asphyxiation. After euthanasia blood was collected through cardiac puncture and NICHE was retrieved from the subcutaneous pocket. Drug reservoir content, cell reservoir tissue, fibrotic capsule, skin above the pocket, ipsilateral axillary lymph nodes, liver, kidney and spleen were collected and processed for molecules quantification.

Ex-vivo molecules concentration measurement

The tissue harvested were homogenized in T-PER buffer supplemented with protein inhibitor cocktail (Thermo Scientific). Tissue homogenates and blood were clarified via centrifugation and stored frozen until analysis.

Molecules concentration in plasma and tissue homogenates was quantified using human CTLA4 ELISA (Invitrogen), Chicken Ovalbumin ELISA (Mybiosource) and human IgG IQELISA (Raybiotech).

Drug reservoir residual concentration was measured through fluorescence using a plate reader (Synergy H4, Biotek) and analyzing the samples in duplicates.

Statistical analysis

All statistical analyses were performed using Prism 9.3.1 (Graphpad). The data is presented as mean ​± ​standard deviation (∗p ​< ​0.05, ∗∗p ​< ​0.01, ∗∗∗p ​< ​0.001, ∗∗∗∗p ​< ​0.0001).

Calculation

Theoretical basis of the model

Molecules are released from NICHE drug reservoir into its cell reservoir where vascularization and tissue ingrowth were driven by MSCs. The permeability and exchange surface of the nanoporous membrane control the release of molecules into the extracellular space of the cell reservoir. There, diffusion through the extracellular space allows them to reach the surrounding fibrotic capsule and subcutaneous tissue. In these areas, the molecules are cleared through the lymphatic and the vascular system at different rates based on size. The molecules are carried by the lymph flow into lymph nodes and eventually in the thoracic duct which drains in the subclavian vein. Once in the vascular network, the blood is responsible for the distribution of the molecules in major organs. The release kinetics, diffusion dynamics, and transport phenomena are described with a mechanistic approach in the PBPK model.

NICHE membrane permeability rate

The membrane permeability of the j-th molecule is calculated using its concentration measured in the NICHE drug reservoir residual collected during necropsy. According to Fick's law of diffusion:where Δm is the difference between the molecule mass loaded in NICHE and the molecule mass measured after necropsy for the k different molecules, Δt is the time between NICHE loading and necropsy, Amem is the exchange surface area of the PES membrane, Pmem is the membrane permeability and ΔC is the difference in concentration between the drug reservoir and the cell reservoir immediately after loading. The membrane permeability is derived as follows:

The membrane permeability is used to calculate the membrane permeability rate (kmem,j) that was used in the model:where Vdr is the volume of the drug reservoir.

NICHE tissue diffusion rate

The tissue diffusion rate described the transport of molecules within NICHE extravascular space. In particular, we modeled the transport between the cell reservoir tissue and the fibrotic capsule (kfc), and between the fibrotic capsule and the surrounding subcutaneous tissue (ksk) as follows:where D is the diffusion coefficient of the j-th molecule in the cell reservoir, Dfs is the diffusion coefficient in the fibrotic capsule and surrounding skin, xfc is the characteristic diffusion length between the cell reservoir and the fibrotic capsule, and xsk is the characteristic diffusion length between the fibrotic capsule and the surrounding subcutaneous tissue (Section S1.1). D is the average of the diffusion coefficients in the center and side macro-areas of the cell reservoir, while Dfs is the average of the diffusion coefficients in the fibrotic capsule and subcutaneous tissue (Section 2.8).

Blood vessels permeability

Molecules transport across the vascular endothelium can happen through bulk transport, when they are carried by the lymph fluid, or through diffusion. The bulk transport of molecules is size-dependent and therefore regulated by the fenestration on the blood vessels. This phenomenon is characterized by the reflection coefficient σ, specific for each tissue i and each molecule j, and defined by the following relation [46,47]:where α represents the ratio between the j-th molecule hydrodynamic radius and the blood vessel pore size of the i-th organ. G and F are decreasing hydrodynamic functions (Section S1.2). Molecular extravasation driven by diffusion is described through the permeability-surface area product P∗S, where the S, is calculated according to the model developed by Dogra et al. [48] (Section S1.1).

Physiologically based pharmacokinetic model development

A PBPK model was adapted to predict the local and systemic pharmacokinetics of drugs following release from NICHE (Fig. 1). The model comprises of 10 compartments, where 4 compartments represent the NICHE and its surrounding tissue (drug reservoir, cell reservoir, fibrotic capsule, and skin), 3 compartments represent the major organs (liver, spleen, kidney), and 3 compartments represent the connective tissues (lymph nodes, lymph, and blood). The various organ compartments and components of NICHE that support vascularization are modeled to contain a vascular and an extravascular sub-compartment. Vascular sub-compartments are connected to the blood compartment via blood flow (Q), represented by red arrows in the model schematic, while extravascular sub-compartments are connected to the lymph node via lymph flow (L), represented by blue arrows. The extravasation of drugs from vascular to extravascular sub-compartments is advection or diffusion-driven and is characterized by parameters defined in section 3.1.3. All vascular sub-compartments include a degradation rate constant (kdeg) to characterize the natural decay of molecules [49]. The various transport and physiological processes in the model are formulated as a system of ordinary differential equations and solved numerically as an initial value problem in Simbiology (MATLAB 2021b, Mathworks).

Fig. 1

PBPK model. Gray box represents the NICHE device and its neighboring tissues. Vascular compartments, sub-compartments, and their connectivity via blood are represented in red. Extravascular compartments, sub-compartments, and their connectivity are represented in light blue. Transport phenomena between sub-compartments are represented by white arrows. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

NICHE compartments

The NICHE microenvironment is described by 4 compartments: a drug reservoir compartment (dr), and 3 tissue compartments to model the cell reservoir tissue (cr), the fibrotic capsule (fc) surrounding NICHE and the skin/subcutaneous tissue above the device (sk). The drug reservoir compartment is connected to the extravascular sub-compartment of the cell reservoir to model molecular diffusion across the nanoporous membrane. The extravascular sub-compartments of cell reservoir, fibrotic capsule, and skin are sequentially connected to describe the diffusion of molecules in the extracellular space of NICHE microenvironment as defined in section 3.1.2.

Blood, lymph and lymph nodes compartments

The blood compartment (b) supplies blood to all vascular sub-compartments and receives the venous return from those organs, in addition to lymph (ly) drainage from the lymphatic node (ln).

Major organ compartments

Three compartments are used to model liver (l), spleen (s), and kidney (k), which are the major determinants of drug pharmacokinetics, especially for macromolecules.

Model parameters

The parameters used in the model to describe the i-th tissue and j-th molecule are obtained from the literature or derived from other mathematical models (Table 1, Table 2, Table 3, Table 4). Lymph flow values for each compartment were assumed to be 500 times smaller than the corresponding blood flow [50]. The blood vessel permeability parameters (P) and degradation rate (kdeg) were estimated through least square fitting.

Table 1

NICHE compartments parameters.

Parameter	Description (Unit)	Value				Source
		Drug reservoir (dr)	Cell reservoir (cr)	Fibrotic capsule (fc)	Subcutaneous tissue (sk)
di	Thickness (cm)	–	0.38	0.0317	0.1872	Section 2.1, Section 2.5
Vi	Total volume (ml)	0.3	0.386	0.364	2.767	Section S1.1
Vi,v	Vascular volume (ml)	–	0.035	0.013	0.05	Section S1.1
Vi,e	Extravascular volume (ml)	–	0.128	0.12	0.915	Section S1.1
Qi	Blood flow (ml/h)	–	7.75	2.84	1.09	Section S1.1
Si	Vascular surface area (cm2/cm3)	–	362.61	141.29	72.43	Section S1.1

Table 2

Blood, lymph and lymph node compartments parameters.

Parameter	Description (Unit)	Value			Source
		Blood (b)	Lymph (ly)	Lymph node (ln)
Vi	Total volume (ml)	8.1	7	1	[50,51]
Li	Lymph flow (ml/h)	–	2.32	0.4	[50,52]

Table 3

Major organs compartments parameters.

Parameter	Description (Unit)	Value			Source
		Liver (l)	Spleen (s)	Kidney (k)
Vi	Total organ volume (ml)	14	2.5	2.15	[50]
Vi,v	Organ vascular volume (ml)	1.2	0.3	0.12	[50]
Vi,e	Organ extravascular volume (ml)	2.3	0.5	0.32	[50]
Qi	Organ blood flow (ml/h)	1044	159.82	325.89	[50,53]
rpore,i	Organ blood vessels fenestration radius (nm)	140	2500	7.5	[54]

Table 4

Molecular parameters.

Parameter	Description (Unit)	Value			Source
		Ovalbumin	CTLA4Ig	IgG
rj	Radius (nm)	3.05	5.1	5.29	[[55], [56], [57]]
Pmem,j	Membrane permeability (cm/s)	1.67 ​× ​10−6	1.03 ​× ​10−6	7.53 ​× ​10−7	Section 3.1.1
σl,j	Reflection coefficient of liver (dimensionless)	0.0029	0.008	0.0086	Section 3.1.3
σs,j	Reflection coefficient of spleen (dimensionless)	9.57 ​× ​10−6	2.67 ​× ​10−5	2.87 ​× ​10−5	Section 3.1.3
σk,j	Reflection coefficient of kidney (dimensionless)	0.51	0.878	0.901	Section 3.1.3
Dj	Diffusion coefficient in cell reservoir (cm2/s)	11.28 ​× ​10−8	7.78 ​× ​10−8	7.39 ​× ​10−8	Section 3.1.2
Dfs.j	Diffusion coefficient in surrounding tissue (cm2/s)	12.67 ​× ​10−8	8.25 ​× ​10−8	8.06 ​× ​10−8	Section 3.1.2

Model equations

The equations representing the model are reported in Section S2 of the supplementary information.

Model parameters estimation

The unknown parameters were estimated for the 3 molecules by fitting the model to the experimental data. The fit was performed by combining a genetic algorithm and a non-linear least square solver, using the ga and lsqnonlin functions in Simbiology.

Local sensitivity analysis

To evaluate the effect of model parameters on drug concentration in various model compartments, we performed a local sensitivity analysis, where we evaluated the time-dependent sensitivities of drug concentration in various model compartments with respect to initial conditions and parameter values in the model. The local sensitivity was obtained from the following equation:where C is the concentration of the j-th molecule in the i-th compartment and k is a parameter. The results were integrated over time to obtain a dimensionless sensitivity value.

Results and discussion

NICHE fabrication and tissue characterization

The NICHE device (Fig. 2A) was designed and fabricated as described in our previous publications [37,38]. The NICHE central cell reservoir is surrounded by a U-shaped drug reservoir (Fig. 2B) and enclosed within 2 overlapping sets of nylon meshes presenting different openings (Fig. 2C): the inner meshes have 300 ​μm ​× ​300 ​μm square openings and ∼120 ​μm threads and provides mechanical support; the outer meshes have 100 ​μm ​× ​100 ​μm openings and finer ∼80 ​μm threads that favors cell retention while permitting vascular and tissue ingrowth. Upon subcutaneous implantation of MSC-loaded NICHE, tissue and neovasculature progressively penetrate across the meshes, filling and engrafting the cell reservoir within 3–4 weeks [37,45]. Long-acting sustained drug elution from the drug reservoir to the vascularized tissue in the cell reservoir is accomplished for months via a nanoporous membrane (Fig. 2D and E). As previously shown, long term deployment of the platform can be achieved via straightforward drug and cell replenishment with transcutaneous injections through the silicone ports [37].

Fig. 2

NICHE description. A) Picture of a fully assembled NICHE (Scale bar 5 ​mm). B) Rendering of longitudinal section of NICHE. C) SEM image of nylon meshes bilayer (500× magnification, scale bar 500 ​nm). D) SEM image of PES nanoporous membrane (50,000× magnification, scale bar 100 ​μm). E) Rendering of subcutaneously implanted NICHE cross section, depicting the diffusive local delivery of immunosuppressant into a vascularized cell reservoir.

Histological sections obtained from NICHE devices explanted 6 weeks post-implantation revealed a thin fibrotic capsule surrounding the device (Fig. 3A). The tissue penetrating the cell reservoir presented higher density in proximity of the membrane (Fig. 3A and B). Fibrotic capsule thickness increased over time up to 317 ​± ​64 ​μm ​at week 6 (Fig. 3C). Image analysis of sections stained with DAPI and α-SMA indicates increased cell density within NICHE cell reservoir as compared to subcutaneous tissue (Fig. 3B). Moreover, vascular-area fraction measurements showed higher blood vessels density in the cell reservoir compared to both fibrotic capsule and subcutaneous tissue (Fig. 3E). In contrast, higher density of collagen was observed in the fibrotic capsule and subcutaneous tissue than in the NICHE cell reservoir (Fig. 3F). As demonstrated in our previous work, tissue growth within the NICHE is influenced by the MSCs retained in the cell reservoir [37,45], which are known to secrete anti-inflammatory and pro-angiogenic factors [[58], [59], [60], [61]]. In fact, lower collagen deposition in the cell reservoir compared to the fibrotic capsule is indicative of the local MSC mitigation of inflammatory foreign body response [62]. Interestingly, tissue adjacent to the membrane showed increased collagen density compared to the tissue in the center of the cell reservoir. Finally, dense vascularization was observed in the cell reservoir, attributable to the angiogenic effect of MSCs, which is congruent with our previous findings [37,45].

Fig. 3

NICHE tissue histology and characterization. A) Histological image of MT-stained cross section of excised NICHE tissues at week 6 post-implantation. Pound signs (#) represent the area occupied by the drug reservoir. The lines represent the division of the cell reservoir tissue in macro-areas, side (s) and center (c). The fibrotic capsule (f) is visible around the drug reservoir (Scale bar 1 ​mm). B) 40× magnifications of H&E, MT and DAPI and α-SMA stained NICHE cell reservoir scans (Scale bars 50 ​μm) explanted at week 6 post-implantation. C) Fibrotic capsule measurements over time. Quantification of D) cell density, E) capillaries area fraction and F) collagen area fraction in NICHE tissues at week 6 post-implantation.

The tissue formed in the cell reservoir showed distinct histological characteristics compared to unaltered subcutaneous tissue. To this extent, diffusion dynamics of molecules within this space were characterized via FRAP (Fig. 4A). Fluorescence recovery curves indicated a faster recovery in the central macro-area compared to the one adjacent to the membrane (Fig. 4B–D), in agreement with the histological observations. The diffusion coefficients, calculated through fitting of the recovery curves of each molecule, decreased in correspondence to the increase in respective hydrodynamic diameters (Fig. 4E). Furthermore, this behavior was visually confirmed by the shorter recovery time of ovalbumin compared to CTLA4Ig and IgG (Fig. 4A–S1). Interestingly, cell reservoir diffusion coefficients were comparable to the ones calculated for the subcutaneous tissue.

Fig. 4

Diffusion coefficient measurement in NICHE tissues. A) Representative FRAP imaging sequence of the cell reservoir center saturated with CTLA4Ig-FITC (Scale bar 300 ​μm). Recovery curves of B) ovalbumin, C) CTLA4Ig, D) IgG in center, side, fibrotic capsule and subcutaneous tissue macro-areas. No measurable fluorescent signal was observed in control tissues. E. Molecule sizes and calculated diffusion coefficients.

Biodistribution of molecules locally released via NICHE

The stability of the molecules was verified for at least 8 weeks under physiological conditions. All molecules showed no signs of degradation, making them suitable for a long-term biodistribution study (Fig. S2). Molecule concentrations measured in various organs and tissues after euthanasia were used to assess their biodistribution. The calculated cumulative molecules release from the drug reservoir showed a steady sustained trend (Fig. 5A–C). The daily release was estimated through linear regression. Ovalbumin, CTLA4Ig and IgG were released at a rate of 1.98%, 1.71%, 1.7% per day, respectively, indicating a size dependent behavior. IgG showed higher concentration in the cell reservoir compared to the other molecules, probably due to its larger size, which slowed down its clearance (Fig. 5D). Ovalbumin levels in lymph nodes were similar to the other molecules (Fig. 5E), however, in plasma its concentration was 10-fold lower. Interestingly, in kidneys, ovalbumin levels were consistently higher than CTLA4Ig and IgG, indicating some degree of accumulation (Fig. 5F). This difference can be explained by the glomerular barrier cut-off size observed in kidneys that allows for the filtration and subsequent elimination of particles smaller than 6–8 ​nm in diameter [63]. CTLA4Ig and IgG hydrodynamic diameters are both larger than the cut-off (Table 4) and therefore are retained in circulation. Differently, the size of ovalbumin (6.1 ​nm in diameter) allows penetration into the glomerular barrier followed by capture in the proximal tubules [64]. CTLA4Ig and IgG concentrations measured in plasma, axillary lymph nodes and major organs were about 100-folds lower compared to the cell reservoir. Furthermore, the measurements in cell reservoir, fibrotic capsule and skin showed a gradient-like decrease in concentration that points to a predominantly diffusive transport in the space surrounding the NICHE (Fig. 5G–I).

Fig. 5

In vivo molecules biodistribution. Calculated cumulative release from the drug reservoir of A) CTLA4Ig, B) IgG and C) ovalbumin. D) Cell reservoir, E) lymph nodes and F) kidney molecules concentration over time. Violin plot of tissue concentrations measured throughout 30 days normalized to plasma of G) CTLA4Ig, H) IgG and I) ovalbumin.

Overall, this data suggests a biodistribution mostly localized within NICHE, especially for large molecules. Furthermore, the residual molecules in the drug reservoir at day 30 were at least 30% of the loaded dose, indicating the potential of the platform for long term deployment.

PBPK model fitting

The concentrations in blood, lymph node, and NICHE compartments were in good agreement with experimental measurements of the CTLA4Ig biodistribution dataset, while levels in major organs were slightly underestimated (Fig. 6). The Pearson correlation coefficient (r ​= ​0.9903) and weighted residuals analysis however demonstrate goodness of fit; it also highlights that the underestimation occurred mostly for concentrations below 1 ​μg/ml (Fig. S3A). The values in the correlation plot were equally distributed around the identity line (Fig. S3B).

Fig. 6

PBPK model fitted to CTLA4Ig biodistribution data.

Similarly, the IgG concentration data were well fitted by the model despite the large variability in measured plasma levels (Fig. 7). Low concentrations were slightly underestimated, according to the weighted residual plot (Fig. S4A). For this dataset, the correlation plot showed values evenly spread around the identity line (Fig. S4B), resulting in a Pearson correlation coefficient comparable to the one obtained for CTLA4Ig (r ​= ​0.9859).

Fig. 7

PBPK model fitted to IgG biodistribution data.

The fitting of the ovalbumin dataset was optimal in the NICHE compartments (Fig. 8). However, the model overestimated the concentrations in the blood compartment and noticeably underestimated the major organ compartments, especially the kidney. This behavior was reflected in the weighted residual and correlation plots (Figs. S5A–B), which resulted in a slightly lower, but still acceptable, Pearson correlation coefficient (r ​= ​0.9820). Overall, the model obeys the principle of conservation of mass and the total mass decreases over time in accordance with the experimental data (Fig. S6).

Fig. 8

PBPK model fitted to ovalbumin biodistribution data.

The fitting estimated permeability coefficients showed a decreasing trend with increasing molecule size, with the exception of the IgG skin and CTLA4Ig fibrotic capsule permeability coefficients (Table 5).

Table 5

Estimated parameters.

Parameter	Description (Unit)	Molecule
		Ovalbumin	CTLA4Ig	IgG
Pcr,j	Permeability coefficient (cm/h)	8 ​× ​10−5	7 ​× ​10−5	1 ​× ​10−5
Pfc,j	Permeability coefficient (cm/h)	1 ​× ​10−7	1 ​× ​10−5	1 ​× ​10−8
Psk,j	Permeability coefficient (cm/h)	3 ​× ​10−3	5 ​× ​10−4	7 ​× ​10−4
kdeg,j	Degradation coefficient (ml/h)	35	14	10

Notably, the model adequately estimated the concentrations in the NICHE compartments and performed well in describing plasma and lymph node levels for all 3 molecules. However, the model could not accurately represent the ovalbumin kidney tubular capture, resulting in discrepant concentrations in the compartment.

We evaluated the effect of membrane parameters, permeability, diffusion and degradation coefficients on the levels of molecules in various compartments (Fig. 9). Membrane permeability (Pmem) and exchange area (Amem) moderately affected the concentration of molecules in all the compartments. Aside from a predictable impact on the NICHE compartments, the cell reservoir permeability coefficient (Pcr) had an effect on molecule levels in the lymph node. In fact, an increase in permeability of cell reservoir capillaries will lead to reduced drug absorption in the lymphatic system. Furthermore, the diffusion coefficients (D, Dfs) had an important influence on the molecule concentrations in fibrotic capsule and subcutaneous tissue. Additionally, the levels in plasma and major organs were significantly impacted by the degradation coefficient (kdeg), similar to that reported by Shah and Betts [50].

Fig. 9

Sensitivity analysis. Sensitivity analysis of the CTLA4Ig, IgG and ovalbumin models. The scalebar represents the dimensionless integrated sensitivity.

Conclusion

In this work, we investigated the localization of PK and biodistribution of model molecules representative of immunomodulatory agents in a bioengineered neovascularized subcutaneous microenvironment within a 3D-printed cell delivery platform. Our results highlight that localization of drugs within a densely vascularized environment with minimal systemic exposure is feasible. This is highly relevant in the context of innovative strategies leveraging local manipulation of immune cells for the prevention, management and treatment of medical conditions. The good agreement between PBPK model and experimental results indicate that local and systemic drug concentration are predictable. Overall, we highlight the potential of the NICHE and the PBPK model to serve as a foundation for the investigation and development of innovative local immunomodulatory technologies.

Funding

This work was supported by the grant 2-SRA-2021-1078-S-B (AG), Vivian L Smith Foundation (AG), the (AG) and in part by the National Institute of Health R01DK132104 (AG). The pharmacokinetic modeling work was in part supported by the grants R01CA253865 (ZW, VC), R01CA226537 (ZW, VC), R01CA222007 (ZW, VC).

Data availability

The data used to support the findings of this study are available from the corresponding author upon request.

Credit author statement

SC: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization. NH: Investigation. JPM: Investigation PD: Validation, Writing – review & editing. ZW: Validation. VC: Validation. CYXC: Writing – review & editing, Supervision. JEN: Writing – review & editing, Supervision. AG: Conceptualization, Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:Alessandro Grattoni reports financial support was provided by JDRF. Alessandro Grattoni reports financial support was provided by Vivian L Smith Foundation. Zhihui Wang reports financial support was provided by National Institute of Health. Vittorio Cristini reports financial support was provided by National Institutes of Health. Alessandro Grattoni has patent Transcutaneously refillable cell confinement platform with local trophic factor delivery licensed to Nanogland, Inc. Simone Capuani has patent Transcutaneously refillable cell confinement platform with local trophic factor delivery licensed to Nanogland, Inc. Jesus Paez-Mayorga has patent Transcutaneously refillable cell confinement platform with local trophic factor delivery licensed to Nanogland, Inc. Corrine Ying Xuan Chua has patent Transcutaneously refillable cell confinement platform with local trophic factor delivery licensed to Nanogland, Inc.