Nanofibrous supramolecular peptide hydrogels for controlled release of small-molecule drugs and biologics

Abstract

Maintaining safe and potent drug levels in vivo is challenging. Multidomain peptides assemble into supramolecular hydrogels with a well-defined, highly porous nanostructure that makes them attractive for drug delivery. However, their ability to extend release is typically limited by rapid drug diffusion. Here, to overcome this challenge, we present self-assembling boronate ester release (SABER) multidomain peptides capable of engaging in dynamic covalent bonding with payloads containing boronic acids. As examples, we demonstrate that SABER hydrogels can prolong the release of boronic acid-containing small-molecule drugs and boronic acid-modified biologics such as insulin and antibodies. Pharmacokinetic studies reveal that SABER hydrogels extend the therapeutic effect of ganfeborole from days to weeks, preventing Mycobacterium tuberculosis growth compared with oral administration in an infection model. Similarly, SABER hydrogels extended insulin activity, maintaining normoglycemia for 6 days in diabetic mice after a single injection. These results suggest that SABER hydrogels present broad potential for clinical translation.

Drug delivery systems can enhance the therapeutic efficacy of pharmaceutical drugs by maintaining drug concentrations within narrow therapeutic windows, thereby extending the duration of drug exposure while alleviating side effects¹. In the United States, poor adherence to medication leads to 10% of hospitalizations, US$100–300 billion in avoidable spending and over 125,000 deaths annually². Controlled-release systems are well positioned to improve dosing schedule adherence by reducing patients’ responsibility to take medications and drug-related side effects that can deter adherence². Contemporary parenteral drug delivery systems include biodegradable microparticles³, nanoparticles⁴, implantable devices⁵ and transdermal patches⁶ Each of these systems, however, possesses limitations, such as high initial burst release, poor drug loading, limited drug compatibility, poor payload stability, low biocompatibility, high cost, challenging scale-up and/or invasive administration.

Several hydrogel drug delivery systems have been developed in an attempt to address the challenges faced by current drug delivery systems. The substantial amount of water in hydrogels provides a biocompatible aqueous environment that facilitates the high loading of hydrophilic drugs and prevents sensitive biological payloads, such as proteins, from denaturing or aggregating⁷. Hydrogels can be formed from a variety of materials, such as carbohydrates, synthetic polymers, proteins, nanoparticles, small molecules and peptides^8,9. Supramolecular peptide hydrogels are particularly promising owing to their modularity and favourable biological properties^10,11. Typically composed of the same amino acids as endogenous proteins, they offer favourable biocompatibility and readily degrade into non-toxic products. In addition, the supramolecular interactions that form these materials enable injection through small-bore needles for minimally invasive administration¹². The translational potential of these materials has been demonstrated by the clinical adoption of PuraStat, which is a peptide hydrogel surgical haemostatic agent¹³. Many preclinical investigations have studied the use of peptide hydrogels for drug delivery, although these have yet to be used in the clinic^14,15. One key barrier to the clinical translation of these materials is their rapid release of small molecules and most proteins, which occurs via diffusion when the hydrogel’s mesh size is larger than its cargo^7,16.

Multidomain peptides (MDPs) are one class of supramolecular peptide hydrogels with a highly porous nanofibrous network¹⁷. Their nanostructure allows the formation of physically crosslinked hydrogels at low concentration (~1% by weight) that can be easily injected through a needle yet form a hydrogel bolus in vivo that remains at the injection site for weeks¹⁸. MDPs have been investigated for the delivery of diverse payloads such as carbohydrates¹⁹, proteins²⁰ and small molecules²¹, yet insufficient release lifetimes remain a challenge. A recent study showed that MDPs released a model protein payload in vivo over the course of 1–9 days depending on the peptide sequence²⁰, which would be insufficient to treat chronic conditions or diseases that require treatment for weeks, months or years.

Affinity-based release modalities, including complementary electrostatic interactions, guest–host binding, hydrogen bonding, ligand binding, hydrophobic interactions and dynamic covalent interactions, have been used to add attractive interactions between a payload and hydrogel, slowing transport out of the hydrogel network to prolong drug release^22–24. Dynamic covalent interactions are substantially stronger than electrostatic or hydrophobic interactions but exist at an equilibrium such that bonding is reversible²⁵. As a result, some fraction of drug remains covalently bonded to the hydrogel, while unbonded drug is free to diffuse until it exits, or bonds to another part of the gel, thereby extending drug release compared with diffusion alone.

The clinically relevant dynamic covalent bond between diols and boronic acid (BA) functional groups is particularly versatile²⁶. There are currently five US Food and Drug Administration (FDA)-approved BA-containing small molecules on the market and dozens more in preclinical and late-stage clinical development²⁷. BA modification is an emerging tool in medicinal chemistry to improve the activity and pharmacokinetics of small-molecule therapuetics^27,28. BAs readily form bidentate ester linkages with diols, even in aqueous solution²⁹. These boronate esters and their constituent free BAs and diols exist in a dynamic equilibrium³⁰. Catechols are the most commonly used diol motifs in BA dynamic covalent chemistry^31,32. These systems are limited, however, by the rapid oxidative decomposition of catechols at neutral pH^33,34. This oxidation can accelerate drug release, reduce payload stability and lead to crosslinked by-products via reactive quinone intermediates^35,36. Catechol-based delivery systems can also exhibit glucose-responsive release owing to competitive BA bonding with carbohydrate polyols, which has been leveraged for glucose-responsive insulin delivery for diabetes treatment³⁷. However, for almost every application outside of diabetes, it is undesirable to have the rate of drug release fluctuating with meals.

Herein we overcome the limitations of previous boronate ester-mediated release systems and covalently crosslinked hydrogels by functionalizing MDPs with 4-nitrocatechol or salicylhydroxamic acid (SHA) to improve oxidative stability, extend the duration of drug release and minimize the effects of glucose concentration on release kinetics^33,38,39. These self-assembling boronate ester release (SABER) hydrogels are capable of engaging in dynamic covalent bonding with BAs to extend the release of BA-containing small molecules, a small-molecule drug modified with a BA and BA-modified biologics (Fig. 1a). We demonstrate that the SHA modification allows SABER hydrogels to control the release of a broad range of drugs, even compounds where traditional catechol motifs fail, without exhibiting glucose sensitivity. By installing BA motifs onto small molecules and biologics, we expand the repertoire of compatible payloads for the SABER system. These approaches resulted in the SABER system’s ability to deliver a wider range of payloads than any other drug delivery system leveraging boronate ester chemistry, to the best of our knowledge. The ability to release multiple classes of drugs from SABER hydrogels makes it well suited for clinical translation.

Fig. 1 |. Characterization of SABER hydrogels.

a, A schematic illustrating how small molecules rapidly diffuse from peptide hydrogels owing to their large mesh size (red). We hypothesized that by adding dynamic covalent attachment moieties to the hydrogel, release can be slowed resulting in prolonged delivery (blue). b, Chemical structures of the three SABER peptides synthesized. c, Fourier transform infrared spectra of the three SABER peptides show that they assemble into antiparallel β-sheets. d, Circular dichroism spectra of unmodified K₂ and SABER MDPs are consistent with β-sheet secondary structure. e–g, Cryo-TEM images of Cat-K₂ (e), nitroCat-K₂ (f) and SHA-K₂ (g) show that all peptides self-assemble into nanofibres. h, Oscillatory rheology frequency sweeps show that the three SABER peptides at 10 mg ml⁻¹ form hydrogels with similar moduli to unmodified K₂. i, Shear recovery rheology experiments where all four peptides were subjected to 200% strain for 1 min, demonstrating that all the hydrogels are shear-thinning and rapidly self-healing. For panels h and i, the storage modulus (G′) is indicated by filled circles and solid lines and the loss modulus (G″) is indicated by open circles and dotted lines. Panel a created in BioRender. McHugh, K. J. (2025) https://biorender.com/093vn0b.

Design and development of SABER nanofibrous hydrogels

Motivated by the shortcomings of catechol–BA bonding, we aimed to identify chemical motifs that (1) engage in dynamic covalent bonding with BAs, (2) resist oxidation and (3) exhibit an equilibrium shifted more towards the bonded (boronate ester) state to increase drug retention in MDP hydrogels. UV–vis analysis showed that SHA and 4-nitrodopamine demonstrated excellent stability over 15 days in solution, whereas a model catechol, dopamine, rapidly degraded (Extended Data Fig. 1a–c). Investigations into the boronate ester association constants of these motifs indicated that SHA formed the strongest dynamic covalent interactions, exceeding the catechol association constants by 6.5- to 9.6-fold (Extended Data Fig. 1d).

On the basis of the oxidation and bonding results, we synthesized MDPs with the sequence K₂(SL)₆K₂ (referred to as K₂) functionalized with catechol (Cat-K₂), 4-nitrocatechol (nitroCat-K₂) or SHA (SHA-K₂) motifs to yield three SABER peptides (Fig. 1b and Extended Data Fig. 1e–h). The Fourier transform infrared spectra of these peptides showed antiparallel β-sheet secondary structure with peaks at 1,618 cm⁻¹ and 1,695 cm⁻¹ (ref. 40) (Fig. 1c), which was confirmed by a minimum approximately 216–220 nm in their circular dichroism spectra (Fig. 1d). Cat-K₂ and nitroCat-K₂ peptides formed long thin nanofibres as seen by cryo-TEM, whereas SHA-K₂ nanofibres appeared to cluster and form sheet-like bundles (Fig. 1e–g). Rheological testing confirmed hydrogel formation, as indicated by storage moduli (G′) that exceed the loss moduli (G″) (Fig. 1h). When subjected to 200% strain, G″ exceeded G′ for all gels, suggesting liquid-like flow under high shear. Hydrogel character rapidly recovered once the shear force was reduced, suggesting that these materials can reform hydrogels after injection (Fig. 1i).

Next, we sought to demonstrate controlled in vitro release of bortezomib and ixazomib (FDA-approved multiple myeloma therapeutics), tavaborole (an FDA-approved antifungal agent) and ganfeborole (a tuberculosis treatment in phase II clinical trials)^27,41. In the cases of bortezomib and ixazomib, unmodified K₂ and Cat- K₂ hydrogels released >82.3% and >57.1% of each drug in the first 2 h, respectively. By contrast, SHA-K₂ and nitroCat-K₂ hydrogels released 15–25% of each drug in 2 h and 50–70% within 24 h (Fig. 2a,b). Many motifs used in dynamic covalent BA chemistry are sensitive to glucose concentration²⁶, which would likely be detrimental in a cancer chemotherapy setting. The release of bortezomib from SHA-K₂ was found to be independent of glucose concentration across a physiologically relevant range (Extended Data Fig. 2a,b). Cat-K₂ had a lower burst release when loaded with tavaborole, with 80.0 ± 4.7% of the drug released in 24 h (Fig. 2c). SHA-K₂ and nitroCat-K₂, however, released significantly less tavaborole at 24 h (49.1 ± 0.3% and 64.8 ± 0.9%, respectively) than Cat-K₂ and K₂. Both nitroCat-K₂ and SHA-K₂ tavaborole release curves appear to plateau at later time points. We verified that the drug loaded into the gels was not irreversibly covalently bound to the peptide by ultra performance liquid chromatrography (UPLC) analysis (Extended Data Fig. 2c). In the case of ganfeborole, K₂ and nitroCat-K₂ had similarly rapid release rates (Fig. 2d), consistent with the low equilibrium constant between ganfeborole and 4-nitrodopamine (Extended Data Fig. 1d). The release of ganfeborole from Cat-K₂ could not be quantified owing to the rapid oxidative degradation of the drug³⁵ (Extended Data Fig. 2d–g). Only SHA-K₂ was able to significantly extend the release of ganfeborole compared with K₂, releasing less drug in 24 h (48.7 ± 2.5%) than K₂ released in 30 min (59.0 ± 1.1%). Varying the drug-to-peptide molar ratio had a limited effect on the peptide secondary structure (Extended Data Fig. 2h) and release below a drug-to-peptide ratio of 1:1 (Fig. 2e,f; additional discussion on SI-4).

Fig. 2 |. In vitro release of BA-containing small-molecule drugs.

a,b, In vitro release of bortezomib (a) and ixazomib (b) loaded at 0.5 mg ml⁻¹ shows that SABER hydrogels prolong the delivery of chemically distinct drugs with BAs (shown in blue) compared with unmodified K₂. c, Release of tavaborole from hydrogels at 0.25 mg ml⁻¹ showed that SABER hydrogels are compatible with benzoxaborole-containing drugs (indicated in red). d, SHA-K₂ was the only peptide able to significantly slow the release of ganfeborole loaded at 0.5 mg ml⁻¹ in vitro. e, Changing the drug-to-peptide molar ratio does not significantly impact the release rate over 24 h until approaching a 1:1 ratio. f, Plotting the cummulative release after 24 h revealed that the release of ganfeborole from unmodified K₂ did not significantly vary as a function of drug loading under a 1:1 peptide-to-drug molar ratio. g, Adding 3-aminophenylboronic acid to 1V209 did not affect the compound’s biological activity as determined by a colorimetric Toll-like receptor 7 activity assay performed using a reporter cell line. h,i, In vitro release of 1V209-PBA (h) and 1V209 (i) shows that the phenylboronic acid motif is necessary for SABER hydrogels to extend the release of the drug. The release of 1V209-PBA from Cat-K₂ was identical to the unmodified K₂ hydrogel. Data points indicate mean ± s.d. (n = 3).

To systematically explore how the structure of the BA influences the release rate, we synthesized six fluorescein compounds modified to contain structurally similar aryl BAs (Extended Data Fig. 3a). All the BA-modified compounds exhibited dramatically slower release from SHA-K₂ compared with unmodified fluorescein (Extended Data Fig. 3b). We observed a more than twofold difference in the linear release rate between the fastest and slowest releasing BAs (Extended Data Fig. 3c,d). These results demonstrate that single atom changes to the BA structure can have significant, and potentially predictable, effects on release rate (additional discussion in SI-4).

Although BAs have become important moieties in medicinal chemistry²⁷, most drugs do not contain a BA. To expand potential applications of the SABER platform, we examined the release of a BA-containing analogue (1V209-PBA) of the small-molecule Toll-like receptor 7 agonist (1V209), which does not have a BA. 1V209-PBA retained similar Toll-like receptor 7 agonism to the unmodified drug (Fig. 2g). In vitro release studies revealed that 1V209-PBA exhibited slower release from nitroCat-K₂ and SHA-K₂ than Cat-K₂ and K₂ hydrogels (Fig. 2h). In the absence of the BA, no altered release kinetics were observed (Fig. 2i). These results underscore the functional importance of the modifications of MDPs with nitroCat or SHA, which succeed at extending drug release where catechols fail. Modifying drugs with BAs may also allow for the programming of release kinetics from SABER hydrogels, as the release rate can be tuned by altering a BA motif (Extended Data Fig. 3d). As adding a BA to an existing drug generates a new molecular entity, requiring renewed safety and efficacy evaluation, this approach may be most beneficial for experimental drugs whose pharmacokinetic profiles hinder their clinical advancement.

In vivo release of therapeutics from SABER hydrogels

Bortezomib has been previously studied in the context of drug delivery⁴². To evaluate the controlled release of bortezomib from SABER hydrogels in vivo, mice were subcutaneously injected with 700 ng of bortezomib alone or loaded in 50 μl of MDP hydrogel. High circulating bortezomib concentrations have been linked to toxicity in patients⁴³. Mice receiving bortezomib loaded in nitroCat-K₂ and SHA-K₂ experienced a greater than 3-fold reduction maximum blood concentration compared with bortezomib alone or in K₂ (Fig. 3a,b). Other hydrogels used to deliver bortezomib have seen high burst release and relatively high maximum circulating concentrations that may increase the risk of adverse effects⁴⁴. Through a dose de-escalation study, we found that a dose of 175 ng of bortezomib alone yielded a maximum concentration of 21.2 ± 3.1 ng ml⁻¹, which is statistically similar to the peak concentration in mice dosed with a 5-fold higher bortezomib dose (700 ng) formulated with nitroCat-K₂ or SHA-K₂ (Extended Data Fig. 4a,b). At this peak-matched drug dose, the blood concentration of the drug reaches parity with the EC₅₀ (half maximal effective concentration) of the compound towards multiple myeloma cells by 336 h, while concentrations were 2.7-fold higher in mice that received bortezomib in nitroCat-K₂ and SHA-K₂ (ref. 45) (Fig. 3c). Mice dosed with bortezomib loaded in nitroCat-K₂ and SHA-K₂ hydrogels had a 2.9-fold and 1.6-fold increased area under the curve (AUC), which is used to measure drug exposure, than the peak-matched control, respectively. Clinical studies have shown that the efficacy of bortezomib treatment for multiple myeloma is associated with AUC and not the maximum drug concentration⁴⁶. Thus, SABER hydrogels may improve the therapeutic benefit of bortezomib by increasing drug exposure while reducing side effects by decreasing the maximum circulating concentration.

Fig. 3 |. SABER peptides prolong the systemic release and local retention of small-molecule drugs in vivo.

a, Pharmacokinetic profile of 700 ng of bortezomib administered subcutaneously without a hydrogel or loaded in 10 mg ml⁻¹ K₂, nitroCat-K₂ or SHA-K₂ (n = 5 for SHA-K₂, otherwise n = 4). b, Delivering bortezomib from nitroCat-K₂ and SHA-K₂ hydrogels significantly reduces the maximum circulating concentration (C_max) of the drug compared with bortezomib alone or bortezomib loaded in K₂ (n = 5 for SHA-K₂, otherwise n = 4). c, Bortezomib without a hydrogel must be administered at a 5-fold lower dose (175 ng) to match the peak drug concentration achieved from delivering 700 ng of bortezomib from nitroCat-K₂ and SHA-K₂ (n = 5 for SHA-K₂, otherwise n = 4). d, Mass spectrometry imaging analysis of bortezomib at the injection site shows that MDP hydrogels retain higher local concentrations of bortezomib. Darker spots in tissues are hydrogels, which suppress bortezomib ionization. e, Pixel intensities over the 1 mm² of tissue with the highest bortezomib concentration at 1 day (top), 7 days (middle) and 21 days (bottom). f, At early time points, the mean pixel intensity in nitroCat-K₂ is the highest of all the groups, while SHA-K₂ is the only group statistically superior to bortezomib alone at 21 days. g, Pharmacokinetic profile of 600 (~1:1 molar ratio of drug to peptide) and 60 μg (~1:10 molar ratio of drug to peptide) of ganfeborole administered alone or in a 20 mg ml⁻¹ SHA-K₂ hydrogel shows that the time above the EC₅₀ is improved by increasing drug loading (n = 4). h, Non-compartmental pharmacokinetic analysis of the groups dosed with 600 μg revealed that the SHA-K₂ substantially improved the drug exposure (AUC) and half-life (t_1/2) of ganfeborole while reducing the maximum drug concentration (C_max). All data points are shown as mean ± s.e.m.

We then investigated the persistence of the drug at the site of injection, which could be beneficial for some applications, such as treating solid tumours⁴⁷. Mass spectrometry imaging of tissue sections from the injection site showed that mice receiving bortezomib loaded in MDPs have higher levels of signal attributable to bortezomib (Extended Data Fig. 4c,d), compared with mice receiving the drug alone at all time points (Fig. 3d and Extended Data Fig. 5a,b). We observed that at 1 and 7 days post-injection, the tissue surrounding nitroCat-K₂ hydrogels had the highest bortezomib concentrations, but at 21 days, only SHA-K₂ resulted in statistically superior local drug concentrations compared with the control (Fig. 3e,f). These data suggest that the duration of local release can be controlled using different BA dynamic covalent bonding motifs. While the exact concentration and duration of local bortezomib exposure needed to treat any individual tumour remain unknown⁴⁵, successful treatment necessitates prolonged high bortezomib exposure at the cancer site⁴⁸.

To demonstrate the compatibility of the SABER platform with multiple drugs and disease applications, we evaluated the in vivo release of the tuberculosis drug candidate ganfeborole. Tuberculosis treatment typically requires months of daily oral drug dosing and is fraught with low patient adherence^49,50. Improving patient adherence to tuberculosis treatment has motivated the field to identify long-acting injectable formulations². Consistent with the in vitro results, 50 μl of 10 mg ml⁻¹ K₂ and nitroCat-K₂ gels loaded with 75 μg of ganfeborole were unable to prolong drug release in vivo. SHA-K₂ gels maintained ganfeborole concentrations above the EC₅₀ over 4-fold longer than other groups⁵¹ (Extended Data Fig. 6). Since ganfeborole is well tolerated in humans at high doses⁵¹, we sought to extend the duration over the EC₅₀ by increasing the peptide concentration, volume of hydrogel administered and drug loading. Mice were subcutaneously injected with 200 μl of 20 mg ml⁻¹ SHA-K₂ hydrogels loaded with 600 μg and 60 μg of ganfeborole. At both drug concentrations, loading ganfeborole into SHA-K₂ significantly prolonged the release of the drug compared with the control (Fig. 3g,h). Mice that received 600 μg of ganfeborole in SHA-K₂ had circulating ganfeborole concentrations above the EC₅₀ for over 508 h compared with under 79 h in mice that received the drug alone. These data highlight that the duration of release from the SABER platform can be modulated by altering the peptide concentration and the amount of drug loaded in the gels. Non-compartmental pharmacokinetic analysis of the 600 μg dose revealed that SHA-K₂ improved the AUC by 2.8-fold and increased the half-life of ganfeborole by 6.5-fold (Fig. 3h). Although these results were promising, K₂ hydrogels were observed to cause some injection-site inflammation. The strongly cationic K₂ MDP is known to cause a localized innate immune response^20,52, which may be well suited to cancer applications but is less ideal for conditions such as tuberculosis. We hypothesized that changing the MDP used in SABER could reduce the immune response to the material.

SABER peptide modularity for improved tuberculosis treatment

To demonstrate the SABER platform’s compatibility with various self-assembling peptides and to develop a less inflammatory alternative to SHA-K₂, we changed the base MDP in our SHA-modified SABER hydrogel from K₂ to the sequence E₂(SL)₆E₂ (called ‘E₂’), a negatively charged peptide (Fig. 4a and Extended Data Fig. 7a,b). In contrast to positively charged MDPs, such as K₂, negatively charged MDPs cause minimal inflammation in vivo⁵². SHA-modified E₂ (SHA-E₂) self-assembled into an antiparallel β-sheet secondary structure, indicated by Fourier transform infrared spectroscopy and circular dichroism spectroscopy (Fig. 4b,c). SHA-E₂ also formed hydrogels with similar rheological properties to E₂ (Extended Data Fig. 7c) and were shear-thinning and self-healing (Fig. 4d). Cryo-TEM images revealed the presence of thin twisted ribbon nanofibres (Fig. 4e). Histology of skin tissue collected from mice after subcutaneous hydrogel injection revealed that SHA-E₂ produces substantially less inflammation than SHA-K₂ (Extended Data Fig. 7d). These data demonstrate that the biological properties of SABER hydrogels can be modulated by altering the peptide sequence.

Fig. 4 |. Characterization of negatively charged SABER hydrogels and their use to treat a mouse model of acute tuberculosis.

a, Chemical structure of SHA-E₂. b, Fourier transform infrared spectra of SHA-E₂ show that the antiparallel β-sheet secondary structure of the peptide remains similar to unmodified E₂. c, Circular dichroism spectra of both E₂ and SHA-E₂ indicate the presence of β-sheet self-assembly. d, Rheological testing shows that both E₂ and SHA-E₂ hydrogels are shear-thinning and recover their viscoelastic properties within 10 min after being subjected to 200% strain for 1 min. G′ is represented by the solid lines and G″ is representated by the dotted lines. e, Cryo-TEM images of SHA-E₂ peptides in solution showing self-assembled nanofibres. f, SHA-E₂ significantly prolongs the release of ganfeborole compared with E₂ in vitro when loaded at 0.5 mg ml⁻¹. Data points indicate mean ± s.d. (n = 3). g, Pharmacokinetic profile after dosing 600 μg of ganfeborole shows that SHA-E₂ prolongs the release of the drug in vivo compared with unmodified E₂. Data points indicate mean (n = 5) ± s.e.m. h, A schematic illustrating the experimental design for the mouse model of acute tuberculosis. i, The colony-forming unit (CFU) counts from the lungs of mice with tuberculosis over 14 days show that a single injection of 600 μg ganfeborole in SHA-E₂ outperforms the same dose injected without the gel or the same total mass of drug given over 10 oral doses. Data for tuberculosis experiments are presented as the mean ± s.d. (n = 4). Panel h created in BioRender. McHugh, K. J. (2025) https://biorender.com/cdfkxjj.

In vitro, SHA-E₂ significantly slowed the release of ganfeborole compared with E₂ (Fig. 4f) and performed similarly to SHA-K₂ under the same conditions (Extended Data Fig. 7e). An in vivo pharmacokinetic study using 200 μl of 20 mg ml⁻¹ SHA-E₂ loaded with 600 μg of ganfeborole similarly showed that SHA-E₂ prolongs drug release compared with the unmodified E₂ hydrogel. This SHA-E₂ formulation maintained a circulating ganfeborole concentration above the EC₅₀ for more than 400 h and enhanced the half-life of the compound by over 10-fold compared with ganfeborole alone (Fig. 4g and Extended Data Fig. 7f).

We then investigated the use of this platform to treat acute tuberculosis. Three days after aerosol infection with Mycobacterium tuberculosis, mice were treated with 600 μg of ganfeborole delivered as a single subcutaneous bolus without a gel or in SHA-E₂. Since ganfeborole is given orally in clinical trials, an additional group was treated orally with the same total mass of drug administered across 10 doses over 14 days (Fig. 4h). Mice given SHA-E₂ without any drug rapidly showed signs of uncontrolled bacterial infection (Fig. 4i). The single injection of drug without hydrogel reduced colony-forming units in lungs of infected mice by 0.95 and 1.55 log₁₀ at 7 and 14 days, respectively. Ten oral doses led to a 1.67 log₁₀ reduction in bacterial burden after 7 days and 3.32 log₁₀ after 14 days compared with the vehicle control. A single injection of ganfeborole loaded in SHA-E₂ significantly enhanced the efficacy of the treatment compared with all groups and resulted in a 2.58 log₁₀ and 4.12 log₁₀ reduction in bacterial lung burden compared with the vehicle control at 7 and 14 days, respectively. No other injectable drug delivery system, to our knowledge, has achieved similar results in tuberculosis with a single therapeutic agent in a single injection. This nearly 10-fold bacterial burden reduction compared oral dosing suggests that SABER hydrogels could simultaneously reduce the necessary dosing frequency and enhance the efficacy of ganfeborole compared with oral dosing strategies currently used in the clinic.

Prolonged local delivery of BA-modified IgG from SABER

Although biotherapeutics, such as monoclonal antibodies, have gained popularity owing to their high specificity and potency, poor pharmacokinetics and/or adverse effects can limit their clinical translation⁵³. We hypothesized that modifying antibodies with a BA motif would enable SABER hydrogels to release antibodies over long periods of time in vivo (Fig. 5a). A model fluorescent immunoglobulin G (IgG) antibody was modified with either a low number of BAs (IgG low, 2.4 BAs per antibody) or a high number of BAs (IgG high, 11.4 BAs per antibody). Fluorescence recovery after photobleaching experiments demonstrated that both SHA-K₂ and SHA-E₂ reduce the diffusion of BA-modified IgG within the gels (Fig. 5b,c). We additionally found that the number of BA motifs per IgG had a minimal impact on IgG diffusion within the range tested (Extended Data Fig. 8a,b; additional discussion on SI-6).

Fig. 5 |. Local delivery of a BA-labelled model IgG antibody from SABER hydrogels.

a, A schematic illustrating dynamic covalent bonds forming between BA motifs on labelled IgG molecules and SABER hydrogels to slow down transport within the hydrogel. b,c, The normalized fluorescence intensity recovery after photobleaching a region of SHA-K₂ (b) or SHA-E₂ (c) hydrogels loaded with fluorescently labelled IgG with 2.4 BAs per IgG (IgG low) and 11.4 BAs per IgG (IgG high). SHA modification of MDPs slowed down the transport of IgG within the hydrogel matrix compared with unmodified MDPs. Data are presented as the mean ± s.d. (n = 3). d, Representative in vivo fluorescence images of mice subcutaneously injected with negatively and positively charged SABER hydrogels loaded with fluorescent IgG low and IgG high at different time points. e,f, Quantification of changes in fluorescence intensity at the injection site over time in mice receiving BA-labelled IgG shows that SHA-K₂ (e) and SHA-E₂ (f) hydrogels prolong the release of the antibodies to more than 28 and 56 days, respectively. Data points indicate mean (n = 4) ± s.e.m. Panel a created in BioRender. Mchugh, K. (2025) https://biorender.com/xzyfqnz.

Fluorescence imaging was used to evaluate the ability of SABER hydrogels to slow the release of BA-modified antibodies in vivo (Fig. 5d). Without a hydrogel, IgG injections were rapidly cleared. K₂ moderately improved the retention of IgG despite the absence of dynamic covalent bonding (Fig. 5e). In the case of IgG low, using SHA-K₂ instead of K₂ increased the half-life of the payload from 2.2 to 5.9 days (Extended Data Fig. 8c). All groups using K₂ as a base peptide had minimal (<5%) IgG remaining after 4 weeks. In comparison, SHA-E₂ extended retention of modified IgG to 8 weeks (Fig. 5f). Using SHA-E₂ to deliver IgG low resulted in a 24-fold increase in the half-life compared with the antibody alone (Extended Data Fig. 8c). Increasing labelling of BAs per antibody led to modest but statistically significant improvements in payload retention (Extended Data Fig. 8c). These data suggest that minimal levels of BA labelling are required to improve local antibody retention and that SHA-E₂ is superior to SHA-K₂ in releasing BA-labelled antibodies. These gels achieve longer local subcutaneous retention of antibody compared with several injectable hydrogels that successfully extend survival in preclinical cancer immunotherapy studies^54,55.

SABER enables days-long glycemic control in diabetic mice

The high clinical burden of basal insulin therapy necessitates innovative insulin delivery systems². Boronate ester chemistry has been investigated for creating glucose-responsive insulin delivery systems, but this responsiveness is undesirable for basal insulin delivery, which aims to provide long-term normalization of insulin levels^56–58. The boronate ester association constant between SHA and phenylboronic acid is >4,000-fold higher than the reported value for the glucose–phenylboronic acid interaction³¹ (Extended Data Fig. 1d). Therefore, we hypothesized that SHA-E₂ could be formulated as a glucose-insensitive delivery system for phenylboronic acid-modified insulin (insulin-PBA; Extended Data Fig. 9a,b). In vitro release assays showed that BA modification was necessary to delay release (Extended Data Fig. 9c) and that SHA-E₂ significantly slowed the release of insulin-PBA at all glucose concentrations compared with E₂ (Fig. 6a). No significant difference in insulin release over 24 h was observed across the glucose range tested (Fig. 6b).

Fig. 6 |. Delivery of phenylboronic acid-modified insulin from SHA-E₂ maintains normoglycemia for up to 144 h in a mouse model of type 1 diabetes.

a, In vitro release of 0.5 mg ml⁻¹ of insulin-PBA loaded in E₂ or SHA-E₂ in media with 0 mg dl⁻¹, 100 mg dl⁻¹ or 250 mg dl⁻¹ of glucose. Data plotted as the mean ± s.d. (n = 3). b, The cumulative 24 h in vitro release of insulin-PBA from SHA-E₂ reveals that the hydrogel is not likely glucose responsive over healthy blood glucose concentrations. Error bars represent ±s.d. from the mean (n = 3). NS indicates that differences between the groups are not statistically significant. c, Diabetic mice treated with a subcutaneous bolus of unmodified insulin without a hydrogel return to hyperglycemic blood glucose levels less than 4 h after injection, while mice administered 3 IU, 6 IU or 12 IU of insulin-PBA within SHA-E₂ remain below 240 mg dl⁻¹. d, The duration over which insulin-PBA delivered in SHA-E₂ can maintain normoglycemia in diabetic mice increases with the loading of insulin-PBA in the gel. e, At 4 h, 24 h and 72 h after administration, SHA-E₂ loaded with insulin-PBA results in significantly lower blood glucose levels than a single bolus of 3 IU of insulin. At 168 h, both 6 IU and 12 IU insulin-PBA gels are statistically superior to the 3 IU bolus control, even though mice are hyperglycemic. Data shown as mean (n = 5) ± s.d. f, Retreating mice with 6 IU of insulin-PBA loaded in SHA-E₂ 6 weeks after the initial treatment yielded statistically similar control of blood glucose levels. The blue region of all blood glucose graphs represents healthy blood glucose levels in humans (80–180 mg dl⁻¹), the grey regions represent mild hyperglycemia (>240 mg dl⁻¹) and mild hypoglycemia (54–80 mg dl⁻¹), and white regions represent critically low or high blood glucose levels. All blood glucose data points in panels c, d and f show mean (n = 5) ± s.e.m.

We then evaluated whether a single injection of insulin-PBA loaded in SHA-E₂ could maintain normoglycemia in diabetic mice for multiple days. Mice received one subcutaneous injection of SHA-E₂ hydrogel loaded with 3 IU, 6 IU or 12 IU of insulin-PBA or 3 IU of unmodified insulin without a hydrogel. Immediately after administration, the blood glucose concentration in all mice dropped below 150 mg dl⁻¹. Mice that received insulin without a hydrogel became hyperglycemic again within 4 h (Fig. 6c). The average blood glucose level in mice receiving 3 IU of insulin-PBA in the SHA-E₂ remained below 240 mg dl⁻¹ for 72 h (Fig. 6d), effectively controlling blood glucose levels 18-fold longer than 3 IU of unmodified insulin. Mice that received gels loaded with 6 IU and 12 IU had blood glucose levels below 180 mg ml⁻¹ for 96 h and 144 h, respectively (Fig. 6d). The 6 IU and 12 IU insulin-PBA gel treatment groups achieved significantly lower blood glucose levels than the insulin-only control group between 4 h and 168 h (Fig. 6e). The blood glucose levels of mice treated with a second dose of 6 IU of insulin-PBA in an SHA-E₂ hydrogel were not statistically significantly different from those obtained with the first dose at all time points measured (Fig. 6f and Extended Data Fig. 9d). These data establish that insulin-PBA loaded in SHA-E₂ can achieve normoglycemia for up to 144 h in a single injection, a 36-fold improvement over the maximum tolerated single dose of unmodified insulin without a hydrogel. This system controls the blood glucose levels of mice, without glucose sensitivity, for ~10-fold longer than next-generation long-acting insulin formulations similar to those that are commercially available for basal insulin therapy⁵⁹.

Conclusion

Modifying MDPs to enable the formation of boronate esters with therapeutics significantly enhances their capacity to control the release of BA-containing payloads while leaving nanostructure and rheological properties intact. Using nitrocatechol and SHA, BA dynamic covalent association motifs not previously used for drug delivery applications, we show that SABER hydrogels can be used to enhance the pharmacokinetics and/or biodistribution of BA-containing compounds. The SABER drug delivery platform is compatible with numerous modalities including synthetically BA-modified drugs to enable drug release and activity for weeks instead of days in vivo. We demonstrate the utility of this system in several clinically relevant models, including better control of tuberculosis with a single injection than with repeated oral dosing and achieving normoglycemia in diabetic mice for 144 h after a single injection. Screening the release kinetics of more BAs may enable the prediction and rational programming of drug release from SABER. We show that SABER works with different peptide hydrogels, and we expect this approach to work with other hydrogels or nanomaterial drug delivery platforms. These results establish the SABER hydrogel platform as a highly flexible drug delivery system with the potential to improve treatment of a wide variety of diseases.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41565-025-01981-6.

Methods

Supplementary methods

Methodology for chemical synthesis, Fourier transform infrared spectroscopy, circular dichroism spectroscopy, rheology, cryo-TEM, UV–vis, binding assays, TLR7 activation assays and fluorescence recovery after photobleaching is included in Supplementary Information (SI-7).

Hydrogel preparation

MDP hydrogels were prepared by dissolving the MDP at 2× the final desired peptide concentration (20 mg ml⁻¹ or 40 mg ml⁻¹) in MilliQ water. The desired BA-containing small molecule or BA-labelled biologic was dissolved at twice the final concentration in 2× Hanks’ balanced salt solution (HBSS). The pH of the 2× HBSS solution was adjusted to pH 7.5–8.5 to assist in the dissolution of the BA-containing small molecules. These two stock solutions were then mixed in a 1:1 ratio to yield a hydrogel with 1× HBSS, and the final 1× peptide and payload concentrations (10 mg ml⁻¹ or 20 mg ml⁻¹). The pH of the gel was adjusted to pH 7–8 as necessary with microlitre additions of NaOH or HCl and the gel was vortexed to ensure homogeneity.

In vitro release assays

Plate reader release assay with fluorescein compounds.

Hydrogels at a final peptide concentration of 10 mg ml⁻¹ (1:4 mass ratio of SHA-K₂ to K₂) were loaded with 312.5 μM fluorescein or fluorescein modified with one of the six BAs were prepared as previously described to achieve a BA-to-SHA molar ratio of 1:4. To start the experiment, 50 μl of each hydrogel was plated in triplicate onto a custom 3D-printed 48-well plate designed previously for plate reader release assays²⁰, and the gels were allowed to recover for 10 min. The gels were then covered gently with 450 μl of 1× phosphate-buffered saline (PBS), sealed with a transparent plate sticker to prevent evaporation, and orbitally agitated at 100 rpm at 37 °C. At the desired time points, 400 μl of the release media was replaced. The amount of the compounds released from the gels was determined using a microplate reader (Tecan) to read the fluorescence at 490 nm/525 nm excitation/emission (Ex/Em). The percentage of payload released was determined by using a compound-matched standard curve.

UPLC release assay.

All BA-containing small molecules were purchased from Ambeed, except ganfeborole, which was purchased from MedChemExpress. Hydrogels, at a final peptide concentration of 10 mg ml⁻¹ and loaded with 500 μg ml⁻¹ of the small molecule, were prepared as previously described earlier in the methods and equilibrated in the dark at ambient temperature overnight. Tavaborole (250 μg ml⁻¹), 1V209 (50 μg ml⁻¹) and 1V209-PBA (50 μg ml⁻¹) gels were prepared with lower concentrations owing to solubility limitations in the hydrogels. The mass of drug loaded into the hydrogel in all cases is below a drug-to-peptide molar loading of 1:2 (1:3 for tavaborole, 1:3.6 for ixazomib, 1:3.84 for bortezomib, 1:2.6 for ganfeborole and 1:48 for 1V209-PBA). In addition, hydrogels with 1V209 and 1V209-PBA were prepared with 25% DMSO to improve the solubility of these hydrophobic small molecules. Gels (50 μl) were then plated in triplicate on a custom 3D-printed 48-well plate as described in previous publications. The gels were then allowed to recover for 10 min before 450 μl of 1× PBS was gently added to each well to start the experiment. The plate was then covered with a 96-well plate cover sticker to prevent evaporation and orbitally agitated at 100 rpm at 37 °C. At the desired sampling time points, 400 μl of the release media was removed and refreshed. The amount of compound released was then assessed by UPLC using a Poroshell C18 (Agilent Technologies) UPLC column and adjusting for the volume remaining in the well after each sampling (Supplementary Table 1). The percentage of released payload was determined by dividing the cumulative mass of drug released by the total loaded in the hydrogel.

To ensure that all the drug loaded in each gel could be released, 20 μl of gels loaded with drug were prepared as previously described. The gels were incubated at room temperature for 1 h and then dissolved in 480 μl of 0.1 M HCl with 20 mM boric acid. The amount of drug recovered from the gels was then determined by UPLC and compared with a drug solution that was prepared without peptide.

Insulin release assay.

Unmodified insulin and insulin-PBA were loaded at 500 μg ml⁻¹ in hydrogels containing 10 mg ml⁻¹ peptide as described above using 1× PBS instead of HBSS to avoid the addition of glucose. Each hydrogel (50 μl) was then pipetted into low protein bind Eppendorf tubes in triplicate and centrifuged (1,000 × g) to settle the gel in the bottom of the tube. After the gels equilibrated for 10 min, 450 μl of 1× PBS with 0 mg dl⁻¹, 100 mg dl⁻¹ or 250 mg dl⁻¹ of glucose was gently pipetted on top of each gel and incubated at 37 °C. At the desired sampling time point, 200 μl of release media was removed and replaced. The concentration of insulin and insulin-PBA in the release medium was quantified by UPLC using a BEH C18 column (Waters Corp.) accounting for the volume remaining in the tube after each sampling (Supplementary Table 1). The percentage of released insulin and insulin-PBA was determined by dividing the cumulative mass of the protein released by the total loaded in the hydrogel.

Pharmacokinetic assays

Bortezomib.

All animal work in this study was performed in compliance with a Rice University IACUC-approved protocol. Animals in all experiments were free fed 5V5R PicoLab Select Rodent diet and were randomly assigned to experimental groups. Data collection and analysis were not performed blind to the conditions of the experiments. Data points and animals were only excluded from analysis if technical errors with data collection (that is, mass spectrometer malfunction as observed by failure of quality control checks) prevented the collection of reliable data.

Hydrogels loaded with 700 ng of bortezomib were prepared as previously described at a final peptide concentration of 10 mg ml⁻¹. Female BALB/c mice (6–8 weeks old; 16–18 g) obtained from The Jackson Laboratory were injected subcutaneously with 50 μl of drug-loaded hydrogel or bortezomib dissolved in HBSS. Blood (10 μl) was collected using untreated Safe-T-Fill plastic hematocrit capillaries (RAM Scientific) and spotted on a Whatman 903 Proteinsaver Card (Cytiva). Blood cards were dried overnight protected from light and then stored with desiccant at 4 °C for up to 1 week or −80 °C for up to 1 month. Extractions were performed following a previously published protocol for bortezomib⁶⁰. The compound was extracted off the blood card by using a 1/8-inch hole puncher to remove the centre of each blood spot, corresponding to 2.4 μl of blood. The blood spots were then submerged in 40 μl of methanol containing 1 ng ml⁻¹ apatinib (APExBIO) as the internal standard. Extractions proceeded for 1 h at ambient temperature on an orbital shaker set to 100 rpm. The extraction solution was then removed and diluted 1:2 with water and analysed by liquid chromatography-mass spectrometery (LC-MS).

Multiple reaction monitoring (MRM) LC-MS analysis of bortezomib was carried out on an Agilent 6470B triple quadrupole (QqQ) mass spectrometer using apatinib as an internal standard. The MS system was interfaced to an Agilent 1290 Infinity ii LC system through an Agilent Jet Spray (AJS) electrospray ionization (ESI) source that was operated in the positive mode. Separations were carried out using a Water’s ACUITY Premier HHS T3 100 mm × 2.1 mm ID, 1.8 μm column that was operated at 0.4 ml min⁻¹. Mobile phase A (MPA) was 0.1% formic acid in water, and mobile phase B (MPB) was 0.1% formic acid in acetonitrile. Initial LC conditions were 20%B up to 80%B over 4.0 min. The column was flushed at 80%B for 2.0 min and then re-equilibrated at 20%B for 2 min before the next injection.

The AJS source and MS data acquisition settings were optimized to meet the needed sensitivity requirements. In brief, the AJS source conditions were as follows: gas temp., 320 °C; gas flow, 8 l min⁻¹; nebulizer gas pressure, 25 psi; sheath gas temp, 400 °C; sheath gas flow, 11 l min⁻¹, capillary voltage, 3,800 V. QqQ MRM data acquisition settings were as follows: cell accelerator voltage, 5 V; dwell time, 50 ms. The MRM ion transitions for bortezomib and apatinib are shown in Supplementary Information (Supplementary Table 2).

Ganfeborole.

Pharmacokinetic analysis of ganfeborole was performed nearly identically to the procedure described above for bortezomib with minor alterations. Hydrogels loaded with 60 μg or 600 μg ganfeborole were prepared as previously described at a final peptide concentration of 10 mg ml⁻¹ or 20 mg ml⁻¹ depending on the experiment. Female BALB/c mice (6–8 weeks old; 16–18 g) obtained from The Jackson Laboratory were injected subcutaneously with the drug-loaded hydrogel or ganfeborole dissolved in PBS. Blood (10 μl) was collected using an untreated plastic capillary and spotted on a Whatman 903 Proteinsaver Card. Blood cards were dried overnight protected from light and then stored with desiccant at 4 °C for up to 1 week or −80 °C for up to 1 month. The compound was extracted off the blood card by using a 1/8-inch hole puncher to remove the centre of each blood spot, corresponding to 2.4 μl of blood. The punches were then submerged in 40 μl of 90:10 methanol:water containing 5 ng ml⁻¹ apatinib as an internal standard. The submerged blood spots were extracted at 37 °C for 1 h while being shaken at 100 rpm. The extraction solution was then removed and diluted 1:2 with water and analysed by LC-MS.

MRM LC-MS analysis of ganfeborole was carried out on an Agilent 6470B QqQ mass spectrometer using apatinib as an internal standard. The MS system was interfaced to an Agilent 1290 Infinity ii LC system through an AJS ESI source that was operated in the positive mode. Separations were carried out using a Water’s ACUITY Premier HHS T3 100 mm × 2.1 mm ID, 1.8 μm column that was operated at 0.4 ml min⁻¹. MPA was 0.1% formic acid in water, and MPB was 0.1% formic acid in acetonitrile. Initial LC conditions were 10%B up to 60%B over 5.0 min from 5.0 min to 5.5 min the gradient was increased to 80%B. The column was flushed at 80%B for 2.5 min and then re-equilibrated at 10%B for 3 min before the next injection.

The AJS source and MS data acquisition settings were optimized to meet the needed sensitivity requirements. In brief, the AJS source conditions were as follows: gas temp, 320 °C; gas flow, 5 l min⁻¹; nebulizer gas pressure, 30 psi; sheath gas temp, 400 °C; sheath gas flow, 12 l min⁻¹; capillary voltage, 3,800 V. QqQ MRM data acquisition settings were as follows: cell accelerator voltage, 5 V; dwell time, 100 ms; MS1 (Q1) resolution, unit; MS2 (Q3) resolution, wide. The MRM ion transitions for ganfeborole and apatinib are shown in Supplementary Information (Supplementary Table 3). The data collected was corrected for dilutions during sample preparation and plotted in GraphPad Prism 10. Pharmacokinetic parameters were determined by noncompartmental analysis using Ubiquity in RStudio⁶¹.

Mass spectrometry imaging

Female BALB/c mice (6–8 weeks old; 16–18 g) obtained from The Jackson Laboratory were subcutaneously injected with 700 ng of bortezomib loaded in 50 μl of 1× HBSS, K₂, nitroCat-K₂ or SHA-K₂. Hydrogels were prepared as previously described. Mice were then euthanized 1, 7 and 21 days post administration and the tissue at the injection sites was collected. The tissue was cryopreserved with liquid nitrogen without the use of any fixative. Mouse skin was cross sectioned at 12 μm thickness using a Thermo NX50 cryostat (Epredia) and collected onto standard plus slides. Optical images of the slides were acquired at 4,800 dpi using an Epson Perfection V600 Photo flatbed document scanner (Epson US). Sections were coated with 10 mg ml⁻¹ α-cyano-4-hydroxycinnamic acid matrix in 70% ACN, 0.1% TFA using an HTX M5 Robotic Reagent Sprayer (HTX Technologies, LLC) as follows: 4 passes, nozzle temperature of 75 °C, flow rate of 100 μl min⁻¹, track speed of 1,200 mm min⁻¹, track spacing of 3 mm, an HH track pattern, and a nozzle height of 40 mm. Serial sections were collected for H&E staining and were digitized using a Hamamatsu NanoZoomerSQ digital slide scanner (Hamamatsu Photonics).

Mass spectrometry images were acquired at 50 μm resolution in positive ion mode using a Bruker timsTOF fleX QTOF mass spectrometer (Bruker Daltonics) over the m/z range 50–1,000 with a summation of 700 shots per pixel. Instrument tuning was as follows: a funnel 1 RF of 100.0 Vpp, a funnel 2 RF of 150.0 Vpp, a multipole RF of 200.0 Vpp, a collision energy of 5.0 eV, a collision RF of 600.0 Vpp, a quadrupole ion energy of 5.0 eV, a transfer time of 60.0 μs and a pre-pulse storage of 6.0 μs. Bortezomib was detected in tissue as the in-source generated fragment at m/z 226.09 that was confirmed through matrix-assisted laser desorption/ionization (MALDI) analysis of a standard.

Image files were loaded into SCiLS Lab Pro 2023b (Bruker Daltonics) for visualization and analysis. Data were root mean square normalized. Hematoxylin and eosin (H&E) images were annotated using Hamamatsu NDP.view2 software for regions of gels in the sections. These annotations were transferred to SCiLS to create regions of interest corresponding to areas of gel and non-gel within each sample. Intensities of bortezomib from each pixel outside the gels were exported to a .CSV file using the SCiLS Lab API for R. Dark spots in the mass spectrometry images are a result of the hydrogel suppressing the ionization of bortezomib; thus, only drug in the tissue outside the hydrogels could be quantified (Extended Data Figs. 4d and 5a). To account for the different sizes of tissue collected, the pixel intensities over the 1 mm² (400 pixels) of tissue with the highest bortezomib signal were analysed for statical comparisons in GraphPad Prism 10.

In vivo fluorescence animal imaging IgG release assay

Sterile 10 mg ml⁻¹ MDP hydrogels containing 2 mg ml⁻¹ of AZ647-labelled IgG-PBA were prepared 1 day before the assay and stored at 4 °C overnight protected from light. SKH1-Elite mice (6–8 weeks old) were obtained from Charles River Laboratories, and tissue background autofluorescence was quantified at Ex/Em 640 nm/700 nm for each mouse using an In Vivo Imaging System (IVIS) small animal imager (PerkinElmer). Background fluorescence intensity was subtracted from all subsequent images. Mice were subcutaneously injected bilaterally in the flank with 50 μl of sample or control and imaged longitudinally. The percentage of material released from the injection site was calculated by drawing equally sized region-of-interest rectangles around the injection sites and dividing the observed total radiant efficiency in that region by the maximum total radiant efficiency measured for that injection on the first day of the experiment. The data were fit to first-order exponential equation in GraphPad Prism 10 using a least-squares regression to model the release from the injection site. The error for parameters extracted from the model is reported as 95% confidence intervals.

Mouse model of acute tuberculosis

M. tuberculosis H37Rv was grown in 7H9 broth supplemented with 10% oleic acid, albumin, dextrose, catalase (OADC; Difco Laboratories) and 0.05% Tween 80 (Sigma-Aldrich) before infection. A log phase growth (acute infection) model infection was used for this experiment. In brief, 6-week-old female BALB/c mice obtained from Charles River Laboratories were infected with a log-phase culture of M. tuberculosis (optical density at 600 nm of approximately 1.0) using an inhalation exposure system (Glas-Col), aiming to implant approximately 4.5 log₁₀ colony-forming units in the lungs. After infection, mice were randomized into treatment groups (12 mice per group). Untreated mice were euthanized at the initiation of treatment to determine pretreatment colony-forming unit counts. Mice were initiated on treatment 3 days post-infection with one of the four different regimens. The vehicle control group received a 200 μl subcutaneous injection of 20 mg ml⁻¹ SHA-E₂ without any drug. The second cohort of mice received a single 200 μl subcutaneous injection containing 600 μg of ganfeborole (MedChemExpress). An additional cohort of mice received 10 doses of 60 μg of ganfeborole administered by oral gavage once per day over the course of 14 days (5 doses for every 7 days). The experimental group received a 200 μl subcutaneous injection of 20 mg ml⁻¹ SHA-E₂ loaded with 600 μg of ganfeborole. Lung colony-forming unit counts were assessed after 1 and 2 weeks of treatment by performing quantitative cultures of lung homogenates on OADC-enriched 7H11 agar (Difco Laboratories).

Mouse type 1 diabetes model

For in vivo studies with insulin, male C57BL/6J mice aged 6–8 weeks were purchased from Charles River Laboratories. After 1 week of acclimatation, mice were treated with 50 mg kg⁻¹ of streptozotocin (MilliporeSigma) for 5 consecutive days. Streptozotocin was dissolved at a concentration of 7.5 mg ml⁻¹ in pH 4.5 sodium citrate buffer immediately before injection. Then, the mice’s blood glucose levels and weights were monitored. For blood glucose measurement, a drop of blood was collected from the tail and tested using a OneTouch UltraMini glucometer (LifeScan). Mice with blood glucose levels that exceeded 350 mg dl⁻¹ for several sequential days were deemed diabetic and suitable for inclusion in the study.

Insulin-PBA was prepared from commercially purchased insulin (35 μg IU⁻¹) as described in the chemical synthesis section of the methods. SHA-E₂ hydrogels were prepared with insulin-PBA with minor modifications to the protocol previously described. Insulin-PBA was dissolved at 8.4 mg ml⁻¹ for 12 IU and 6 IU gels and 4.2 mg ml⁻¹ for 3 IU gels in 2× PBS. PBS was used instead of HBSS to avoid injecting diabetic mice with glucose, which is part of the HBSS buffer. These solutions were then mixed 1:1 with a stock solution of 20 mg ml⁻¹ of SHA-E₂ dissolved in MQ water to form insulin-PBA-loaded hydrogels. Diabetic mice were then injected with 100 μl of 12 IU gels, 50 μl of 6 IU and 3 IU gels or with 3 IU of unmodified insulin dissolved in 50 μl of 1× PBS. Right after treatment administration, blood glucose levels were measured every hour until 4 h and every 2 h until 10 h. Blood glucose levels for mice that received insulin-PBA gel injections were subsequently monitored daily for 9 days. Mice that had blood glucose levels below 240 mg dl⁻¹ were considered normoglycemic.

Statistical analysis

Multiple group comparisons were calculated by one- or two-way analysis of variance with Tukey’s multiple comparisons test. Statistical calculations were performed in GraphPad Prism 10. Statistical significance is denoted with asterisks as follows: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Error bars and reported error represent standard error of the mean (s.e.m.) unless otherwise specified as standard deviation (s.d.) or as a 95% confidence interval. No statistical methods were used to pre-determine sample sizes, but our sample sizes are similar to those reported in previous publications^24,55,58. Data distribution was assumed to be normal, but this was not formally tested.

Extended Data

Extended Data Fig. 1 |. Design and development of SABER peptides.

a) Catechol oxidation and degradation occurred rapidly, appearing as changes in the UV-Vis spectrum of dopamine (DOPA) over time in pH 7.4 PBS. The increase in the baseline absorbance indicates increased turbidity of the sample as insoluble oxidation products form black precipitates in solution. b) The UV-Vis spectrum of the nitrocatechol 4-nitrodopamine (nDOPA) shows that the molecule is stable over the course of 15 d. c) Longitudinal analysis of the UV-Vis spectrum salicylhydroxamic acid (SHA) demonstrates that the molecule is largely stable with minor changes observed after 11 d. d) Boronate ester equilibrium constants between chemically distinct boronic acids (BAs) and the dynamic covalent association motifs alizarin red S (ARS), nDOPA, SHA, and DOPA. For all BAs, SHA forms the strongest boronate ester interactions. Equilibrium constants are presented as the mean value of three replicates ± SD (see Supplementary Fig. S1). e-f) Chemical structures, mass spectra, and UPLC chromatograms of e) Cat-K₂, f) nitroCat-K₂, g) SHA-K₂, and h) K₂.

Extended Data Fig. 2 |. In vitro release and drug stability.

a) Cumulative release of bortezomib from SHA-K₂ is not significantly altered within physiologically relevant concentrations of glucose. b) The total amount of bortezomib released after 24 h was not statistically different between glucose concentrations. Data are presented as the mean value of three replicates ± SD. c) Gels loaded with drug were dissolved in water and run on UPLC to ensure that all of the drug could be recovered. Bars with ‘#’ reached 100% release in the release experiments shown in Fig. 2 and cumulative release from those experiments after 24 h is shown here. These data demonstrate that all the drug loaded into the gels can be released. Data are presented as the mean value of three replicates ± SD. d) The change in UPLC retention time of ganfeborole and ganfeborole released from Cat-K₂ in UPLC suggests that the hydrogel is degrading the drug. The new peak at 3.2 min seen in Cat-K₂ + ganfeborole is also observed when the drug is reacted with 280 mM hydrogen peroxide, suggesting that this new peak corresponds to oxidized ganfeborole. e) Electrospray ionization mass spectrometry of ganfeborole alone contains the expected mass of the drug at 258.1 m/z. Ganfeborole released from Cat-K₂ has a peak at 248.1 m/z, (f) corresponding to a loss of a boron atom due to oxidative deboronation. g) The stability of ganfeborole loaded in K₂, Cat-K₂, nitroCat-K₂, and SHA-K₂ quantified by UPLC shows that the majority of the drug remains stable over the course of two weeks in all hydrogels except for Cat-K₂, which rapidly degrades the compound. Data are presented as the mean value of three replicates ± SD. h) SHA-K₂ without drug and the hydrogel maximally loaded (1:1 drug-to-peptide molar ratio) have the same β-sheet secondary structure as determined by a minimum at 220 nm seen by circular dichroism, suggesting that drug loading does not perturb peptide self-assembly.

Extended Data Fig. 3 |. In vitro release of fluorescent BAs.

a) Structures of fluorescein and the six fluorescein BA conjugates synthesized. b) In vitro release of the fluorescent BAs from SHA-K₂ over the course of 336 h. All BA-modified compounds demonstrate significantly lower release than unmodified fluorescein. c) A rescaled release curve focusing on the six BA-modified compounds. The higher burst release observed in compound 6 may be due to the presence of a small amount of unconjugated fluorescein. The compounds demonstrate zero-order ‘linear’ release from 72–336 h. d) The linear release regime for each sample was fit to a line to calculate the release rate. All fits resulted in an R² > 0.97. Compound 5 demonstrated the fastest release rate, which is over two-fold faster than the slowest releasing compounds (2, 4, and 6). These data demonstrate that minor changes to BA structure can impact the release rate from SABER hydrogels. All data is presented as the mean value of three replicates ± SD.

Extended Data Fig. 4 |. In vivo release of bortezomib.

a) Pharmacokinetics of decreasing bortezomib doses administered as subcutaneous boluses without hydrogel. Data is presented as the mean value of four replicates ± SEM. b) The maximum circulating concentration (C_max) of bortezomib decreases linearly with the initial dose. Dotted lines represent the C_max for 700 ng of bortezomib delivered from nitroCat-K₂ (blue) and SHA-K₂ (purple), indicating that a bolus bortezomib dose of 175 ng yields the same C_max as these hydrogel formulations loaded with 5-fold more drug. c) Chemical structure of the bortezomib fragment observed in mass spectrometry imaging. d) K₂, nitroCat-K₂, and SHA-K₂ hydrogels loaded with 700 ng of bortezomib imaged by mass spectrometry imaging in vitro show very little bortezomib signal, suggesting that MDP peptides suppress the ionization of bortezomib within the gels. This ionization suppression results in dark spots in mass spectrometry images at the location of the hydrogels.

Extended Data Fig. 5 |. Hematoxylin & eosin staining and mass spectrometry imaging of injection site tissues.

a) Tissue sections from mice that received subcutaneous injections of bortezomib alone or in a hydrogel at 1-, 7-, and 21-days stained with hematoxylin & eosin. Large dark purple sections in K₂, nitroCat-K₂, and SHA-K₂ are the hydrogels in the skin samples. b) Mass spectrometry imaging of the same tissue samples stained with hematoxylin & eosin shows that bortezomib signal does not significantly overlap with that from heme (616.178 m/z), illustrating that the drug observed in the tissue is not in circulation but in the local environment of the injection site.

Extended Data Fig. 6 |. Pharmacokinetics of ganfeborole in 50 μL of 10 mg/mL hydrogels.

Longitudinal concentrations of ganfeborole in the blood of mice after a single subcutaneous injection of 75 μg of the drug loaded into 50 μL hydrogels or PBS. Data presented as the mean value of four replicates ± SEM. Injections of ganfeborole without a hydrogel, ganfeborole in K₂, and ganfeborole in nitroCat-K₂ all led to the rapid release of the drug. SHA-K₂ hydrogels were able to retain ganfeborole concentrations over the EC₅₀ for more than 200 h.

Extended Data Fig. 7 |. In vitro characterization of SHA-E₂ and in vivo release of ganfeborole.

a) Chemical structure of SHA-E₂ with the mass spectrum and UPLC chromatogram of the material confirming the identity and purity of the peptide. b) Mass spectrum and UPLC chromatogram of the unmodified E₂ peptide. c) Frequency sweep collected by oscillatory rheology shows that SHA-E₂ hydrogels are more frequency dependent than unmodified E₂ and form slightly weaker gels, as indicated by a reduction in the difference between the storage modulus (G′), indicated by filled circles and solid lines, and the loss modulus (G″), indicated by open circles and dotted lines. d) Representative histological sections of SHA-K₂ and SHA-E₂ gels excised three days after subcutaneous injection. Sections were stained with hematoxylin & eosin (left two images) or Masson’s trichrome (right four images) stains. Black 750 μm scale bar applies to the left four full sized images. Histological analysis shows that SHA-K₂ gels swell significantly and are heavily infiltrated by cells, consistent with an inflammatory response. SHA-E₂ remains minimally infiltrated by cells, has less collagen deposition on the border of the gel, and does not swell in size. e) Cumulative release of ganfeborole after 24 h from SHA-K₂ and SHA-E₂ are statistically similar, suggesting that changing the peptide used in SABER hydrogels does not compromise its ability to control the release of BA-containing small molecules. Data is presented as the mean of three replicates ± SD. f) Pharmacokinetic parameters extracted by performing a non-compartmental analysis on the in vivo release of ganfeborole from SHA-E₂ show that using the SABER hydrogel improves drug exposure (AUC), half-life (t_1/2) and reduces the maximum circulating concentration (C_max) of the compound. Pharmacokinetic parameters are presented as the mean of n=4–5 replicates ± SD.

Extended Data Fig. 8 |. Local release of BA-modified IgG.

a) Representative confocal images illustrating fluorescence recovery and photobleaching (FRAP) experiments. Fluorescence recovery was quantified by monitoring the return of fluorescence signal in the bleached region over 10 min using a 640 nm excitation laser. The white scale bar in the bottom left represents 40 μm. b) FRAP data (n=3 for each group) was fit to a first-order exponential equation to extract the FRAP half-time (t_1/2) and the mobile fraction (M_f). Loading BA-modified IgG in SABER hydrogels reduced the M_f and increased the t_1/2, suggesting that the rate of payload diffusion in these samples is slower. We observed minimal differences between IgG labeled with 11.4 BAs per antibody (IgG high) and 2.4 BAs per antibody (IgG low), demonstrating that the degree of labeling may not play a large role in controlling the rate of diffusion. All data were well-modeled by the first-order exponential equation and had R² values above 0.95 except for SHA-E₂ + IgG high (denoted with an asterisk), which was poorly fit by this model and thus extracted parameters may not accurately describe the data. c) In vivo release data of IgG with low and high degrees of BA modeling was modeled with a first-order exponential equation to determine the half-life (t_1/2) and burst release from the site of injection. All SABER hydrogels had a moderate burst release of 20% but significantly extended the t_1/2 of the antibody at the injection site. All fits adequately modeled the data (R² > 0.95). All numerical data in this figure is presented as the mean value of four replicates ± 95% confidence interval.

Extended Data Fig. 9 |. Basal insulin delivery from SABER hydrogels.

a) Mass spectrum of BA-modified insulin (insulin-PBA) showing that a single phenylboronic acid was added to insulin. b) The UPLC chromatogram of the synthesized insulin-PBA confirmed the purity of the material. c) In vitro release of unmodified insulin from SHA-E₂ and E₂ illustrates that the phenylboronic acid modification is necessary for the SABER peptide to achieve the delay in the release of the payload observed in Fig. 6a. Data presented as the mean value of three replicates ± SD. d) First 10 h of the initial treatment of diabetic mice with 6 IU of insulin-PBA in SHA-E₂ plotted with a repeat treatment with the same formulation in the same mice 6 weeks later. The repeat dose resulted in statistically similar blood glucose levels to the initial dose at all time points. Data points for blood glucose measurements are the mean value of five replicates ± SEM.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41565-025-01981-6.

Data availability

The authors declare that all the data supporting the findings of this research are available within the article and its Supplementary Information. Source data are available upon reasonable request. Source data are provided with this paper.