Light-Triggered Drug Release from Red Blood Cells Suppresses Arthritic Inflammation

Abstract

Arthritis is a leading cause of disability in adults, which can be intensely incapacitating. The location and intensity of the pain is both subjective and challenging to manage. Consequently, patient-directed delivery of anti-inflammatories is an essential component of future therapeutic strategies for the management of this disorder. We describe the design and application of a light responsive red blood cell (RBC) conveyed dexamethasone (Dex) construct that enables targeted drug delivery upon illumination of the inflamed site. The red wavelength (650 nm) responsive nature of the phototherapeutic was validated using tissue phantoms mimicking the light absorbing properties of various skin types. Furthermore, photoreleased Dex has the same impact on cellular responses as conventional Dex. Murine RBCs containing the photoactivatable therapeutic display comparable circulation properties as fluorescently labelled RBCs. In addition, a single dose of light-targeted Dex delivery is 5-fold more effective in suppressing inflammation than the parent drug, delivered serially over multiple days. These results are consistent with the notion that the circulatory system be used as an on-command drug depot, providing the means to therapeutically target diseased sites both efficiently and effectively.

Graphical Abstract

Red blood cells containing a dexamethasone (Dex)-vitamin B12-Cy5 conjugate serve as a circulating drug depot that releases Dex, an anti-inflammatory agent, upon exposure to red light (650 nm). The photoactivatable therapeutic construct displays anti-inflammatory efficacy in a rheumatoid arthritis mouse model.

1. Introduction

Joint pain and inflammation are a leading cause of disability amongst working age adults with staggering societal costs.^[1] For example, the economic burden associated with arthritis is estimated to be greater than $300 billion in the United States alone.^[2] As a consequence of the persistent nature of these diseases, frequent and long-term therapeutic administration is required, which results in moderate to severe undesired side effects. Furthermore, the therapeutic needs are accentuated during periods of profoundly increased disease activity, which can be intensely debilitating. Efforts to improve efficacy and reduce undesired systemic toxicity have focused on technologies which can selectively deliver therapeutics to inflamed joints.^{[3, 4]} Indeed, it has been known for decades that the direct injection of glucocorticoids into arthritic joints provides temporary benefits;^[5] however, multiple injections into joints on a routine basis is not an acceptable therapeutic option.^[3] Nonetheless, the repeated delivery of therapeutic agents to afflicted joints is required to silence local inflammation and repair damage.^[6] A stimuli-responsive drug delivery system could potentially be used to intermittently dispense therapeutic agent(s) at the diseased site in a patient-directed, as-needed, fashion. As noted above, light as a stimulus enjoys a number of potentially useful attributes, including that it can be easily focused on inflamed and painful joints using readily available 600 to 1000 nm laser and light-emitting diode (LED) light sources.^[7] A light-activated form of an anti-inflammatory agent, such as dexamethasone (Dex), would ideally be maintained in the circulatory system in an inactive state, and subsequently released using light by the patient when needed. Indeed, the circulatory system represents an opportune drug depot since all cells in the human body are positioned within 100 μm of a blood vessel. However, repeated delivery to inflamed joints over the course of days or weeks requires a circulatory presence significantly longer than the half-life (a few hours) of the parent drug. We have addressed this issue by installing a photoactivable Dex inside red blood cells (RBCs) and have employed these engineered cells to successfully treat a mouse model of inflammatory arthritis. For clinical relevance, the overwhelming majority of RBC drug loading studies have been performed with human erythrocytes (hRBCs). However, in preparation for animal studies, loaded mouse RBCs were also prepared and characterized.

2. Results and Discussion

2.1. Design of a Circulating Photoresponsive Anti-inflammatory Drug Depot

Novel therapies designed to maintain Dex’s circulatory presence while minimizing its systemic side effects employ slow release carrier-based systems such as liposomes, polymeric-drug conjugates, and RBCs.^[8] Of these carriers, RBCs present a potentially elegant solution to the challenge of creating a long term drug depot that circulates throughout the body in an innocuous, dormant form.^{[9, 10]} Indeed, internally loaded RBCs have been reported to circulate for at least 1 month.^[10] The strategy outlined herein employs engineered RBCs that stably house a Dex derivative that is released upon exposure to red light 650 nm. A key element of the design strategy is the covalent attachment of Dex to vitamin B12 (cobalamin, Cbl), where Cbl serves three roles (Figure 1A).^[11] First, Cbl is membrane impermeable, which ensures that the internally loaded Cbl-Drug is retained by the RBC (Figure 1B). Second, Dex is appended to the central Co of Cbl via a light cleavable Co-C bond. Exposure to the appropriate wavelength severs Dex from the Cbl anchor, enabling the drug to freely diffuse out of the RBC. Third, although the corrin ring of Cbl absorbs only short wavelength light (330 – 575 nm), installation of Cy5 on the Cbl^[12–14] adjusts photo-release of the appended drug to a longer, tissue-penetrating wavelength (650 nm).^[15]

Figure 1.

Assembly of Dex-Cbl-Cy5 RBC Phototherapeutics. (A) Structures of photoactivatable Dex-Cbl-Cy5 (1) and control compound H₂O-Cbl-Cy5 (2). (B) Schematic representation of the isotonic-to-hypotonic-to-isotonic method by which phototherapeutics are loaded into RBCs and subsequently photochemically released. Dex is represented by the blue sphere. Pore formation in the RBC membrane occurs in the presence of 1 or 2 under hypotonic conditions (4 °C for 40 min). Pores are subsequently resealed by direct addition of high salt followed by incubation at 37 °C for 20 min. Conjugates 1 or 2 remain trapped inside the RBC due to the membrane impermeability of the B12 anchor. Upon photolysis of the C-Co bond, the now membrane permeable Dex is released from the RBC carrier.

2.1.1. Synthesis of Cbl Conjugates

Both Dex-Cbl-Cy5 (1) and H₂O-Cbl-Cy5 (2) (Figure 1A) were synthesized and introduced into RBCs to assess drug photo-delivery as a potential therapeutic strategy for the treatment of inflammatory arthritis. H₂O-Cbl-Cy5 serves as an inactive control that lacks the Dex therapeutic agent. In brief, Dex was appended to the aminopropyl ligand on the Co and Cy5 subsequently coupled to an ethylenediamine linker on the ribose of Cbl (Scheme S1 and Scheme S2, Figure S1 – S4, Supporting Information). The Cy5 fluorophore extends the light capturing wavelengths of Cbl from 330 – 575 nm to 650 nm, where the latter displays greater tissue penetration than that of the former (Figure S5, Supporting Information).^{[11, 13]} LC-MS analysis of the resulting mixture after photolysis of Dex-Cbl-Cy5 confirms the expected photoproducts free Dex and H₂O-Cbl-Cy5 (Scheme S3 and Figure S5, Supporting Information).

2.1.2. Assembly of Phototherapeutic RBCs

RBCs internally loaded with either 1 or 2 were prepared using a hypotonic swelling procedure (Figure 1B).^{[11, 12, 16]} Exposure of RBCs to a low ionic strength buffer solution induces cell swelling and pore formation within the cell membrane, which enables otherwise impermeable compounds to enter RBCs. The pores are subsequently resealed upon exposure to a high salt solution to reestablish an isotonic environment, which internally entraps the Cbl derivatives inside the RBCs (Figure S6 and Figure S7, Supporting Information).^[17] Although loading conditions are well established for hRBCs, mouse RBCs (mRBCs) are less stable than their human counterparts as demonstrated by their accelerated hemolysis and aging.^{[11, 18]} We found that modification of the established loading protocol improves the stability of the loaded mRBCs.^[19] Key optimized parameters include lengthening the drug loading and membrane resealing times, the high salt addition to return the RBCs to an isotonic environment, and the presence of ATP (see Materials and Methods section for details).^[20]

2.2. Characterization of Phototherapeutic RBCs

The hRBCs and mRBCs loaded with 1 or 2 were assessed for overall volume and hemoglobin content, cellular distribution of the phototherapeutic, and loading homogeneity and quantity of Dex in loaded RBCs. The results from these studies were compared with those obtained for unmodified RBCs and for RBCs surface-loaded with the lipidated indocarbocyanine fluorophore DiI (1, 1’-dioctadecyl-3, 3, 3’, 3’- tetramethylindocarbocyanine perchlorate; λ_ex 550 nm, λ_em 570 nm). The latter is noncovalently anchored via insertion of the lipid tails into the outer membrane sheath of the RBC. In addition, the DiI surface-loaded cells were used as a control circulation population serving as a comparison to internally loaded RBCs for in vivo studies (Figure S8).^[21] Specifically, no pores were opened in the membrane of DiI surface-loaded cells and, as a consequence, their internal contents are not perturbed. The mean cell volume (MCV), mean cell hemoglobin (MCH), and mean cell hemoglobin concentration (MCHC) of both hRBCs and mRBCs were assessed by automated hemocytometry (Table 1).^[22]

Table 1. Mean corpuscular volume (MCV), mean cell hemoglobin (MCH), and mean corpuscular hemoglobin concentration (MCHC) in human and mouse RBCs.

Data presented as mean ± SD, n = 3; where NA = not applicable and Q = quantitative.

Properties	Native*	DiI*	1*	2*	Native†	DiI†	1†	2†
MCV [fl]	98 ± 5	100 ± 1	70 ± 7	76 ± 5	34 ± 1	36 ± 3	29 ± 3	33 ± 4
MCH [pg]	28 ± 2	30 ± 1	18 ± 1	21 ± 2	13 ± 1	14 ± 1	8 ± 1	8 ± 1
MCHC [g/dL]	29 ± 1	30 ± 1	25 ± 3	28 ± 2	38 ± 1	40 ± 1	27 ± 4	23 ± 1
Cell recovery after loading	NA	Q	Q	Q	NA	Q	25–50%	25–50%

^*human RBCs;

^†mouse RBCs

RBCs internally loaded with 1 or 2 are smaller and contain less hemoglobin on average than native RBCs. As expected, RBCs that are surface loaded with DiI display values similar to native RBCs (Table 1). Trends are comparable for both mRBCs and hRBCs, consistent with the notion that opening and resealing RBCs results in the loss of some of the intracellular contents. In addition, although the recovery of hRBCs following loading is essentially quantitative, a significantly lower recovery was obtained for mRBCs. These results are consistent with the observation that hRBCs are more stable, under ex vivo conditions, than their murine counterparts.^[11]

Imaging flow cytometry was used to qualitatively visualize the various RBC populations and to quantitively assess changes in diameter and SSC (side light scattering) (Figure S9, Supporting Information). These studies confirm that RBCs containing either 1 or 2 exhibit a decrease in both size (diameter) and a decreasing shift in the SSC compared to native or DiI surface-loaded RBCs (Figure 2B–C and 2E–F). The decrease in diameter of internally loaded RBCs is consistent with hemocytometry MCV analysis and suggests that loss of some of the hemoglobin during the drug loading process is responsible for the smaller cell size. SSC is commonly related to the internal complexity (microparticles) of the cell. Since RBCs are presumed to have a relatively homogenous refractive index,^[23] this implies that the refractive index is altered upon the partial replacement of hemoglobin with the Cbl conjugates 1 or 2.

Figure 2.

Characterization of native and loaded RBCs. (A) Imaging flow cytometry of hRBCs. Brightfield images (left) and fluorescent images (right) of native, surface loaded (DiI), and internally loaded (1 or 2) hRBCs. Scale bar represents 7 μm. B-C) Plot of SSC versus diameter of loaded hRBCs. (B) hRBCs loaded with 1 (red) display a greater variability in SSC compared to DiI (yellow) and native hRBCs (black x). (C) hRBCs loaded with 2 (blue) exhibit morphological changes comparable to those displayed by hRBCs containing 1 (red). (D) Imaging flow cytometry of mRBCs. Brightfield images (left) and fluorescent images (right) of native, surface loaded (DiI), and internally loaded (1 or 2) mRBCs. Scale bar represents 7 μm. E-F) Plot of SSC versus diameter of loaded mRBCs. (E) mRBCs loaded with 1 (red) display a greater variability in SSC compared to surface loaded DiI (yellow) and native mRBCs (black x). (F) mRBCs loaded with 2 (blue) exhibit morphological changes comparable to those displayed by mRBCs containing 1 (red). The minor population of cells with the low SSCs in 2B-F are likely RBC ghosts.^[24] Imaging data was used to calculate the diameter of each cell/event as previously described.^[25]

We also assessed the uniformity of drug loading in RBCs and quantified the amount of Dex loaded. Fluorescent imaging flow cytometry revealed a uniform peak of loaded 1 and slightly less uniform 2 in hRBCs and mRBCs (Figure S10, Supporting Information). The amount of Dex-Cbl-Cy5 and H₂O-Cbl-Cy5 loaded was quantified by ethanol extraction of loaded RBC pellets and subsequent measurement of the absorbance of Cy5 at 649 nm (Figure S11 and Figure S12, Supporting Information). mRBCs contain 0.4 μg Dex in 100 μL of a mRBC pellet. Approximately 5% of Dex is loaded into RBCs.

2.3. Red Light Stimulates the Release of Dex from RBCs

We subsequently examined the light-triggered release of Dex from loaded hRBCs, which was quantified by LC-MS (Figure 3A and S13 and Scheme S3, Supporting Information). In addition, following treatment, the RBC-containing solution was centrifuged and the presence of free Dex examined in both the pellet and the supernatant. We did not detect free Dex in the absence of illumination (Figure 3A). By contrast, 5 min exposure to a 645 nm light source released the vast majority (90 ± 5.8%) of Dex that had been loaded into RBCs, consistent with the notion that Dex, once released from the membrane impermeable Cbl, is free to diffuse out of the cellular carrier (Figure 3A). Our previous studies with an alkyl-Cbl-Cy5 derivative revealed a quantum yield of ~0.1.^[26]

Figure 3. Photolysis of Dex-Cbl derivatives and release from hRBCs.

(A) hRBCs containing 1 were kept in the dark or exposed to 645 nm light for 5 min. Following illumination, the RBCs were centrifuged and the presence of free Dex in the supernatant (black bars) and pellet (white bars) quantified via LC-MS. Free Dex is not detected in the absence of illumination whereas, upon illumination, 90 ± 5.8% of free Dex is present in the supernatant. In addition, the photolyzed byproduct, H₂O-Cbl-Cy5 (2) is not detected in the supernatant. (B) Illumination (645 nm) of a buffered solution of 1 in the presence of Fitzpatrick phantom solutions. After 2 min of light exposure, there was no significant difference detected in Dex-Cbl-Cy5 photolysis with any of the Fitzpatrick types as measured by one-way ANOVA (n=3, NS). (C) Illumination (645 nm) of 1 embedded in hRBCs in the presence of Fitzpatrick phantom solutions. After 2 min of light exposure, there was no significant difference detected in Dex-Cbl-Cy5 photolysis from RBCs with any of the Fitzpatrick types as measured by one-way ANOVA (n=3, NS). All samples were illuminated with a light intensity of 1.0 mW/cm² for up to 6 min.

The Dex-Cbl-Cy5 phototherapeutic is designed to respond to red photons, which are not as extensively absorbed by tissue as compared to blue light. Indeed, wavelengths in the far red/near IR achieve tissue depths of up to several cm under optimized conditions.^[27] We explored the relative efficacy of Dex photorelease from Dex-Cbl, which lacks a red light absorbing antenna, and Dex-Cbl-Cy5, in the presence of the Fitzpatrick series of tissue phantoms. The latter are devised to mimic the tissue absorbing properties of human skin color, from lightly to heavily pigmented (Figure 3B and 3C).^{[26, 28]} We examined the photolysis of 1 filtered through Fitzpatrick skin phantom solutions that reproduced the properties of light (Type I-II; [melanin] = 8.8 μg/mL), brown (Type III-IV; [melanin] = 66 μg/mL), and dark (Type V-VI; [melanin] = 130 μg/mL) skin.^[26] Illumination (510 nm) through the Fitzpatrick phantoms fails to produce significant photolysis of a buffered solution of Dex-Cbl (Scheme S4 and Figure S14, Supporting Information). Nearly 90% of Dex-Cbl is unphotolyzed even after 6 min of illumination in the presence of type I/II Fitzpatrick phantom (Figure S15). As expected, tissue phantoms containing greater melanin concentrations are even more effective at blocking photolysis. By contrast, analogous experiments performed at 645 nm with 1 result in the near complete photolysis after only 3 min (Figure 3B). Finally, we examined the photorelease (645 nm) of Dex from hRBCs containing 1 in the presence of the tissue phantoms (Figure 3C). Reassuringly, illumination triggers the rapid photolysis of 1 and subsequent release of Dex from RBCs.

2.3.1. Photoreleased Dex Triggers Glucocorticoid Receptor α Translocation

Our initial studies on the potential therapeutic efficacy of photoactivatable Dex were performed on a variety of cultured human cells. Dex mediates its therapeutic action via the glucocorticoid receptor α (GRα), which is normally found in the cytoplasm but localizes to the nucleus in response to glucocorticoid binding.^[29] Upon entry into the nucleus, GRα associates with glucocorticoid responsive elements and stimulates or represses target gene expression.^{[29, 30]} We assessed the ability of Dex, photoreleased from RBCs bearing 1, to trigger GRα in HeLa cells and primary fibroblast-like synoviocytes (FLS). HeLa cells were used as a model cell line due to their established sensitivity to conventional Dex (Figure S16, Supporting Information).^[31] In contrast, FLS are non-immune cells that participate in rheumatoid arthritis (RA) pathogenesis. The FLS employed in this study were collected and cultured from RA patients.

Both HeLa and FLS cells were exposed to hRBCs bearing 1 in the dark and at 660 nm. In addition, RBCs containing 2 (i.e., no Dex) was used as a negative control and Dex itself was employed as a positive control. HeLa cells respond in a manner consistent with literature precedent (Figure 4A).^[31] In the absence of 660 nm exposure, GRα is primarily retained in the cytoplasm in untreated HeLa cells and in cells co-incubated with hRBCs containing either 1 or 2. By contrast, robust GRα nuclear localization is clear in Dex exposed HeLa cells. Analogous experiments were performed in the presence of 660 nm light. Untreated HeLa cells, as well as those co-incubated with hRBCs containing 2 (negative control), fail to display a nuclear GRα migration. However, hRBCs loaded with 1 trigger the anticipated GRα migration in a fashion consistent with that observed with the parent glucocorticoid. These experiments were recapitulated using primary FLS cells from RA patients. We do note that FLS cells are phenotypically heterogeneous^[32] and do not display the near 100% uniform GRα nuclear migration observed with HeLa cells. However, both 660 nm-exposed hRBCs bearing 1, and Dex itself, trigger the same degree of GRα relocation in FLS cells (Figure 4B, Figure S17 and Figure S18, Supporting Information).

Figure 4. Efficacy and Toxicity of Dex-Cbl-Cy5 hRBCs.

(A) Immunocytochemical analysis of GRα nuclear localization in HeLa cells. HeLa cells were cultured in serum free media for 24 h and then left untreated (column 1), exposed to hRBCs containing 2 (column 2), Dex (250 nM; column 3), or hRBCs containing 1 (column 4) in the absence (row 1) or presence of 660 nm light (row 2). Cells were fixed and permeabilized, exposed to anti-GRα, and subsequently anti-rabbit secondary antibodies conjugated to Alexa Fluor 488 (green). Cells were also labeled with the nuclear Hoescht 33342 stain (blue). Images display merged green and blue channels where scale bars represent 30 μm (Representative 1 of 5). (B) Immunocytochemical analysis of GRα nuclear localization in FLS cells. FLS cells were exposed to the conditions for HeLa cells as described above without serum starvation. Scale bars represent 50 μm (Representative 1 of 4). (C) The effect of 660 nm illuminated hRBCs loaded with 1 or 2 on the viability of FLS and Sup-B15 B cells. hRBC treatments with 1 have a modest impact on FLS cell viability, but a more substantial effect on Sup-B15 cells (n = 3, *P<0.05, **P<0.01).

2.3.2. Photoreleased Dex Does Not Impact FLS Viability

Glucocorticoids are known to induce apoptosis in certain cell types, which serves as both a mechanism of action for anti-inflammatory effects as well as a contributor to certain negative side effects (such as osteoporosis).^[29] With this in mind, we examined the impact that free Dex, or hRBCs containing 1 or 2, have on Dex resistant and sensitive cells. FLS cells experience minimal growth inhibition in response to treatment with Dex (Figure S19 and Figure S20, Supporting Information). The viability of FLS cells is likewise minimally impacted upon exposure to illuminated hRBCs bearing either 1 or 2 (Figure 4C). However, unlike synoviocytes, B cells are known to be sensitive to Dex (Figure S21 and Figure S22, Supporting Information).^[33] We examined the effect of hRBCs loaded with either 1 or 2, in the presence of 660 nm light, on Sup-B15 lymphoma B cell viability. As expected, both Dex photo-released from Dex-Cbl-Cy5 hRBCs, as well as the parent drug (Dex), impact B cell viability (Figure 4C). These results suggest that Dex, delivered to the site of inflammation, should reduce the localized immune response responsible for symptoms associated with arthritis. Furthermore, since synoviocytes play a key role in producing extracellular components of the synovial fluid, it is reassuring that the photorelease of Dex from RBCs does not impact FLS viability.

2.4. Photoactivated Treatment of Collagen Antibody-Induced Arthritis in a Mouse Model

2.4.1. The Circulatory Integrity of Internally and Externally Modified RBCs is Similar

The circulatory integrity of modified RBCs was examined with a 1:1 mixture of two mRBC populations: (i) internally loaded with H₂O-Cbl-Cy5 (2) and (ii) surface-loaded with DiI (Figure 5A). A mixture of the two populations were tail vein injected into mice and a blood sample was subsequently acquired after 20 min. Flow cytometry revealed that both cell types constitute approximately 3% of all circulating RBCs. Subsequent blood sampling after 1 h revealed minimal loss of circulating modified RBCs, indicating that Dex-loaded mRBCs circulate to the same extent as their fluorophore surface labelled counterparts.

Figure 5. Light-stimulated Dex treatment of CAIA mice.

(A) Percent of loaded mRBCs circulating in mice (n = 3) at 20 min and 60 min. White bars represent control surface-loaded DiI RBCs, while black bars represent RBCs containing 2. (B) There is significantly less inflammation (*P = 0.01) in mice treated with RBCs containing 1 (dotted line) than mice exposed to RBCs loaded with 2 (grey line) after the first 24 h of treatment. IP Dex serves as a positive treatment control (black line) but does not achieve a statistically significant improvement at 24 h. (C) Over multiple days mice treated with mRBCs loaded with 1 and treated with serial laser (dotted line, n = 13, P = 0.0002) continued to maintain remission of inflammation while arthritis in IP Dex treated mice (black line, n = 11, P = 0.0001) improved after 1–4 injections, depending on the individual mouse. Both treatments are significantly different over time when compared to the illuminated daily inactive control (2) loaded mRBCs which continue to worsen (gray line, n = 12) despite daily laser treatment, but are comparable to each other (P = 0.09). (D) Shaded bars furnish the ratio of mice that received additional doses (black) of IP Dex over the treatment course to achieve remission and the cumulative dose received. Dex from loaded mRBCs (gray, n = 13) is compared to the average IP Dex (hatched, n = 11). Mice treated with 1 loaded mRBCs only required 1 intravenous infusion on day 0 and subsequent daily laser application to the arthritic paw to achieve remission of inflammation (gray). The RBC group containing 1 received 78% less Dex (mg/kg) than mice treated with IP Dex to achieve similar clinical remission of inflammation (P<0.0001).

2.4.2. Phototherapeutic RBCs Suppress Inflammation in an Arthritic Mouse Model

The phototherapeutic efficacy of mRBCs containing 1 was evaluated using the collagen-antibody induced arthritis (CAIA) mouse model. CAIA is an accepted animal model of inflammatory arthritis ^[34], including a dependence on inflammatory mononuclear cells in acute inflammation and joint erosion in end-stage disease.^[35] Particular advantages to the CAIA model include its rapid onset and highly consistent penetrance of disease with minimal inter-experimental variability.^[36] CAIA was induced by an intraperitoneal (i.p.) injection of arthrogenic mAb 5-clone cocktail into DBA1/J mice on day 0 followed by i.p. injection of lipopolysaccharide (LPS) on day 2 according to manufacturer instructions. Arthritis was measured daily by a blinded observer using a clinical disease score index 0 – 4, where 0 = normal paw; 1 = mild but definite swelling of the ankle or digits; 2 = moderate redness and swelling of an ankle ± any number of digits; 3 = moderate redness and swelling of the entire paw; and 4 = maximal redness and swelling of the entire paw and digits, with or without ankylosis.^[37] CAIA mice were tail vein injected with a 90% hematocrit of allogenic, strain-specific mRBCs containing 1 or 2 (negative control). A single arthritic paw was illuminated with a 3 mW 635 nm laser for 5 min (Scheme S5, Figure S23, and Figure S24, Supporting Information). In addition, a positive control treatment group was treated using a standard protocol^[38] (0.5 mg/kg) dexamethasone sodium phosphate i.p. (“IP Dex”). Within 24 h following treatment, only mice exposed to mRBCs containing 1, with illumination of the inflamed paw, displayed statistically diminished arthritic severity relative to animals treated with mRBCs containing 2 (P = 0.01). By contrast, the difference in arthritic severity of laser treated paws of animals receiving IP Dex at 24 h was not significant when compared to inactive control 2 mRBC treatment (Figure 5B). If clinical paw swelling was still present in an IP Dex-treated mouse, IP Dex was continued in an individual mouse until clinical remission was achieved. Thus, IP Dex mice received between 1 and 4 doses of steroid, depending on severity of disease (Figure 5C). Specifically, on the second day of treatment, 8 of 11 mice IP Dex mice had clinical inflammation and were treated with a second dose of IP Dex. Similarly, mice treated with RBCs containing 1 had serial laser treatments on the arthritic paw until clinical inflammation was resolved. All treatments were completed, and the experiment terminated after 4 days for all groups. The disease curves of both IP Dex and mRBCs containing 1 achieve statistical difference in arthritis severity when compared to the RBC 2 inactive control (P = 0.0001 and P = 0.0002 respectively) and are comparable to each other (P = 0.9). However, it is notable that 78% less Dex was administered in the single 1 mRBC treatment dose with serial laser treatments as compared to average systemic treatments of the IP Dex to achieve comparable remission (Figure 5D). It is also noteworthy that the RBC group containing 2 had progressive worsening of clinical inflammation in the paw that was treated with laser daily for 4 days (Figure 5C).

3. Conclusion

Arthritis therapies are constrained by (1) the failure to deliver sufficient drug quantities to the inflamed site to achieve the desired therapeutic effect, (2) the range of moderate to severe side effects associated with long-term systemic exposure, and (3) the inability of the patient to self-administer therapeutics in a site-targeted as-needed fashion. There is a compelling biomedical need to develop a technology to address these issues, especially given the chronic nature and prevalence of arthritis, the capricious acute episodes of pain and inflammation, interpatient heterogeneity, and different disease subtypes of arthritis that affect a broad swath of the population. We have developed a light activation strategy that triggers the delivery of the anti-inflammatory agent, Dex, in a site-specific fashion. Given the therapeutic potential established by clinical studies of internally loaded RBCs and the Red Cell Loader and ERYcaps RBC loading devices, RBCs conveying a photoactivatable therapeutic offer the opportunity to target inflamed diseased sites in a patient-directed fashion by employing the circulatory system as a drug depot.^{[10, 39]} We have shown that RBCs harboring the phototherapeutic agent display properties similar to those of native and surface labelled erythrocytes. Furthermore, given the red-light responsiveness of the carrier system, drug release is efficiently triggered through tissue phantoms that mimic the entire range of pale to dark skin types. We have demonstrated that the RBC-conveyed photo-anti-inflammatory agent not only induces the desired reduction in inflammation but does so at a dose that is significantly lower than the standard of care for the parent drug. Finally, we note that the inherent flexibility of the drug delivery platform, particularly with respect to designating the wavelength of photorelease, may ultimately enable a variety of arthritis drugs to be separately delivered to designated sites via wavelength modulation. These studies are in progress.

4. Experimental Section/Methods

Synthesis of Dex-Cbl-Ethylenediamine (Dex-Cbl-EDA, Scheme S1, Supporting Information):

Succinyl-Dex^[40] (0.041 mmol) was dissolved in dry DMF (5 mL) along with DIPEA (0.32 mmol) and HATU (0.047 mmol). The solution was stirred for 5 min to activate the carboxylic acid. After 5 min, Propylamine-Cbl-Fmoc (0.041 mmol) was added to the solution, and the reaction vessel was placed on a shake plate for 4 h. After 4 h, the solution was diluted with diethyl ether (to 50 mL) and spun down to pellet the product. Then the supernatant was decanted. The Dex-Cbl-Fmoc pellet was allowed to dry for 4 h, then dissolved in dry DMF (5 mL). Piperidine (0.29 mmol) was added to the solution to deprotect the Fmoc group. The solution was stirred for 4 h, then diluted with diether ethyl (to 50 mL) to precipitate the product, which was pelleted via centrifugation. The pellet was dissolved in MeOH (10 mL) and then diluted with DI water (to 40 mL). A 12 g Biotage C18 column with sample was prepped with 5:95 MeOH:H₂O (0.1% TFA) and the solution was eluted using an MeOH:H₂O (0.1%TFA) gradient from 10% to 80% MeOH. Fractions were analyzed for purity via LC-MS (Figure S1, Supporting Information) and combined, concentrated under reduced pressure, and lyophilized to yield a powdered orange solid. ¹H NMR (Figure S2, Supporting Information; 500 MHz, MeOH-d₄).

Synthesis of Dex-Cbl-Cy5 (1, Scheme S2, Supporting Information):

Cy5 (0.0204 g, 0.04 mmol) was dissolved in dry DMF (5 mL) creating a dark blue solution. DIPEA (0.0481 g, 0.37 mmol) and HATU (0.0113 g, 0.03 mmol) were added to solution and the reaction mixture was placed on a shake plate for 10 min. Dex-Cbl-Eda (0.0495 g, 0.025 mmol) was then added to the solution and the reaction vessel shaken for an additional 2.5 h. The solution was diluted with diethyl ether (to 50 mL) and centrifuged to pellet the Cbl conjugate. Residual ether was removed by “air drying” for 4 h in the dark. The pellet was then dissolved in MeOH (4 mL) and diluted with DI H₂O (to 15 mL). A 30 g Biotage C18 column with sample was prepped with MeOH:H₂O (0.1% TFA) and the solution was eluted using an MeOH:H₂O (0.1%TFA) gradient from 0% to 80% MeOH. Fractions were analyzed for purity via LC-MS (Figure S3, Supporting Information) and combined, concentrated under reduced pressure, and lyophilized to yield a blue solid (49.2 g, 80.2% yield). ¹H NMR (Figure S4, Supporting Information; 500 MHz, MeOH-d₄).

Encapsulation of 1 and 2 in mRBCs:

Mouse blood was collected from DBA-1J mice, and the RBCs immediately isolated by filtering through Ficoll-Paque at 400 g for 30 min. The serum and white blood cells were removed and packed mRBCs were washed 3 times with 1X PBS at 500 g for 4 min. RBCs were then immediately loaded with Cbl derivatives as previously described with slight modifications to allow for stable loading of larger RBC pellets.^[11] Most commonly, mRBC pellets (400 μL, 100% hematocrit) were prepared for loading by addition of a solution of 1 (13.8 mM, 8 μL) or 2 (19 mM, 6 μL) in DMSO and diluent C (164 μL) to create a mixture of mRBCs (70% hematocrit) and Cbl conjugate (200 μM). The RBC/Cbl mixture was dialyzed for 40 min in 400 mL of dialysis buffer (80 mOSm/L PBS, 0.25% glycerol, 10 mM glucose, 2 mM ATP) at 4 °C. RBCs were removed from dialysis tubing and resealed by adding 0.1 volume 10X PBS per vol of dialyzed RBCs and by incubating for 20 min at 37 °C. Loaded, resealed RBCs were washed 5x with 1X PBS or 3x with 1X PBS and 2x with FBS. All manipulations of light responsive compounds occurred in the dark room to prevent exposure to ambient light. The entire isolation and loading of mRBCs was completed within 6 h.

Imaging Flow Cytometry:

RBCs loaded with the Cbl conjugates 1 or 2 or surface-loaded with DiI were washed with 1X PBS and then diluted 1,000-fold with 1X PBS to achieve a concentration of 9 × 10⁷ RBCs/mL. Loaded cells were then evaluated using the ImageStreamX Mark II (Amnis, Seattle, USA). Samples were run at a speed of 25 μL/min and 100,000 unique cells were imaged in two brightfield channels (intensities of 64.6 mW and 93.94 mW), a side scatter channel (785 nm laser at an intensity of 2 mW with a 740 – 800 nm filter), a DiI channel (488 nm solid state laser at an intensity of 200 mW with a 560 – 595 nm filter), and a Cy5 Channel (658 nm diode laser at an intensity of 150 mW with a 660 – 740 nm filter) at 60X magnification. Cells were gated, first filtering for focus by requiring a gradient root mean square greater than 60, then analyzing for singlet cells using a gate within an Area vs Aspect Ratio scatter plot (Figure S9, Supporting Information). The fluorescence of the cells was then analyzed using an intensity histogram of the appropriate fluorescence channel. General morphology of cells was analyzed by comparing cells using the diameter feature (Erode mask, pixel 03) as established in previous methods and intensity of channel 6 (the side scatter channel).^[25] Images were analyzed using the IDEAS Software. Characterization of mRBCs was performed on the same day as their isolation and loading.

Automated Hemocytometry:

hRBCs were prepared as 10% hematocrit hRBC samples as described above and then analyzed via the Siemens Advia 2120i – automated CBC. RBCs were sphered and fixed and then automatically processed through a Flowcell where RBC Count, MCV and hemoglobin content were measured using a laser diode. Hemoglobin was measured in a colorimeter by lysing all RBCs to free hemoglobin. The free hemoglobin was converted to methemoglobin and then porphyrin and the color change was measured at 546 nm. MCH was calculated by the analyzer using the following equation: MCH = Hgb × 10/RBC. mRBCs were prepared as 10% hematocrit mRBC samples as described above. Samples were then analyzed using the IDEXX ProCyte Dx automated hematology instrument (software version 00–34 Build57) using settings for mouse whole blood. Cell recovery was assessed by measuring the RBC pellet volume before and after loading.

Photorelease of Dex from Dex-Cbl-Cy5 (1) hRBCs (Scheme S3, Supporting Information):

Dex-Cbl-Cy5 (Loaded into hRBCs, 20% hematocrit in human platelet rich plasma, 500 μL) were photolyzed as described below. Release of Dex into supernatant after photolysis was assessed by extracting Dex from plasma and pellet with acetonitrile followed by LC-MS analysis, comparing photolyzed product fragment ion 510 and 373 in the supernatant and pellet. Exposure to the light source was performed 5 – 10 min so that Dex-Cbl-Cy5 was fully photolyzed, as revealed by the absence of un-photolyzed starting material via LC-MS (Figure S13, Supporting Information).

Photolysis of Dex-Cbl-Cy5 (1) and Dex-Cbl with Fitzpatrick Skin Phantom Solutions:

Fitzpatrick skin phantom solutions were prepared following a previously reported protocol.^[26] Hematocrit lysate stock solution (25% hematocrit) was prepared via sonicated lysis of RBCs in water followed by centrifuge to remove residual membranes. All skin phantom solutions contained 0.3% v/v Intralipid®, 1% hemoglobin, and erythrocyte lysate solution (1% v/v) in PBS. Melanin stock solution (synthetic melanin from Sigma-Aldrich; ε at 645 nm ~1 – 3 mg/mL⁻¹ cm⁻¹; 3 mg/mL in 100 mM NH₄OH solution) was added at varying concentrations depending on Fitzpatrick skin type (I – II ~8.8 μg/mL; III – IV ~66 μg/mL; V – VI ~130 μg/mL). The skin phantom solution was inserted between the sample and the light source before illumination. Dex-Cbl (20 μM in MeOH/PBS 1:1, 200 μL) or Dex-Cbl-Cy5 (20 μM in MeOH/PBS 1:1, 100 μL) were placed in a cuvette and illuminated with an Oriel Xe Flash lamp (Model 60000) with a 510 nm filter (Newport 10BPF10–510, band width 10 nm) for Dex-Cbl and a 645 nm filter (Newport HPM645–50, band width 50 nm) for Dex-Cbl-Cy5. The sample was exposed to a light intensity of 1.0 mW/cm². For Dex-Cbl, the progress of photolysis was monitored by observing an absorbance increase at 350 nm as previously described.^[41] For Dex-Cbl-Cy5, the progress of photolysis was monitored by observing the fluorescence decrease using at 675 nm (with λ_ex at 645 nm). Completion of photolysis was confirmed by LC-MS. Analogous experiments were conducted using Dex-Cbl-Cy5 loaded human RBCs (20% hematocrit in human platelet rich plasma, 100 μL), which were added to 0.5 mL of human platelet rich plasma (Zen-Bio, Inc). Photolysis of the samples was performed using an Oriel Xe Flash lamp (Model 60000) (800 mJ, 62 Hz) as the light source with selective bandpass filters for 645 ± 10 for 5 min (dark samples were incubated in the dark for 5 min). Photolyzed RBC suspensions were centrifuged at 1000 g for 3 min, and the supernatant then analyzed by LC-MS (solvent A: 0.1% formic acid/H₂O; solvent B: 0.1% formic acid/methanol). The pellet was extracted with 0.5 mL of acetonitrile and analyzed by LC-MS as well. The photolyzed Dex product was monitored by its fragment ions 510 (Dex-aldehyde) and 373 (hydrolyzed Dex) and compared with standard photolyzed product from photolysis in PBS/MeOH (1:1). Acetonitrile was added to the supernatant in a 1:1 ratio, centrifuged at 21,000 g for 5 min, and removed from the pellet.

GRα Localization in Human Cells:

Human HeLa or FLS cells were seeded in 24-well glass-bottom plates treated with polylysine at a density of 5×10⁴ cells/mL and allowed to adhere overnight. HeLa cells were then serum starved for 24 h while FLS cells were immediately treated after adhering overnight. The cell cultures were subsequently exposed to 100 μL of RBCs-(1) at 10% hematocrit (7.9 × 10⁸ cells/mL), or 100 μL of RBCs-(2) at 10% hematocrit (8.5 × 10⁸ cells/mL), or 500 μL Dex at 250 nM, or 500 μL ethanol vehicle in plain DMEM using Millicell Hanging Cell Culture Inserts (1 μm, polyethylene terephthalate, Millipore). Light treated samples had been exposed to a 660 nm LED board for 30 min at RT. After 30 min of exposure, hanging wells were removed and cells were incubated for an additional 30 min at 37 °C in a humidified environment at 5% CO₂. Cells were then washed 3x with 1X PBS, fixed with 4% PFA in PBS for 10 min, washed with 2× 1X PBS (1 mL), and treated with methanol for 5 min. Fixed cells were treated with a blocking buffer (5% Donkey serum; 0.1% Triton X-100; PBS) and stained for GRα using monoclonal rabbit anti-GRα antibody (Abcam 181327) at 1:200 dilution in antibody dilution buffer (1% BSA; 0.1% Triton-X-100; PBS). Cells were then washed with PBS (3 × 5 min) while stirring before incubation with anti-rabbit AlexaFluor 488 secondary antibody (Life Technologies A11034) at 1:500 dilution in antibody dilution buffer for 1 h at room temperature. Cells were washed (1 × 5 min) with PBS and imaged on an inverted Olympus IX81 microscope equipped with a Hamamatsu FLASH 4V3, 60X oil objective, and a FITC filter cube (Semrock). Images were analyzed using ImageJ. FLS cells and quantified using ImageJ’s Coloc 2 program. Regions of Interest (created from the Hoechst nuclear stain) were analyzed for colocalization between the Hoechst nuclear stain and anti-GRα 488 fluorescence in the nuclear space. Pearson’s coefficients without thresholding were analyzed from four replicates (Figure S18, Supporting Information).

Cell Viability Upon Exposure to Dex:

Sup-B15 and FLS cell seeding densities were optimized to ensure cells linearly responded to MTS (Figure S19 and Figure S20, Supporting Information). Sup-B15 cells were then seeded in 24-well glass-bottom plates at a density of 1.7×10⁶ cells/mL and allowed to settle for 1 h, while primary FLS cells were plated at a density of 1×10⁴ cells/mL and allowed to adhere overnight. Cells were then treated with 100 μL of RBCs-(1) at 50% hematocrit (4.45 × 10⁹ cells/mL), or 100 μL of RBCs-(2) at 50% hematocrit (4.3 × 10⁹ cells/mL), or 500 μL free Dex at 2 μM, or 500 μL plain DMEM which was added to wells in Millicell Hanging Cell Culture Inserts (1 μm, polyethylene terephthalate, Millipore). Samples were exposed to a 660 nm LED board for 30 min at RT. After 30 min of exposure, hanging wells were removed and cells were incubated for 24 h at 37 °C in a humidified environment at 5% CO₂. Cells were then treated with 100 μL MTS/1 mL media for 2 h (Abcam 197010) after which absorption at 492 nm was measured using a HTS 7000 BioAssay Reader (Perkin Elmer, Waltham, Ma).

Assessment of Circulatory Viability of Modified mRBCs:

Whole blood of DBA-1J, 8 – 10 week, male donors was harvested on the day of loading by cardiac puncture. Internally loaded and control RBCs were prepared as described above. mRBCs loaded with 2 were resuspended in PBS at 50% hematocrit and 100 μL of cell suspension was drawn into syringes circumferentially covered with tape to block light. Surface-loaded control mRBCs (DiI) and internally (2) loaded mRBCs were mixed 1:1 v/v and 100 μL of the mixture was loaded in a syringe at 90% hematocrit and circumferentially taped to block light. Samples were injected into DBA-1J mice via the tail vein. At the described time points, blood was collected from mice via tail nicks. Mice were euthanized via CO₂ inhalation according to approved UNC IACUC protocol #19–048.0. Whole blood was diluted approximately 1,000–1,500-fold and analyzed on the Attune NxT flow cytometer as described below.

Attune NxT Flow Cytometry:

mRBCs loaded with the Cbl conjugates 1 or 2, or surface-loaded with DiI, or unmodified were diluted 1,000-fold with 1X PBS to achieve a concentration of 1 × 10⁷ RBCs/mL. Diluted samples were analyzed using the Attune NxT (ThermoFisher). First, a scatter plot of side scatter area (SSC-A) to side scatter height (SSC-H) was used to gate for single cells. Then, Cy5 from 2 was excited and measured using the 637 nm 100 mW laser and the RL1 detector (Mirror 654LP, Filter 670/14). Data was processed using FCS Express v.7.04 ensuring even flow of sample through the instrument by plotting time vs forward scatter area (FSC-A). Final data allowed for RBC size characterization via scatter plots of FSC-A vs SSC-A and loading uniformity via Cy5 fluorescence-RL1-area and DiI fluorescence-BL1-area.

Collagen Antibody Induced Arthritis (CAIA) Animal Model:

Collagen Antibody Induced Arthritis (CAIA) was induced in forty-five 8–10 week old male DBA/IJ mice via the i.p. injection of Chondrex Arthrogen-CIA 5-Clone Cocktail (1.6 mg/mouse) followed by an i.p. injection of lipopolysaccharide from E. coli (25 mg) 3 days later, as per the manufacturer’s instructions. Arthritis was measured daily by a blinded observer with a clinical disease score index and measurement of paw swelling. The clinical disease score index is performed with the following scoring system: 0 = normal paw; 1 = mild but definite swelling of the ankle or digits; 2 = moderate redness and swelling of an ankle ± any number of digits; 3 = moderate redness and swelling of the entire paw; and 4 = moderate redness and swelling of all three joints in the paw.^{[37, 42]} The maximum score/paw is 4. This method has been previously validated in inflammatory arthritis models.^[42] Assessment of disease occurred while the mouse was anesthetized using isoflurane.

Treatment of Mice with IP Dex or RBCs Containing 1 or 2 (Scheme S5, Supporting Information):

Mice were randomized after symptom onset in at least one paw with a clinical score ≥ 1 to one of three groups that received the following injections: mRBCs containing 1 delivered intravenously (1 mRBC; experimental treatment); mRBCs containing 2 delivered intravenously (2 mRBC; inactive control, no drug); or Dex sodium phosphate solution delivered i.p. (IP Dex; positive control). The 1 mRBC group received mRBCs (90% hematocrit, 100 μ L) internally loaded with approximately 0.0065 mg Dex-Cbl-Cy5 (1), equivalent to 0.31 mg/kg/mouse, and the IP Dex group received 0.5 mg/kg/mouse. Individual mice in the IP Dex group with persistent disease would be treated with additional injections of 0.5 mg/kg daily to achieve a clinical score of 0 in the arthritic paw. Injections of loaded RBCs were performed via the tail vein by a single, blinded technician using a 27G needle. A laser was applied to the first observed affected joint (only one paw per mouse) in all RBC groups for 5 min immediately following injection and each day until experiment termination at day 4. Mice were euthanized via CO₂ inhalation according to approved Duke IACUC protocol #A1851708. An RS laser diode (Power Technology Inc., Alexander AR, USA), 635 nm, 3 mW (Figure S23, Supporting Information) was used for laser treatment of the arthritic paw (clinical score ≥ 1). If two or more joints were simultaneously affected, then the laser-treated paw was selected at random. A single, blinded experimenter performed all laser treatments and both lower and upper extremity paws were treated in the study. The laser was applied to the selected joint at an angle that allowed for illumination of the entire joint and paw and was held in position for a total of 5 min per animal. Mice were anesthetized using isoflurane during laser treatment. Treatment was applied to the same paw daily from day of injection until termination.

Statistical Analysis:

All plots are presented as the mean ± standard deviation. Data was analyzed by an unpaired Student’s t test. GraphPad Prism software was used for statistical analysis. A P value of less than 0.05 is statistically significant. We have previously demonstrated that 10 animals per group are needed to achieve 80% power to detect a medium effect size (0.22) at a 0.05 level of significance in the CAIA model.^[42] There were >10 animals per group for each of the three experimental groups in this study. The area under the curve (AUC) of the arthritis severity score-time curve was calculated for laser paws in each treatment group. AUC differences were estimated by ANOVA. Comparisons were made across all three treatment groups, and pairwise comparisons were performed using Tukey’s test based on the LSMEANS statement. All analyses were conducted using SAS 9.4 (Cary, North Carolina). A one-way ANOVA was used for comparing laser-paw treatment groups at 24 h and was found to be significant (P = 0.05). A subsequent Mann Whitney U test comparing laser-paw Dex-Cbl-Cy5 (1) RBCs (n = 13) with H₂O-Cbl-Cy5 (2) RBC control (n = 12) identified significant difference (P = 0.01) at 24 h, whereas IP DEX (n = 11) did not achieve statistical significance from the control. An unpaired T test was used to analyze the cumulative steroid dose used to achieve clinical remission of the laser-treated paw between IP Dex (n = 11) mice and Dex-Cbl-Cy5 (1) RBCs (n = 13) mice, where IP Dex mice were treated daily with 0.5 mg/kg IP until the arthritis subsided (ranging between 1–4 steroid doses + laser) and Dex-Cbl-Cy5 (1) RBC mice received a single dose (0.31 mg/kg intracellularly loaded within 50% Hct) injected intravenously followed by daily laser for 3 additional days.