Backpack-mediated anti-inflammatory macrophage cell therapy for the treatment of traumatic brain injury

Abstract

Traumatic brain injury (TBI) is a debilitating disease with no current therapies outside of acute clinical management. While acute, controlled inflammation is important for debris clearance and regeneration after injury, chronic, rampant inflammation plays a significant adverse role in the pathophysiology of secondary brain injury. Immune cell therapies hold unique therapeutic potential for inflammation modulation, due to their active sensing and migration abilities. Macrophages are particularly suited for this task, given the role of macrophages and microglia in the dysregulated inflammatory response after TBI. However, maintaining adoptively transferred macrophages in an anti-inflammatory, wound-healing phenotype against the proinflammatory TBI milieu is essential. To achieve this, we developed discoidal microparticles, termed backpacks, encapsulating anti-inflammatory interleukin-4, and dexamethasone for ex vivo macrophage attachment. Backpacks durably adhered to the surface of macrophages without internalization and maintained an anti-inflammatory phenotype of the carrier macrophage through 7 days in vitro. Backpack–macrophage therapy was scaled up and safely infused into piglets in a cortical impact TBI model. Backpack–macrophages migrated to the brain lesion site and reduced proinflammatory activation of microglia in the lesion penumbra of the rostral gyrus of the cortex and decreased serum concentrations of proinflammatory biomarkers. These immunomodulatory effects elicited a 56% decrease in lesion volume. The results reported here demonstrate, to the best of our knowledge, a potential use of a cell therapy intervention for a large animal model of TBI and highlight the potential of macrophage-based therapy. Further investigation is required to elucidate the neuroprotection mechanisms associated with anti-inflammatory macrophage therapy.

Significance Statement

Traumatic brain injury (TBI) is a debilitating disease characterized by rampant neuroinflammation, with no current therapies outside of acute clinical symptom management. Anti-inflammatory macrophages hold potential for TBI neuroprotection, given their ability to chemotactically migrate into neuroinflammatory lesion sites. However, brain-infiltrating macrophages are prone to adopting a proinflammatory phenotype in the lesion microenvironment. To counteract this, we developed interleukin-4/dexamethasone-releasing discoidal microparticles, termed backpacks, which durably adhere to the macrophage surfaces. In a piglet TBI model, backpack–macrophages migrated to the brain contusion, reduced proinflammatory microglia, decreased proinflammatory serum biomarkers, and reduced lesion volume by 56%. These results demonstrate a potential use of a cell therapy intervention for a large animal TBI model and highlight the potential of macrophage-based immunotherapy.

Introduction

Traumatic brain injury (TBI) afflicts 3 million people in the United States annually, with around 55,000 fatal cases (1). While there are standard guidelines for the acute clinical management of TBI that focus on reducing intracranial pressure or preventing coagulopathies (2), there are no clinically approved therapeutics for the treatment of TBI that address all phases to mitigate pathology progression (2). One key mediator in disease progression is neuroinflammation, which plays both positive and negative roles after brain injury (3). Neuroinflammation has been shown to promote neurogenesis after injury by directing migration of neural stem cells via chemokine gradients (4) and initiating proliferation of neuronal progenitors (5). Neuroinflammation also plays an important role in axonal regeneration and remyelination after injury (4). In the case of focal TBI, the primary brain injury results in the immediate death of neurons and other cells in the contusion, leading to the extracellular release of proinflammatory damage-associated molecular patterns. While a controlled, acute inflammatory response after TBI is desired to promote debris clearance and regeneration, excessive inflammation contributes to secondary brain injury. Inflammatory activation of microglia, the brain-resident macrophages, and astrocytes leads to further production and secretion of proinflammatory cytokines, along with oxidative stress, local hypoxia, and excitotoxicity (6–10). These processes contribute to blood–brain barrier impairment and production of chemokines, resulting in the infiltration of peripheral myeloid cells, such as monocytes and neutrophils, from circulation into the injured brain tissue, providing additional sources of inflammation. Infiltrating monocytes can differentiate into macrophages and can remain in the contusion site for weeks after the primary injury, contributing to ongoing, chronic neuroinflammation (11). Rampant inflammation after TBI contributes to the expansion of the lesion and increases the risk of downstream development of posttraumatic epilepsy, deficits in sensorimotor and memory function, depression, and dementia (1, 12). Thus, managing the inflammatory cascade in the brain is vital for ameliorating TBI sequelae.

While the benefits of anti-inflammatory therapeutic strategies for TBI are evident in principle, such strategies should specifically target damaged brain regions, rather than exert global immunosuppression to prevent inadvertent side effects. Cell therapies may offer targeted therapy due to the intrinsic ability of cells to chemotactically migrate to injured tissues. However, immune cell therapy strategies for TBI have been underexplored. Although clinical studies have used stem cells for neuro-regeneration (13), to our knowledge, there are no studies utilizing immune cell therapies in large animal models of TBI (14). Recently, Williams et al. (15, 16) published papers utilizing human mesenchymal stem cell (MSC)-derived exosomes, but not the cells themselves, for treating TBI in swine models. Clinical trials utilizing stem cell therapies for the acute treatment of TBI have been completed or are ongoing (17). These clinical trials use bone-marrow-derived autologous stem cells (NCT02795052), modified SB623 stem cells (NCT02416492), umbilical cord-derived stem cells (NCT05018832), and adipose-derived MSCs (NCT04744051) or bone-marrow mononuclear cells (NCT01575470, NCT02525432, NCT02028104, and NCT00254722). Several other clinical trials explore the use of stem cells for delayed treatment of TBI 6 months or more after the primary injury (NCT05951777, NCT04063215, NCT01251003, and NCT05293873). All of these clinical trials focus on the therapeutic potential of stem cells, rather than immune cells. Some preclinical studies in rodents have used regulatory T cells (18), B cells (19, 20), and bone-marrow mononuclear cells as adoptive cell therapies for TBI (17). Overall, the use of immune cells, and more specifically macrophages, is a new frontier for TBI treatment. Given the role of macrophages and microglia in the dysregulated inflammatory response after TBI, their use holds substantial promise for the treatment of TBI. Macrophages perform vital roles of debris clearance in the lesion core and produce growth factors conducive to neurogenesis and angiogenesis (21–23). Accordingly, we hypothesize that delivery of anti-inflammatory macrophages to the lesion microenvironment after injury may confer therapeutic benefits by mitigating damage and promoting tissue repair.

One major barrier to the development of effect TBI therapies is the use of appropriate models. Though rodent studies have significantly contributed to the understanding of TBI, consequential species differences in models of TBI vs. TBI in humans are a major limitation in the field. Over 150 therapies have been shown to reduce lesion volume in rodent models of TBI, but have failed to work in swine or humans (24, 25). There are numerous differences in brain anatomy between rodents and humans, such as location of the hippocampus and abundance of white matter, coupled with differences in injury induction (direct hit in rodents compared to diffuse TBI or secondary cascades in humans). Furthermore, postinjury lesion development deviates between species, manifesting as a cavity in rodents compared to a remodeled area with gliotic scarring in gyrencephalic species. Other host factors that affect the development of post-TBI encephalopathy among species include cellular diversity (26–29), microglia responses (30), extracellular matrix interactions (31), genetic variations, and immune responses (32, 33), confounding our understanding of the development of TBI in humans. Indeed, the immune response to trauma can diverge dramatically in rodents vs. humans (32, 33). Accordingly, in the studies reported here, we tested adoptive macrophage therapy in a biofidelic porcine model to prioritize clinical translatability. Focal cortical contusion of the rostral gyrus, the somatosensory cortex of the snout (34, 35), was chosen as the model to be tested as we have extensively characterized how the impact results in a significant pathological lesion that is clinically silent with pigs recovering quickly after injury (24, 36–43).

Adoptively transferred macrophages can infiltrate into the brain following the chemotactic gradients generated by neuroinflammation and tissue damage (44). However, the excessively proinflammatory TBI microenvironment can force infiltrating macrophages to undergo proinflammatory phenotype switch, which can further exacerbate inflammation. It is critical that adoptively transferred macrophages maintain their anti-inflammatory phenotype in vivo, amid the proinflammatory TBI milieu (45). To prevent proinflammatory repolarization and retain an anti-inflammatory phenotype in vivo, we use discoidal microparticles, termed backpacks (46), which encapsulate a mixture of two anti-inflammatory agents: interleukin-4 (IL-4) and dexamethasone (47). Owing to their discoidal shape, backpacks remain adhered to the macrophage surface without internalization and deliver the anti-inflammatory drug cocktail to the carrier macrophage, thereby maintaining it in an anti-inflammatory phenotype. In a clinically relevant, gyrencephalic porcine cortical impact model of focal TBI, we demonstrate that treatment with anti-inflammatory backpack–macrophages results in a 56% reduction in lesion volume at 7 days postinjury. For studying the expansion or resolution of inflammation, we focused on microglia in the lesion penumbra. The lesion penumbra is the region of interest that may be capable of salvation with therapeutic intervention to prevent the expansion of secondary brain injury (48, 49), as the lesion core that has already undergone neuronal death cannot be rescued by therapeutics. We demonstrate potential remodeling of the lesion penumbra toward a resolved inflammation state. Overall, backpacks are an effective strategy for extending the anti-inflammatory phenotype of adoptively transferred macrophages, resulting in enhanced therapeutic efficacy against TBI neuroinflammation. The results reported here constitute, to the best of our knowledge, a potential use of a cell therapy intervention in a porcine model of TBI and demonstrate the potential safety and efficacy of anti-inflammatory backpack–macrophage approach, serving as a potential step in evaluating this approach before translation to the clinic. Future studies will focus on elucidating the mechanisms of how anti-inflammatory macrophage intervention might affect blood–brain barrier integrity to prevent hemorrhagic transformation.

Results

Backpacks adhere to porcine macrophages

Backpacks were designed to adhere to the surface of macrophages. The high aspect ratio of backpacks is a key design feature that allows them to avoid phagocytosis and remain on the cell surface (50). Backpacks possessed three layers: the two outer layers comprised of dexamethasone in poly(lactic-co-glycolic acid) (PLGA) and the middle layer comprised of IL-4 in poly(vinyl alcohol) (PVA; Fig. 1A). The combination of dexamethasone and IL-4 has been shown to induce a synergistic effect in the induction of anti-inflammatory myeloid cell phenotype (47). Backpacks were prepared through serial spin coating steps, as described previously (47). Backpacks possessed an average diameter of 8.2 µm, thickness of 914 nm, and stiffness of 7.6 GPa (Fig. S1). The loading and release of IL-4 and dexamethasone from backpacks was quantified previously (47).

Fig. 1.

Backpacks durably adhere to porcine macrophages. A) The schematic of the concept shows that backpacks loaded with IL-4 and dexamethasone are adhered to porcine macrophages. Upon intravenous infusion, backpack–macrophages respond to inflammatory chemotactic cues to migrate to and traverse across the disrupted blood–brain barrier. Backpack–macrophages extravasate into the inflamed brain lesion to remodel the inflammatory milieu. B) Percentage of porcine macrophages with ≥1 backpack, as determined by flow cytometry (mean ± SD, n = 2–3). C) Percentage of human macrophages with ≥1 backpack as determined by flow cytometry (mean ± SD, n = 4). D) Confocal micrograph of porcine macrophage (actin, nucleus) with backpack (BP). Scale bar = 5 µm. E) Percentage of macrophages with backpacks attached following shear studies as determined by flow cytometry (mean ± SD, n = 3). For B and E, data were analyzed by 1-way ANOVA with Tukey's honestly significant difference (HSD) test. BP, backpack; ns, not significant. **P < 0.01, ***P < 0.001.

Porcine macrophages were cultured and differentiated from bone-marrow and prestimulated with IL-4 (Fig. S2). Backpack–macrophage complexes were prepared by incubating backpacks with plate-adhered macrophages for backpack attachment to the macrophage surface. Flow cytometry confirmed that backpacks adhered reproducibly to porcine macrophages, with at least one backpack attaching to 33.9% of macrophages at a backpack:macrophage incubation ratio of 3:1 (Fig. 1B). Adhesion of backpacks to human macrophages differentiated from blood-derived human monocytes was also confirmed, with 35.2% of human macrophages attached to at least one backpack, as quantified by flow cytometry (Fig. 1C). Confocal microscopy confirmed that backpacks remained adhered to the macrophage surface without internalization (Fig. 1D). Backpacks remained adhered to macrophages when exposed to a physiologically relevant range of shear stress conditions: 0, 2, 6, and 12 Pa (Fig. 1E). Physiological shear stress ranges from 0.1 to 7 Pa in the vasculature (51, 52).

Backpacks induce durable phenotype shifts in porcine macrophages

Backpacks induced a durable shift in polarization in vitro as assessed by the expression of key proinflammatory (iNOS, CD80) and anti-inflammatory (Arg1, CD206) markers. The expression of each biomarker was normalized to that of macrophages alone. In both unstimulated (Fig. 2A) and inflammatory media (Fig. 2B), backpacks decreased the expression of proinflammatory biomarkers iNOS and CD80 and increased the expression of anti-inflammatory biomarkers Arg1 and CD206 to a level comparable to or greater than that induced by an equivalent dose of a free drug bolus dose over 7 days. Furthermore, backpacks did not adversely impact the viability of carrier macrophages (Fig. 2C). Backpacks did not induce aggregation of macrophages as confirmed by the analysis of singlets via flow cytometry and confocal microscopy (Fig. 2D). Backpacks undergoing one freeze-thaw cycle also adhered to macrophages with a minimal loss in adhesion efficiency compared to that of freshly prepared backpacks (Fig. 2E).

Fig. 2.

Backpacks promote a robust anti-inflammatory phenotype in porcine macrophages. Macrophages (MΦ), backpack–macrophages (BP-MΦ), or macrophages with an equivalent dose of soluble IL-4/dexamethasone (MΦ + sol) were cultured for up to 7 days in A) unstimulated media or B) inflammatory media supplemented with 1 ng/mL interferon-γ (IFN-γ). Cells were analyzed for expression of proinflammatory (iNOS, CD80) and anti-inflammatory (Arg1, CD206) markers. The expression of each marker was normalized to the expression by macrophages alone. Complete data for A is shown in Fig. S3. Complete data for B is shown in Fig. S4. C) Percentage of live cells at 24 h for macrophages (MΦ), macrophages with soluble drug (MΦ + sol.), and backpack–macrophages (BP-MΦ), as determined by flow cytometry (mean ± SD, n = 4). D) Left: percent of singlets of cells as a function of backpack:macrophage ratio, as determined by flow cytometry (mean ± SD, n = 2–3). Right: confocal micrograph of porcine macrophages (actin, nucleus) with backpacks (BP). White arrows denote backpack–macrophage complexes. Scale bar = 20 µm). E) Fold change in backpack adhesion, as quantified by flow cytometry, for freeze-thaw backpacks compared to freshly printed backpacks (mean ± SD, n = 3). For A and B, data were analyzed by two-way ANOVA with Sidak's multiple comparisons test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). For C and D, data were analyzed by 1-way ANOVA with Tukey's HSD test. BP, backpack; ns, not significant.

Backpack–macrophage scale-up for in vivo porcine cortical impact studies

A workflow for high-throughput scale-up of backpack–macrophage production was established to enable studies in porcine cortical impact studies. We reproducibly printed 2.4 × 10⁶ backpacks per template (Fig. S5). Backpacks were adhered to porcine macrophages cultured in 100 mm tissue culture-treated dishes for large-scale in vivo studies, compared to 12-well plates for in vitro testing. With this scaled up method, 17.75% of macrophages had ≥1 backpack on average (Table S1). On average, 141 million macrophages were adoptively transferred per treated piglet (84–262 million, Table S1).

Backpacks accumulate at the site of the contusion in an outbred, gyrencephalic porcine model of TBI

In pigs subjected to cortical impact, backpack–macrophages were administered intravenously and 7 days later, backpacks were observed in the lesion penumbra in treated subjects (Fig. 3A–C). For treated subjects, backpacks penetrated intact tissue of the lesion penumbra. Backpack density was greater in the lesion penumbra (∼80 backpacks per 20× field) compared to analogous regions in the contralateral hemisphere (∼25 backpacks per 20× field) of treated piglets (Fig. 3D). A whole brain scan with magnifications is provided in Fig. S6 to demonstrate the regions of interest provided in Fig. 3A–C. A representative image of the contralateral hemisphere is shown in Fig. S7. Analysis of the background signal at 568 nm in saline pigs is shown in Fig. S8. Among other organs, backpacks exhibited higher density in the lung compared to the kidney and liver (Fig. 3E).

Fig. 3.

Macrophage–backpacks migrate to the lesion penumbra and accumulate in the lung and spleen. Cortical impact was induced in Yorkshire pigs. The pigs were administered backpack–macrophage (with backpacks labeled with rhodamine-B) or saline via jugular vein catheter. On day 7, the brain was harvested for ex vivo imaging. Rhodamine-B (568 nm) signal was visualized in the lesion penumbra of the rostral gyrus of the cortex in A) saline (n = 11) and B) treated piglets (n = 8). C) Backpacks were covisualized with cell nuclei labeled with 4',6-diamidino-2-phenylindole (DAPI). Scale bar = 50 µm. D) Backpack density in the lesion penumbra of the rostral gyrus in the ipsilateral vs. contralateral hemisphere (mean ± SD) (five fields per region per piglet). E) Backpack density in organs in treated subjects (n = 5 treated, 1–2 fields of view per organ per pig). For D, data were analyzed by Welch's t test (*P < 0.1). For E, data were analyzed by one-way ANOVA with Tukey's HSD test (*P < 0.1). Given the variability intrinsic to large animals, outbred species, and the porcine cortical impact model, *P < 0.1 was considered significant for in vivo assessments (53).

Backpack–macrophages show promise in reducing lesion size

The lesion resulting from cortical impact was assessed via macroscopic and microscopic analysis. The macroscopic lesion volume measures lesion volume similarly to lesion volume estimates via T₂ MRI (Fig. 4A and B). Only anti-inflammatory backpack–macrophage treatment with a single administration window was chosen in this study to increase the ability to detect potential treatment differences in this heterogeneous model. Treatment with backpack–macrophages reduced total macroscopic lesion volume by 56% compared to that of subjects receiving saline (196 vs. 86 mm³, Fig. 4C). As the difference in hemorrhage between the groups was visually striking (Figs. 4A and S9), the volume of hemorrhage alone was also estimated macroscopically (Fig. 4D). Treatment with backpack–macrophages decreased hemorrhage volume (73 vs. 21 mm³, Fig. 4D). Hemorrhage volume was positively correlated to lesion volume among all subjects (Pearson correlation, r² = 0.86, P < 0.0001), indicating that lesion volume was heavily impacted by how much the lesion hemorrhaged as the lesion evolved.

Fig. 4.

Backpack–macrophages show promise in reducing lesion size in an outbred, gyrencephalic model of TBI. One-month-old Yorkshire pigs received a cortical impact to the rostral gyrus. Subjects were treated with saline or backpack–macrophages within 4 h postimpact (n = 11 saline, n = 8 treated) via jugular vein catheter. A) Whole brains 7 days after cortical impact in subjects that received saline (left column) or backpack–macrophage treatment (right column). B) Coronal sections demonstrate the lesion 7 days after cortical impact in those that received saline (left column) or backpack–macrophage treatment (right column). On each coronal slab for the entire brain, areas of hemorrhage (top, black), swelling (middle, yellow), or hemorrhage alone (bottom, red) were outlined, multiplied by the slice thickness (5 mm), and added together to determine C) macroscopic lesion volume, and D) hemorrhage volume (mean ± SD). E) The areas of lesioned tissue viewed microscopically were outlined (top, purple dots), filled in to determine area (bottom, yellow), multiplied by the slice thickness (5 mm), and added together to determine, F) microscopic lesion volume (mean ± SD). G) Percent of subjects with microscopic lesion volume >20 mm³ (2 × 2 contingency table with number of lesions < or >20 mm³ among all subjects, X² = 4.23, P = 0.039). The lesion volume of 20 mm³ was determined based on blinded assessment of the distribution of lesion volumes. For C, D, and F, comparisons were made with a one-sided, unpaired, and Student's t test (*P < 0.1).

The microscopic lesion volume is an assessment of permanent tissue damage (Figs. 4E and S10). Compared to saline administration, backpack–macrophage treatment reduced lesion volume by 47% (94 vs. 50 mm³; ns, P = 0.1571) when assessed microscopically (Fig. 4F). Subjects receiving backpack–macrophages had fewer large lesions (lesion volume >20 mm³, Fig. 4G, treated = 2 out of 8, saline = 8 out of 11). Quantifying raw lesion area vs. lesion area as a ratio to contralateral hemisphere area, a helpful metric when evaluating treatments among different developmental stages, resulted in a similar trend (Fig. S11). The two indentor devices used to induce cortical impact were alternated between subjects in each group and exhibited no differences between indentor and lesion volume (Fig. S12).

Backpack–macrophages reduce inflammation locally and systemically

Microglia in the lesion penumbra of the rostral gyrus were assessed for activation status using CD80 expression via immunohistochemistry (IHC) (Figs. 5A and S13). CD80 is a marker of proinflammatory phenotype for antigen presenting cells. The increase in CD80 pixel intensity was lower in the lesion penumbra in treated subjects (6.86%) compared to that of the saline group (9.03%), demonstrating reduced proinflammatory activation (Fig. 5A). Iba1 IHC was used to quantify microglia density (although Iba1 can mark both microglia and infiltrating peripheral macrophages) (Figs. 5B and S14). Microglia density in the lesion penumbra was not significantly lower in subjects treated with backpack–macrophages (483.9 Iba1+/mm²) compared to that of saline subjects (541.3 Iba1+/mm²) (Fig. 5B). Microglia density at the lesion penumbra was greater than microglia density at the comparable location on the rostral gyrus in the contralateral hemisphere for both treated and saline groups, as expected (Fig. 5B).

Fig. 5.

Backpack–macrophages reduce inflammation locally and systemically after TBI. One-month-old Yorkshire pigs received a cortical impact to the rostral gyrus. Subjects were treated with saline or backpack–macrophages within 4 h postimpact via jugular vein catheter. Twenty-four hours and 7 days after injury, blood was drawn. Seven days after injury, brains, and CSF were harvested. A) Top: Tissue sections of lesion penumbra of the rostral gyrus of the cortex were analyzed to determine CD80 pixel intensity increase compared to background (saline n = 6 pigs with n = 2,484 CD80 cells, treated n = 4 pigs with n = 1,664 CD80 cells). Bottom: brightfield 20× microphotographs depict representative images of CD80 for the ipsilateral hemisphere of saline and treated groups (scale bar = 100 µm). B) Bottom: Tissue sections of lesion penumbra of the rostral gyrus were analyzed to determine Iba1+ microglia count per field (saline n = 7 pigs × 5 fields = 35, treated n = 6 pigs × 5 fields = 30). Bottom: brightfield 20× microphotographs depict representative images of Iba1 expression for the ipsilateral hemisphere of saline and treated groups (scale bar = 100 µm). C) TNF-α and GFAP concentrations were assessed in serum at 24 h and 7 days postinjury. D) TNF-α and GFAP concentrations were assessed in CSF at 7 days postinjury. Box and whisker plots reported as median with whiskers from minimum to maximum. For C and D, postinjury analyte concentrations are presented as percentages normalized to the baseline analyte concentration of each subject at −1 h preinjury. Data in A and B were analyzed by one-way ANOVA with Tukey's HSD test (****P < 0.0001). Data in C and D were analyzed by Welch's t test (*P < 0.1).

Peripheral biomarkers of inflammation, tumor necrosis factor alpha (TNF-α), and glial fibrillary acidic protein (GFAP) were analyzed in the serum and cerebrospinal fluid (CSF). Twenty-four hours after injury, serum TNF-α concentration compared to baseline was less in backpack–macrophage-treated subjects (82.7%) vs. that of subjects receiving saline (117.5%) (Fig. 5C). Seven days after injury, serum GFAP concentration compared to baseline was less in treated (75.2%) vs. saline (158.4%) groups (Fig. 5C). There were no significant differences in TNF-α and GFAP concentrations in the CSF between the groups 7 days after injury (Fig. 5D).

Backpack–macrophages are safe

Clinically, the rate of adverse events for treated piglets did not differ from that of piglets receiving saline (Table S2). No signs of toxicity due to backpack–macrophage treatment were observed in the spleen, liver, kidney, and lungs (Fig. S15). The rate of specific pathologic findings was not different between the groups and within normal range of farm-derived piglets (Table S3). Issues with scours occurred presurgically/pretreatment in both groups and were resolved with the administration of probiotics. Both groups had a similar rate of brief apnea at infusion of backpack–macrophages or saline, as well as an equivalent rate of postsurgical swelling at the catheter incision site at the neck, which resolved with cold compression. No piglet had any clinically concerning symptoms in the days following infusion of backpack–macrophages.

Discussion

Despite their potential benefits, cell therapies have not been extensively explored for treating TBI, with the exception of stem cells (18, 54–56). A combination of cell therapies with engineered materials such as backpacks can further modulate and enhance cell functions (57, 58). Owing to their discoidal geometry, backpacks exhibit a unique ability to adhere to the macrophage surface without internalization and control macrophage phenotype with controlled release of loaded cargo (50). Backpacks loaded with interferon-γ (IFN-γ) have been shown to retain an antitumoral macrophage phenotype in the anti-inflammatory tumor microenvironment, resulting in a reduction in tumor volume and improvement in survival in mice with 4T1 mammary carcinomas (46). Additionally, backpacks loaded with IL-4 and dexamethasone have been shown to adhere to monocytes and maintain them in an anti-inflammatory state to treat multiple sclerosis in a mouse model of experimental autoimmune encephalomyelitis (47). Here, we demonstrate an anti-inflammatory backpack–macrophage therapy in a clinically relevant porcine model of TBI. Backpacks maintain the anti-inflammatory phenotype of adoptively transferred macrophages against the proinflammatory TBI microenvironment, resulting in therapeutic efficacy, and offering promise for eventual clinical translation.

We show that IL-4 and dexamethasone-loaded backpacks can efficiently and durably adhere to primary bone-marrow-derived porcine macrophages. Previously, backpack attachment to monocytes required backpack functionalization with anti-CD45 for effective binding (47). In contrast, backpack attachment to macrophages did not require an adhesive coating due to the enhanced phagocytic nature of macrophages, simplifying the backpack design. Backpacks effectively maintain porcine macrophage anti-inflammatory phenotype for up to 7 days in both unstimulated and inflammatory media conditions. After undergoing a freeze-thaw cycle, backpacks retain adhesive ability to macrophages, making the backpack technology amenable to longer term storage. Backpack–macrophages can be produced at scale, as evidenced by our preparation of over 200 million backpacks and over 100 million macrophages per treated piglet for this study. Considering the translational potential of this cell therapy approach, the high-throughput scale-up processes reported here are readily translatable to the magnitude of cells and materials necessary for human clinical trials.

Despite significant efforts and preclinical data, no therapies have yielded success in clinical trials of TBI (25). Attempts at global immunosuppression often result in worse outcomes and higher death rates from comorbid infections (59, 60). While TBI research in rodents has advanced the field's understanding of the disease, rodents are anatomically different from humans with very different responses to trauma (33). The rodent cortex is lissencephalic and contains 15% white matter, while the hippocampus is located directly underneath the thin cortex, resulting in the rapid development of a lesion cavity after TBI. Meanwhile, the human cortex is gyrencephalic, contains 50% white matter, and the hippocampus is embedded deep within the temporal lobe. TBI in humans results in malignant swelling and glial scarring with extensive remodeling. Furthermore, TBI-induced gene expression profiles in humans overlap with only 13% of the transcriptome in rodent models of TBI. Additionally, humans have populations of glia, including subpopulations of microglia, which are distinct from those of rodents (61, 62). The porcine cortical impact model of TBI used here addresses these weaknesses of rodent models with relevant brain anatomy, TBI pathophysiology, microglia populations, and immune responses that more closely resemble that of humans (62, 63).

In the study reported here, all piglets received backpack–macrophages within 4 h postinjury. Administration of backpack–macrophages in the acute window resulted in accumulation of backpacks in the lesion penumbra at 7 days after injury. The striking visual difference in the brains of treated vs. saline subjects was due to the lack of hemorrhage or reduced hemorrhage (Figs. 4A and S8). In other inflammation-mediated brain injuries such as stroke, injury might not result in extensive breakdown of the blood–brain barrier and blood vessels, inducing a lesion without hemorrhage. However, with greater inflammation, the integrity of the blood vessels is compromised. As a result, the ischemic lesion undergoes hemorrhagic transformation, where peripheral blood extravasates outside of the blood–brain barrier into the brain inducing further expansion of the resulting lesion, swelling, and worse outcomes, including increased morbidity (64). We conjecture that backpack–macrophages may have reduced the proportion of subjects that had large lesions and reduced lesion size by reducing inflammation-mediated conversion to hemorrhage. Furthermore, backpack–macrophages may have reduced inflammation-mediated cytotoxic and/or vasogenic edema as there was markedly less swelling in treated subjects. Future studies will seek to validate this hypothesis. Reduction in lesion volume was more striking with macroscopic analysis, which includes swelling, as opposed to that of microscopic analysis (Fig. 4A and B). A reduction of hemorrhagic transformation and swelling could be important to stopping evolving inflammation at the site of the contusion, with the potential to reduce the incidence of morbidity. Backpack–macrophage administration did not result in any clinical complications or organ toxicity different from what was observed in saline piglets. Though backpacks were observed in the lung and spleen in treated piglets, organ pathology analysis did not indicate any specific pathologic findings worse than that of saline piglets.

In response to inflammation after brain injury, microglia undergoes proliferation (65–67). CD80 is a proinflammatory marker of both microglia and peripheral macrophages (68). CD80 is also expressed on activated B cells, T cells, and dendritic cells (69), but their acute presence after TBI is minimal (70). While microglia density was similar in the lesion penumbra for piglets receiving backpack–macrophages or saline, microglia of saline piglets exhibited an increase in CD80 expression, corresponding to more proinflammatory activation and damage that may be expanding into the neighboring regions (66). For backpack–macrophage-treated piglets, the presence of less inflamed microglia in the lesion penumbra suggests that inflammation is resolving compared to that of the saline group.

In addition to studying microglia in the lesion penumbra to gain insight into neuroinflammatory status, peripheral serum, and CSF protein concentrations are clinically relevant biomarkers of neuroinflammation. TNF-α and GFAP are being increasingly used as biomarkers of lesion volume and to prognosticate outcome. TNF-α is one of the most potent proinflammatory cytokines and plays a vital role in exacerbating inflammation, oxidative stress, and excitotoxicity (71). GFAP is a cytoskeletal marker of astrocytes and increases expression due to astrocyte proliferation in response to inflammation (65). In the hours and days after TBI, TNF-α and GFAP originating from the brain diffuse into the CSF and leak out of the impaired blood–brain barrier into the peripheral blood, elevating serum concentrations (72, 73). Concentrations of both TNF-α and GFAP in CSF and serum correlate to extent of injury after TBI and in other neurological diseases in rodent, porcine, and clinical studies (72, 73). Reduced TNF-α serum concentration at 24 h postinjury for backpack–macrophage-treated piglets is indicative of an ameliorated acute inflammatory response. Reduced GFAP serum concentration at 7 days postinjury for treated piglets signifies that ongoing astrocyte activation for formation of a glial scar was reduced, potentially indicating that TBI damage was reduced to an extent that less scarring was required (72). In combination with CD80 and Iba1 analysis, TNF-α and GFAP levels further support evidence of reduced local inflammation following backpack–macrophage treatment.

While the findings reported here offer a snapshot of lesion size and serum inflammation markers at 7 days after focal TBI, ongoing rampant neuroinflammation can lead to an expansion in the lesion size beyond 7 days. Considering chronic microglia activation is thought to contribute to ongoing cognitive issues after TBI, the dampening of microglial inflammation in the lesion penumbra demonstrated in these studies suggests that the therapeutic efficacy of backpack–macrophages may be even more pronounced at later timepoints with a greater relative reduction in lesion size. Furthermore, our work does not assess other potential long-term positive impacts of backpack–macrophage interventions, as reduced acute TBI damage correlates to reduced elevated risk of long-term sequelae such as posttraumatic epilepsy, mental health disorders, and dementia (74). Backpack–macrophages might also be therapeutic in other types of TBI and neurological diseases. For example, rotational TBI may also benefit from backpack–macrophage therapy considering the dysregulation of microglia for up to 1 year after TBI in a swine model of mild rotational injury (75). In future studies, repeat dosing of macrophage–backpacks might further increase the number of backpacks at the contusion site and might elicit an even more robust anti-inflammatory effect, further reducing lesion size. Areas of future exploration include testing additional control groups, varying therapy dose and administration time, varying the timing of biological readouts postinjury, evaluating the percent injected dose of backpack–macrophages at the site of contusion, evaluating sex differences (76), and elucidating the mechanism behind backpack–macrophage inhibition of hemorrhagic transformation.

Overall, TBI is a debilitating disease resulting in an array of post-TBI sequelae with no current treatments besides clinical management of malignant swelling. Cell therapies provide a unique opportunity to leverage chemotactically guided migration to overcome barriers to brain delivery and achieve accumulation at the target site in the brain. We demonstrate that IL-4- and dexamethasone-loaded backpacks can be adhered to porcine macrophages and promote an anti-inflammatory macrophage phenotype for at least 7 days in vitro. We produced backpack–macrophages at high-throughput scale and safely infused them into piglets after cortical impact. Backpack–macrophages trafficked to the lesion site in the brain and remodeled the local and systemic inflammatory milieu, evidenced by decreased microglia proinflammatory expression in lesion penumbra of the rostral gyrus of the cortex and decreased serum concentrations of proinflammatory biomarkers. These immunomodulatory effects contributed to a 56% decrease in lesion volume. The results reported here demonstrate, to the best of our knowledge, a potential use of a cell therapy intervention for a large animal model of TBI.

Materials and methods

Materials

PLGA Resomer 502H, dexamethasone, PVA, heparin, Roswell Park Memorial Institute (RPMI) 1640 media, dipotassium ethylenediaminetetraacetic acid (K₂EDTA), fetal bovine serum (FBS), penicillin and streptomycin (P/S), phosphate-buffered saline (PBS), and LIVE/DEAD blue dye were purchased from Sigma-Aldrich. PLGA-rhodamine-B was obtained from PolySciTech Akina. Recombinant murine macrophage colony-stimulating factor (M-CSF) was obtained from PeproTech. Recombinant porcine IL-4 was obtained from Thermo Fisher. Porcine TNFa and GFAP enzyme-linked immunosorbent assay (ELISA) kits were purchased from RayBioTech and MyBioSource, respectively. Sylgard 184 Silicone Elastomer kit was purchased from Dow. All primary antibodies for immune cell staining were purchased from Invitrogen, Novus Biologicals, and R&D Systems. All secondary antibodies were purchased from Invitrogen. Cell staining buffer was purchased from BioLegend. Cell fixation/permeabilization kits were obtained from BD Biosciences.

Animals

Male, Yorkshire piglets aged 30 days were used (n = 24; 8–11 kg; Parsons Farm, Hadley, MA, USA). All protocols used were approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee and the Animal Care and Use Review Office of the United States Army Medical Research and Development Command and adhere to the guidelines of the NIH Guide for the Care and Use of Laboratory Animals. Every effort was made to reduce animal numbers, animal discomfort, and suffering.

Backpack fabrication

Polydimethylsiloxane (PDMS) templates were prepared as described previously (46, 77). Briefly, silicon wafers were fabricated with patterned photoresist in an array of 8 µm holes. PDMS mixed in a 10:1 base-to-crosslinker ratio from the Sylgard 184 kit was poured onto silicon wafers in petri dishes in 20 g aliquots. The PDMS was degassed and cured at 65°C overnight and cut away from the silicon wafers. A solution of 80 mg/mL PLGA Resomer 502H (7 to 17 kDa) and 15 mg/mL dexamethasone in acetone was prepared. For fluorescently labeled backpacks, PLGA-rhodamine-B was incorporated at a ratio of 100:1 fluorescent to nonfluorescent PLGA. Two hundred and twenty microliters of PLGA solution were spin-coated onto each PDMS quadrant at 2,000 × g for 35 s (at a 200 × g/s ramp). Quadrants were plasma-ashed with O₂ for 60 s. A solution of 0.5 wt% PVA (146 to 186 kDa, 99+% hydrolyzed) and 0.5 wt% heparin in PBS was prepared with IL-4 (25 µg/mL). Heparin was used to stabilize IL-4 loading for greater loading and bioactivity (47, 78). Immediately after plasma treatment, 50 µL of PVA/IL-4 solution was spread onto each quadrant. Quadrants were dried in a desiccator for 1 h and then a second PLGA layer was deposited using the same procedure as a potential. Backpacks were then stamped onto PVA-coated dishes by microcontact printing, as described previously (46). To collect backpacks, PVA-coated dishes were washed twice with 3 mL of PBS. The solution was filtered through 20 µm cell strainers and pelleted at 2,000 × g for 5 min. Backpacks were resuspended in media of choice.

Backpack characterization

Backpacks were harvested from dishes and centrifuged at 2,500 × g for 5 min and resuspended in RPMI + 0.1% bovine serum albumin. Atomic force microscopy (AFM, JPK Nanowizard, Bruker) was used to characterize the topology and stiffness of backpacks. Backpacks were adhered to glass slides, mounted on the AFM (JPK Nanowizard, Bruker), and imaged in Qi (single-point contact) mode using all-in-one-Al cantilever D with a stiffness of ∼40 Nm⁻¹. About 10 µm × 10 µm regions were scanned for quantifying backpack topography, followed by a 2 µm × 2 µm scan on backpack surface for probing stiffness. Topography and stiffness were recovered using JPK DP data processing software. Stiffness was obtained by fitting corrected deflection curves to a Hertz model assuming a pyramid tip.

Yorkshire porcine bone-marrow cell extraction

The rib cage was collected from 4- to 6-week-old piglets before perfusion (detailed above) and stored on ice until bone-marrow extraction. The surface of the rib cage was cleaned in 70% ethanol and dried before transfer into a sterile biosafety cabinet. All media and surgical instruments used were sterile. Surgical scissors and tweezers were used to cut and remove the adipose and muscle tissue from the rib cage, yielding individual ribs. Exposed ribs were cleaned in 70% ethanol and dried once more. Approximately, 0.5 cm of the end of the exposed bones and the costal cartilage on the opposite end of the rib were cut and removed. Subsequently, the bone-marrow was flushed out in cold bone-marrow extraction media (BMEM: RPMI 1640 with 5 mM K₂EDTA) with a 21-G syringe needle and collected into a sterile 50 mL falcon tube. Once the ends of the bones were thoroughly flushed, the bone was cut in ∼2 cm increments for subsequent flushing until the bone-marrow from the entire rib was extracted. After bone-marrow cells (BMCs) from all the ribs were extracted, the BMC solution was passed through a 40-μm cell strainer into a new sterile 50 mL falcon tube to remove debris. The filtered BMEM solution was centrifuged at 300 × g for 10 min, aspirated, resuspended in 4°C PBS, and again pelleted. BMCs were resuspended in 5 mL of Ammonium-Chloride-Potassium (ACK) lysing buffer for 2 min at room temperature, brought up to 50 mL with cold PBS, followed by pelleting and washing until the lysed red blood cells were washed away. BMCs were resuspended in 4°C macrophage media (MM−: RPMI 1640 with 1% P/S and 10% FBS) with 10% DMSO at a concentration of 40 million cells/mL and frozen for future studies.

Porcine bone-marrow-derived macrophage culture

Cryovials were thawed in a bead bath until a small ice crystal remained. The cryovials were collected with prewarmed MM− at a volume ratio of 1:3 into 50 mL falcon tubes, followed by pelletting at 300 × g for 10 min. BMCs were resuspended in prewarmed MM+ (MM− with 20 ng/mL murine M-CSF) and counted. BMCs were seeded in tissue culture (TC)-treated 100 mm dishes at ∼10 × 10⁶ BMCs/dish in 12 mL MM+ and placed into an incubator at 37°C and 5% CO₂. To promote cell attachment to the 100 mm dish surface, extreme caution was taken throughout culturing to not disturb the plates. After 6 days of incubation, the media was gently aspirated, followed by a gentle addition of 12 mL of prewarmed MM+ with 20 ng/mL porcine IL-4. On day 9 after seeding, BMCs had matured into bone-marrow-derived macrophages and were ready for subsequent experiments.

Backpack attachment to porcine macrophages for in vitro studies

For in vitro studies, porcine macrophages were replated in nontissue cultured treated 24-well plates with 150,000–175,000 cells suspended in 500 mL of growth media and allowed to adhere for 24 h. Backpacks were washed from dishes and centrifuged at 2,500 × g for 5 min and resuspended in media. Backpacks were counted using a hemocytometer and added in the predetermined ratio (0.75:1 to 3:1 backpacks to macrophages) to each well. Plates were centrifuged at 300 ×  g for 7.5 min and incubated for 1 h at 37°C and 5% CO₂. To harvest the backpack–macrophage complexes, the media from the wells was collected. The wells were washed with cold PBS, and a cold solution of 5 mM K₂EDTA in PBS was added. The plates were placed in a 37°C incubator for 15 min. The plates were then removed and tapped to dislodge adherent cells, and the K₂EDTA solution was collected. Cells were washed once more with cold PBS, pelleted, and resuspended in media or buffer of choice for downstream use. For phenotyping studies, after backpack–macrophage incubation, the growth media was replaced with unstimulated growth media, unstimulated growth media containing an equivalent dose of soluble IL-4/dexamethasone (18.31 ng/mL IL-4 and 3.74 μg/mL dexamethasone). For some conditions, inflammatory media supplemented with 1 ng/mL IFN-γ was used. To determine activation status longitudinally, cells were cultured and harvested at 1, 4, and 7 days. During the harvesting steps, cells were collected as described previously and stained using LIVE/DEAD Blue (BioLegend). Samples were washed and blocked with Anti-Pig CD16 (BioRead) in a solution of either 1% goat serum (samples for surface staining) or 1% donkey/chicken serum (samples for intracellular staining). For surface staining, samples were stained with primary antibodies anti-CD80-SuperBright600 (Invitrogen) and anti-CD206 (Novus Biologicals) and secondary antibodies antirat-AlexaFluor488 (Invitrogen). For intracellular staining, samples were fixed, permeabilized, stained with primary antibodies anti-iNOS (Life Technologies) and anti-Arg1 (Life Technologies), and stained with secondary antibodies antirabbit-AlexaFluor647 (Invitrogen) and antigoat-AlexaFluor488 (Invitrogen). Samples were resuspended in stain buffer and assayed on the Cytek Aurora analyzer. Data were analyzed with FlowJo V10. The expression of each biomarker was normalized to macrophages alone for each respective day.

Backpack attachment to porcine macrophages for in vivo studies

For in vivo studies, porcine macrophages were plated in nontissue culture-treated 100 mm² dishes. Backpacks were harvested, pelleted, and resuspended in media at a concentration of 3–4 million backpacks/mL. One milliliter of media was removed from the macrophage dishes, and 1 mL of backpack solution was added per dish and the dishes were swished to disperse the backpacks. Dishes were then placed in a cell culture incubator for 1–1.5 h to allow backpacks to adhere to macrophages. To harvest the backpack–macrophage complexes, the media was collected as described above using cold PBS and K₂EDTA solution with 5 mL per step per 100 mm² dish. Cells were centrifuged and pellets were resuspended in 10–20 mL saline for in vivo injections.

Cortical impact induction and surgical procedures

Piglets were sedated with an intramuscular injection of Telazol (0.5 mg/kg), xylazine (1–2.2 mg/kg), and atropine (0.04 mg/kg) or midazolam (0.05 mg/kg), xylazine (1–2.2 mg/kg), and atropine (0.04 mg/kg). To address slow recovery from anesthesia, preanesthetic regimen was changed to xylazine (1 mg/kg) and atropine (0.04 mg/kg) only.

Anesthesia was induced by inhaled isoflurane with a snout mask. Surgical sites were clipped and scrubbed. Piglets were draped. Buprenorphine (0.02 mg/kg) was administered at least 15 min before the incision for analgesia. A Bair hugger blanket with forced hot air was used to maintain core body temperature between 37 and 39°C as body temperature influences lesion size after TBI (40). An ear IV was placed and propofol (0.5–2 mg/kg) was used as an adjunct for anesthetic induction. The piglet was intubated and anesthesia was maintained with isoflurane and mechanically ventilated. Oxygen saturation, heart rate, blood pressure, respiration rate (via mechanical ventilation), end-tidal CO₂, and core body temperature as recorded with a nasal temperature probe were monitored continuously and recorded every 5 min. Ventilation was adjusted to maintain end-tidal CO₂ within 35–45 mmHg. Before injury, animals were switched from ventilation with oxygen to ventilation with room air to parallel TBI in humans. Prophylactic cefazolin (20 mg/kg) was administered intravenously. Local anesthetic bupivacaine (1.5–2.5 mg/kg) was administered subcutaneously at the incision sites before incision.

The external jugular vein was catheterized to allow venous access for backpack–macrophages or saline delivery and was tunneled under the skin for an exit near the back of the neck. The incision site was closed. Venous blood (5–7 mL) was collected before injury for later analysis. The catheter was flushed with heparinized saline (1 to 3 cc, 10 units/mL) if not actively infused with saline.

A straight incision was made ∼15 cm long, running down the sagittal midline of skull from above the snout to the crown of the head. The intersection of the right coronal and sagittal suture is exposed, and a craniectomy is performed resulting in a 2 cm window over the rostral gyrus, which is the somatosensory cortex for the snout (35). This cortical impact model is well-characterized resulting in a clinically silent, pathoanatomic contusion (40, 41). The dura was cut in a stellate manner and reflected back. The stand for the indentor was screwed on securely to the skull such that the indentor was perpendicular to the cortical surface. The spring-loaded device was screwed into the stand until contacting the dura. The spring-loaded indentor tip was deployed and the indentor device was removed. No sham animals were used. The skull was not replaced. The cortical surface was gently irrigated with saline. Both incision sites were closed with interrupted subcutaneous suture (2-0 PDS2) followed by a running subcuticular suture (3-0 Monocryl).

Recovery from anesthesia and postcortical impact injury care

Swine was fitted with a vest (SAI Infusion Technologies, Lake Vila, IL, USA) to secure and protect the external jugular catheter, and a fentanyl patch (2–3 μg/kg/h for 72 h) was applied. Piglets were lightly sedated for the 1 h postinjury blood collection then were placed back on 100% oxygen, removed from mechanical ventilation, and encouraged to start breathing on their own and extubated while on the OR table.

Postcortical impact, therapy administration

Piglets were serially assigned to receive treatment or saline. The backpack–macrophages or saline were infused with an IV pump (278 mL/h) in a 10–20 mL volume. In a subset of piglets (n = 1 treated, n = 1 saline), cells with backpacks or saline were infused at 1 h postinjury while still in the operating room. In the remaining piglets (n = 20), backpack–macrophages or saline was administered 4 h postinjury. Initially, piglets (n = 16) were re-anesthetized at this time point with brief exposure to isoflurane mixed with 100% oxygen delivered via snout mask. At the same time, piglets received a second dose of buprenorphine (0.02 mg/kg) to provide analgesia until the fentanyl from the patch was absorbed. However, after some piglets were apneic during the second bout of anesthesia, the last four piglets remained awake for their blood collection and administration of backpack–macrophages or saline. In this case, backpack–macrophages or saline was administered by hand slowly over 1–2 minutes instead of using an IV pump for logistical reasons.

Cortical impact study endpoints

Seven days after cortical impact, piglets were anesthetized with an intramuscular injection of ketamine and xylazine, anesthetized with isoflurane (2%), and intubated. CSF (2–4 mL) was collected from the cisterna magna with an 18-G spinal needle. Piglets were deeply anesthetized (4–5% isoflurane). The chest cavity was opened, and the ribs were removed. In a subset of piglets, ribs were put on ice for harvest of BMCs. Piglets were then perfused through the heart with PBS followed by 10% formalin. The brain was collected as well as samples from the liver, spleen, kidney, and lung. Organs were postfixed in 10% formalin for 3–5 days before being moved to PBS. One treated piglet underwent the above steps 6 days after cortical impact, due to surgical room availability.

Indentor devices

Two indentor devices were machined in November 2021 for this study. Each is a spring-loaded device previously described (39, 40). Deployment characteristics were analyzed in indentor 1 by a biomechanical engineering lab. The uncocked indentor tip was positioned 150 mm away from the top of the laser displacement sensor. The indentor was cocked and deployed 5 times with data sampled at 10,000 Hz with a data acquisition system (Labview Signal Express 2015; National Instruments, Austin, TX, USA). A custom MATLAB script was written to analyze the data. A fourth-order Butterworth low pass filter with a cutoff frequency of 150 Hz was used on each data set. The region of interest was identified, and the impact velocity was calculated using the last half of the displacement data and polyfit to calculate the slope (velocity) of the line. The impact velocity data location was chosen to eliminate the edge effects (getting the mass up to speed at the start; metal on metal and the resulting rebound at the end). The indentor was found to operate within expected limits: velocity at impact of 1.597 m/s, and a time to deploy of 3.5 ms. Indentors were alternated so that a similar number of saline and treated piglets received the same indentor (indentor 1: saline n = 6, treated n = 4; indentor 2: saline n = 5, treated n = 7).

Pig brain histology

Brains were photographed and cut into 5 mm slabs in a standardized manner separating hemispheres ipsilateral and contralateral to the unilateral cortical impact. Brains were paraffin embedded and 5 µm slices were sectioned, mounted, and baked for 30 min at 60°C. The majority of sectioning, hematoxylin and eosin (H&E), and all of the IHC were performed by Comparative Pathology and Genomics Shared Resource at Tufts University Cummings School of Veterinary Medicine. Some sectioning and H&E staining were performed in the Brain Trauma Lab as well as the MGH Pathology Core at Charlestown Navy Yard.

For IHC for Iba1 and CD80, sections were deparaffinized and hydrated. Antigen retrieval was accomplished by heating in a pressure cooker for 25 min in citrate buffer then cooling for 20 min at room temperature. Sections were rinsed twice in distilled water (rinses were in distilled water unless specified), loaded onto IntelliPATH FLX automated IHC staining system (BioCare Medical, Pacheco, CA, USA) and flooded with TBS Auto Wash Buffer (“Buffer”; BioCare Medical) with 3% hydrogen peroxide, 2 changes, 5 min each, rinsed twice with Buffer. Sections were blocked for 10 min (Background Punisher, BioCare Medical). Without rinsing, primary antibodies (Iba1: anti-Iba1 1:2k, Polyclonal Rabbit, Wako, FujuFilm 019-19741; CD80: anti-CD80 1:50, Polyclonal Rabbit, Invitrogen PAS83990/E3585469) or a universal negative were applied and incubated for 60 min at room temperature. Slides were rinsed twice in Buffer and the secondary Horseradish Peroxidase (HRP)-antibody was applied (Mach2 Rabbit HRP-Polymer, BioCareMedical) for 30 min at room temperature. Sections were rinsed twice in Buffer and 3,3'-Diaminobenzidine (DAB) chromagen was applied and incubated at 4 min at room temperature and rinsed. Sections were counterstained with hematoxylin 1:1 for 4 min, rinsed twice, and exposed bluing reagent for 2 min, rinsed twice, removed from IntelliPATH FLX, rinsed again, and dehydrated 2 changes each 95% ethanol, 100% ethanol, xylene, and were coverslipped.

Microscopic evaluation of backpacks

For backpack quantification, sections were deparaffinized, hydrated, and were coverslipped with a soft set mounting medium containing DAPI (VECTASHIELD Antifade, Vector Laboratories). Backpacks were tagged with rhodamine-B, a fluorescent probe that emits at 568 nm.

The number of backpacks was determined in the lesion penumbra or the analogous region on the rostral gyrus in treated or control piglets. Some backpack-like signals were detected in saline piglets via the automated analysis due to red blood cell presence at the contusion site from hemorrhage. Red blood cells exhibit autofluorescence and are similar in size as the backpacks. The area was brought into focus using the DAPI/405 nm filter and then switched to the red/595 nm filter. A set of qualitative photos were taken of both backpacks (595 nm filter) and DAPI (405 nm filter) and merged. The protocol was tested with piglets that did and did not receive macrophage–backpacks to ensure a few false positives. Additionally, backpack density was determined in the kidney, liver, lung, and spleen in 4–5 piglets per group. Five standardized images using Zen Microscopy image acquisition software were obtained with the 20× objective (ZEISS, Germany; 3,200 K white balance, X-cite at 50% for photos in red and 20% for blue/DAPI; exposure time for DAPI photos 400 ms; 494/red 3,800 ms with the contrast adjusted using the “black” at 35). Only structurally intact tissue was evaluated to focus attention on potentially salvageable tissue and avoid red blood cells. Fields photographed avoided hemorrhages or cavitations in the tissue, which were filled with cells and cell debris.

Backpack counting was automated using ImageJ. The threshold was adjusted until only backpacks were highlighted in red, binarized, and then underwent Watershed function. The Erode function was used if the image contained extremely small points. The processed image was saved and then analyzed using the Analyze Particles function. The background signal is displayed in Fig. S6.

Microscopic evaluation of lesion volume

The lesion was marked on one section from 4 to 5 blocks (5 mm thick; Fig. S5A and B) spanning the lesion or the comparable location in sections stained with H&E. Areas that were marked as lesion included pyknotic and vacuolized neurons, hemorrhage, and areas where there was loss of structure of tissue with either the cavitations or the beginnings of a glial scar (disorganized white or gray matter; Fig. S9). White matter was designated as a lesion when there was loss of tightly packed fibers resulting in rarefaction and/or contained hemorrhage. However, white matter that demonstrated mild edema, including swelling of oligodendrocytes, which is often fixation artifact, was not included as it is nonspecific and/or might resolve.

The marked slide was then photographed with a ruler, and the lesion area was determined with the ruler tool, paint can tool, and polygonal lasso tool adding layers in Photoshop. The volume of lesion was calculated by multiplying lesion area on each section by the slice thickness (5 mm) and calculating the sum. Alternatively, the area of lesion as a percent of the contralateral hemisphere was calculated to allow comparison to future studies that might test the effect of age, scaling by brain size. The ratio of the lesion to the contralateral hemisphere was multiplied by 100 and averaged among sections. The lesion size as a percent area was very similar in pattern between groups to lesion volume estimates (Fig. S10).

Macroscopic evaluation of lesion volume

To evaluate differences in swelling and/or edema in addition to microscopic lesions, similar to what might be measured via T₂ MRI imaging, macroscopic lesion volume was calculated via photographs of the coronal 5 mm slabs exhibiting visual abnormalities. The area of the lesion was marked in Photoshop, and volume was calculated as described above (microscopic evaluation of lesion volume). Marked areas included hemorrhage, dusky, or dark areas of cortex, as well as areas of swelling that were asymmetric from the uninjured, contralateral hemisphere. Both lesion volume and lesion area as a percent of the uninjured hemisphere were calculated as described above.

Similarly, areas of hemorrhage only were marked in Photoshop and volume calculated.

Microscopic evaluation of Iba1 and CD80

Sections of brain hemispheres were scanned under brightfield at 20× magnification on the ZEISS Axio Scan-Z1 Slide Scanner to obtain .czi files. Five subset fields were obtained for analysis per slide. Five fields were randomly sampled from the gray matter in the lesion penumbra of the rostral gyrus of the cortex and the comparable region on the contralateral hemisphere and ranged in size from 3.7–6.2 × 10⁶ μm². Fields were analyzed via ImageJ analysis macros developed in-house. The authors were blinded to whether fields corresponded to saline or treated piglets and ipsilateral or contralateral hemispheres when developing the ImageJ analysis macro. The .czi files were imported into ImageJ as RGB images.

For Iba1 analysis after duplicating the blue channel from the original RGB image, the blue channel was binarized via Yen auto-threshold and converted to mask. The mask then was processed with the Analyze Particles function with a 1,300 minimum pixel size, outputting the object count per field representing the number of Iba1+ cells (Fig. S16). The fields were separately processed to obtain the area. Microglia cell density was calculated with the following equation:

$$\mathrm{Microglia}\mathrm{cell}\mathrm{density}={\displaystyle \frac{\mathrm{Microglia}\mathrm{cell}\mathrm{count}}{\mathrm{Field}\mathrm{area}(\mathrm{m}{\mathrm{m}}^2)}}$$

(1)

For CD80 analysis, after duplicating the blue channel from the original RGB image, the blue channel was binarized via Otsu auto-threshold and converted to a mask (Fig. S17). The mask was then inverted and underwent Create Selection to add the cell bodies to the region of interest (ROI) manager. The mask then underwent the Analyze Particles function to create a new mask with 1,300 minimum pixel size and 0.4–1.0 circularity to remove dark signals from debris and nuclei without CD80 positivity. The new mask underwent Create Selection to obtain an ROI of CD80⁺ objects. This ROI was superimposed onto a duplicate of the red channel, underwent the Clear Outside function, was binarized via Isodata auto-threshold, converted to mask, inverted, and underwent Create Selection to obtain CD80⁺ objects without dark nuclei centers. The mask was then inverted and underwent Create Selection for a new ROI. This new ROI was then superimposed onto a duplicate of the blue channel, underwent the Clear Outside function, and was binarized via Isodata auto-threshold and converted to a mask to remove light nuclei centers. The mask was then inverted and underwent Analyze Particles with 200 minimum pixel size to add an ROI of each individual object containing the soma of CD80⁺ cells without nuclei. The IPI of each individual object ROI was then measured on a duplicate of the red channel.

To obtain the mean background pixel intensity per field, after duplicating the red channel from the original RGB image, the red channel was binarized via Otsu auto-threshold, converted to mask, inverted, and processed with Create Selection to add cell bodies ROI (Fig. S18). The cell bodies ROI was then superimposed onto a duplicate of the red channel and underwent the Clear Outside function to remove the cell bodies. A duplicated red channel without cell bodies was then binarized via Yen auto-threshold, converted to mask, and underwent Create Selection to add the void white spaces ROI. This void white space ROI was superimposed onto the red channel without cell bodies and underwent Clear Outside function to yield background only, which was measured for pixel intensity.

IPI and field background pixel intensity (FPI) were subtracted from 255 to obtain an inverted pixel intensity value, where 0 represents no signal, and 255 represents a fully saturated signal. The CD80 pixel intensity increase was calculated with the following equation:

$$\begin{array}{rl}\mathrm{CD}80\mathrm{pixel}\mathrm{intensity}\mathrm{increase}=& {\displaystyle \frac{\mathrm{Inverted}\mathrm{IPI}-\mathrm{Inverted}\mathrm{FPI}}{\mathrm{Inverted}\mathrm{FPI}}}\\ & \times 100\text{\%}\end{array}$$

(2)

Inflammatory marker analysis of serum and CSF

Blood was obtained from an external jugular vein catheter from swine at −1 h preinjury, and 1, 4, 24, and 168 h postinjury when logistically possible. Blood was stored for 15–30 min at room temperature to allow for clotting, followed by centrifugation at 1,000 × g for 10 min at 4°C for collection of the serum supernatant. CSF was obtained via lumbar puncture at 168 h postinjury. Serum and CSF were prepared in 120 μL aliquots and stored at −80°C. Serum and CSF samples were transferred from −80 to 4°C to thaw overnight. The porcine TNF-α ELISA kit (RayBioTech) and porcine GFAP ELISA Kit (MyBioSource) were performed following the manufacturer's instructions, with samples tested in duplicate. The 7 d serum sample from pig 484 and the 7 d CSF sample from pig 483 were rerun for TNF-α detection due to a high (>20%) duplicate coefficient of variation (CV). The overall average CV for TNF-α and GFAP were 4.64 and 3.52%, respectively. The samples from piglets 479 and 480 were excluded from TNF-α analysis due to improper processing of blood/CSF at collection. Samples were diluted 2-fold in assay diluent before running the ELISAs. Porcine TNF-α concentration was measured at 450 nm based on 3,3',5,5' tetramethylbenzidine (TMB) substrate colorimetric detection, and porcine GFAP concentration was measured at 450 nm based on HRP substrate colorimetric detection system on a BioTek Synergy H1 microplate reader. TNF-α and GFAP concentrations were normalized to the −1 h preinjury concentrations of each individual respective piglet to obtain the biomarker baseline percentage with the following equation:

$$\begin{array}{rl}{\mathrm{Baseline }\text{\%}}_{\mathrm{TNF}\text{-}\mathrm{\alpha }or\mathrm{GFAP}}=& {\displaystyle \frac{{\mathrm{Concentration}}_{\mathrm{TNF}\text{-}\mathrm{\alpha }or\mathrm{GFAP}}}{-1\mathrm{h}{\mathrm{concentration}}_{\mathrm{TNF}\text{-}\mathrm{\alpha }or\mathrm{GFAP}}}\times 100\text{\%}}\end{array}$$

(3)

Transparency and reproducibility paragraph

In this first attempt at determining the efficacy of a cell-based therapy to treat TBI, only male piglets were used and both 1 and 4 h delay to treat was chosen. Due to initiation of the coordination and logistics of producing the large volume of engineered cells, piglets were not randomized but treatment was alternated. If there was an issue in the batch of cells where infusion was not attempted, then both piglets scheduled that day were assigned to receive saline that day. Our previous lesion size data after cortical impact in 1-month-old male piglets was used for a power analysis. We calculated that 10 piglets per group would detect a 78% difference in lesion size at an alpha level of 0.10 and a power of 95% (53). All tissue analyses were performed by those blinded to treatment. A total of five piglets were excluded from lesion analyses (Table S4). Two piglets were excluded as they did not finish the experiment. An additional three piglets were excluded due to insufficient macrophage yield. Data generated here will inform future power tests to detect a 50% reduction of lesion volume at an alpha level of 0.05. A total of 19 piglets were in the study: 11 in the saline group, and 8 in the backpack–macrophage group. No individual data points were excluded.

Statistics

For in vitro assessments concerning backpack–macrophage binding and viability, data were tested via one-way ANOVA with Tukey's honestly significant difference (HSD) test (*P < 0.05, **P < 0.01, ***P < 0.005). For in vivo assessments, the main effects of treatment, hemisphere, and the interaction on backpack density at the contusion site were tested via two-way ANOVA followed by Tukey's HSD test. The main effects of treatment, organ, and the interaction on backpack density in the kidney, liver, lung, and spleen were tested via two-way ANOVA followed by Sidak post hoc testing. A reduction from treatment with backpack–macrophages was tested in macroscopic and microscopic lesion volume or lesion area as a ratio of the contralateral hemisphere via two-sided, unpaired Student’s t test. Because of the variability intrinsic to large animal, outbred species, and of the porcine cortical impact model in this study, *P < 0.1 was considered significant for in vivo assessments (53): backpack counts, lesion analysis, and serum/CSF biomarker concentrations. A χ² test was performed to detect differences in complications in saline vs. treated piglets.

Acknowledgments

The authors acknowledge Wyss Institute of Biologically Inspired Engineering and John A. Paulson School of Engineering at Harvard University for their support. The authors acknowledge Kerry Daley, Solomina Darko, and Ani Srinivasan for backpack fabrication. They acknowledge Tammy Hayes, Dian Taylor, and Denise Elwart of Tufts University Cummings School of Veterinary Medicine Comparative Pathology and Genomics Shared Resource. The authors also acknowledge the staff of the MGH Knight Surgical Research Laboratory for anesthesia during surgeries and the MGH Center for Comparative Medicine, including Kimberly Degrenier, Julie Larson, and Charlotte Chuen, for animal husbandry and postsurgical care. The authors thank Dr Jibing Yang for consultation on postsurgical complications. They acknowledge James Luck for help with backpack counts and postsurgical procedures, Shruthi Rozario for lesion analysis and postsurgical procedures, Elizabeth Horan for scheduling piglet surgeries and facilities and equipment and lesion analysis, Robert H. Petrillo for surgery preparation, assistance with piglet surgeries, postsurgical procedures, and photographs for backpack counts analysis, Monica Tynan for piglet surgeries, postsurgical procedures, David Kim for postsurgical procedures, Praneel Sunkvalli for postsurgical procedures, Benjamin Baskin for postsurgical procedures, and Chloe Hyun for lesion analysis. The authors acknowledge Dr C. Wyatt Shields, Dr Ann-Christine Duhaime, Dr Michael Whalen, Amogh Vaidya, and Dr Abhirup Mandal for their helpful discussions. They acknowledge the Harvard Center for Biological Imaging; the Allston Science and Engineering Complex's Molecular and Cellular Biology Core; and Harvard University Center for Nanoscale Systems (CNS), a member of the National Nanotechnology Coordinated Infrastructure Network (NNCI) supported by the National Science Foundation ECCS-2025158. The authors also thank the Dana-Farber/Harvard Cancer Center in Boston, MA, USA, for the use of the Rodent Histopathology Core and its histological section preparation service. They acknowledge the use of https://www.biorender.com in creating schematics.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

The authors acknowledge support from the Defense Medical Research and Development Program by the Department of Defense (W81XWH-19-2-0011). N.Ka. acknowledges support from the National Science Foundation Graduate Research Fellowship under grant no. 1122374. B.C.-B. acknowledges support from National Institute of Health Eunice Kennedy Shriver National Institute of Child Health and Human Development R01 HD099397. M.E.J. acknowledges support from the National Science Foundation Graduate Research Fellowship under grant no. 1745302.

Author Contributions

N.Ka. and R.L. led and contributed equally to the work, performing backpack fabrication, macrophage–backpack preparation, data analysis, manuscript concept ideation, manuscript writing, and manuscript editing. B.C.-B. led the in vivo work, performing piglet surgeries and postsurgical procedures, backpack counts, lesion analysis, manuscript writing, and manuscript editing. L.S. performed backpack counts. T.S. performed piglet surgeries, postsurgical procedures, and lesion analysis. S.P. performed backpack fabrication and macrophage–backpack preparation and assisted manuscript concept ideation. N.Ku. performed silicon wafer photolithography, backpack fabrication, and macrophage–backpack preparation, and assisted manuscript concept ideation. V.C.S. performed backpack fabrication, macrophage–backpack preparation, and AFM. L.L.W. performed silicon wafer photolithography and assisted manuscript concept ideation. M.F., R.R., and D.V. performed backpack fabrication. S.S., M.J., K.S.P, M.D., and K.A. performed backpack fabrication and macrophage–backpack preparation. B.G. performed piglet surgeries and postsurgical procedures. A.H. performed lesion analysis. J.C. assisted with manuscript concept ideation. A.S. performed veterinary consultation and nonbrain organ toxicology. D.M. performed nonbrain organ toxicology and manuscript editing. The corresponding author, S.M. performed manuscript concept ideation and manuscript editing.

Data Availability

All relevant data are included in the manuscript or the Supplementary material.

Author notes

N.K. and R.L. contributed equally to this work.

Present address: Stephenson School of Biomedical Engineering, University of Oklahoma, Norman, OK 73019, USA

Competing Interest: N.Ka, L.L.W., N.Ku., S.P., and S.M. are inventors of patent applications related to backpacks (owned and managed by Harvard University), and S.M. is a shareholder and a member of the board of directors of Hitch Bio.