Intracellular GRP78-Directed Delivery of Rapamycin by Biomolecular Condensates of Hydra-Elastin-like Polypeptides

Abstract

Rapamycin (Rapa) is a potent inhibitor of the mammalian target of rapamycin complex 1 (mTORC1) with possible applications in multiple diseases; however, it and its analogues exhibit low solubility, variable bioavailability, and dose-limiting side effects. To engineer a long-release carrier, we employ Rapa’s cognate receptor (FKBP12) to modulate its solubility, rate of release, and cellular uptake. To target its internalization into cancer cells under stress with an unfolded protein response (UPR), we use an L-peptide that binds cell-surface glucose-regulated protein 78 (GRP78). Herein, the L-peptide was fused to five FKBP domains linked by an elastin-like polypeptide (ELP) selected to form a biomolecular condensate depot at body temperature. This novel GRP78-targeted carrier (L-5FV) was characterized by UV–vis spectrophotometry, dynamic light scattering (DLS), surface plasmon resonance (SPR), and dialysis under sink conditions to assess its thermosensitivity, particle assembly, binding kinetics to both Rapa and GRP78, and drug release, respectively. Functional delivery of cellular internalization and mTORC1 inhibition were confirmed using fluorescence microscopy and Western blot in dose- and time-dependent manners in a breast cancer cell line, BT-474. Both targeted and untargeted formulations are phase-separated at physiological temperatures and exhibit nanomolar affinity for FKBP12 and Rapa. Notably, L-5FV demonstrated a more significant cellular association and inhibition of p-rpS6, a mechanistic target of mTORC1 activity. This report provides insight into how to construct long-release, molecularly targeted drug carriers with applications in UPR-active cancers.

Graphical Abstract

INTRODUCTION

Rapa or sirolimus (Rapamune) is a U.S. Food and Drug Administration (FDA)-approved medication for kidney transplantation.¹ Everolimus, a rapalogue, is also FDA-approved for a spectrum of clinical applications, including hormone receptor-positive breast cancer.^2–7 Despite their therapeutic promise, they are limited by challenges, such as aqueous solubility that reduces oral administration,⁸ off-target effects,⁹ and resistance.¹⁰ Their macrolide structures are allosteric inhibitors of FK506-Binding Protein 12 (FKBP12).¹¹ Once bound, they directly inhibit mTORC1, the mechanistic target of Rapa.¹² Inhibition of mTORC1 reduces the level of activation of S6 kinase 1 (S6K1), a major downstream target of mTORC1, and the level of downstream phosphorylation of ribosomal protein S6 (rpS6). p-rpS6 activates 5′ terminal oligopyrimidine (5′TOP) mRNA, which modulates translational efficiency relevant to cancer progression.^13–15 Rapa signaling is mechanistically linked to cell cycle arrest and cancer attenuation.¹⁶ Our group established a new method for Rapa delivery by forming a complex with its cognate receptor, FKBP12, which both solubilizes the drug and allows its control through fusion proteins.^17,18 Compared to traditional encapsulation of therapeutic cargos using physical entrapment or covalent linkage in carrier systems,^19–22 the FKBP delivery approach combines drug-specific, high-affinity drug interaction with sustained release.^23,24

To control FKBP/Rapa, our approach links them with repetitive ELPs. ELPs are genetically engineered, biocompatible, and thermosensitive polymers with tunable properties.^25,26 They are expressed as pentapeptide sequences, (VPGXG)_n, at which ‘n’ controls the molecular weight and the guest residue ‘X’ determines the polymeric solubility. Both elements define their reversible physiological transition temperature, T_t, and subsequently their pharmacokinetics. Below T_t, they are soluble; however, above T_t, they coacervate into larger microparticles. Coacervation both enables their purification from bacterial lysates and slows the rate of absorption from a point of subcutaneous injection.^18,27,28 We first engineered one copy of FKBP into a micellar structure composed of ELPs with hydrophilic serine and hydrophobic isoleucine blocks, which was called FSI. FSI assembled micelles above 25 °C that entrap the drug in their FKBP; however, these micelles required intravenous administration and were only stable in vitro and in vivo for about a day.^17,29,30 To enhance the formulation stability for subcutaneous administration, we next fused two copies of FKBP, spaced by 192 repeats of an ELP selected to remain soluble at body temperature, which we called FAF.¹⁸ This biheaded FKBP fusion increased the subcutaneous bioavailability by 8-fold compared to FSI, showed less toxicity than the free drug, and suppressed autoimmune inflammation in the lacrimal gland of a mouse model of Sjögren’s disease. Compared with FSI, FAF yielded more Rapa encapsulation efficiency with greater stability. To further sustain Rapa delivery through a long-acting formulation, we next developed a fusion of five FKBP domains, linked via (VPGVG)₂₄, which we called 5FV.²⁴ The valine ELP set phase separation to occur at 21 °C for a 300 μM depot concentration. A single injection of this Rapa-loaded carrier suppressed inflammation in the lacrimal gland of a nonobese diabetic mice model using 16-fold less Rapa relative to soluble Rapa. This carrier demonstrated the potential for FKBP-ELP encapsulation and release. Simultaneously, we developed a soluble version of the five “Hydra” FKBP domains linked by (VPGAG)₂₄, known as 5FA,³¹ which was evaluated in combination with a panel of GRP78-targeting peptides to enhance the potency of the formulation through target-mediated cellular uptake. We compared four GRP78-targeting ligands, identified from various sources,^32–34 and investigated their effect on cellular internalization and p-rpS6 inhibition.³¹ In vitro assays demonstrated that the L-peptide (RLLDTNRPLLPY) linked to 5FA (L-5A) best promoted the cellular uptake and suppression of p-rpS6 signaling. Since 5FA does not phase-separate at body temperature, these were not designed to extend release from a depot; however, they demonstrated how the FKBP-ELP system could be made more potent through incorporation of GRP78-targeted ligands.

With the advent of antibody–drug conjugates, molecularly targeted cancer therapies have become useful to combat resistance and nonspecific side effects.^35–37 They facilitate selective and controlled release through direct molecular interactions, and that can be combined with different encapsulation strategies.^38–40 GRP78 is among the broadly upregulated cancer hallmarks.⁴¹ In the same family as heat shock protein 70 (HSP70), GRP78 normally resides in the endoplasmic reticulum (ER). There, it acts as a critical chaperone mediating protein folding.⁴² GRP78 is also a master regulator of the UPR such that in unstressed cells, it binds ER transmembrane proteins (inositol-requiring enzyme 1 “IRE-1”, activating transcription factor 6 “ATF-6”, and PKR-like ER kinase “PERK”) and maintains them in an inactive form. Under stress, GRP78 dissociates from them, triggering the UPR, and at the same time, GRP78 binds to unfolded proteins in the ER to maintain cellular homeostasis under stress.⁴³ In stressed cells, including cancer cells which commonly exhibit intrinsic and extrinsic stress, a subfraction of GRP78 translocates to the cell surface, where it assumes a new function acting as a coreceptor that is susceptible to binding and internalization of cargo.^44,45 Being a multifaceted contributor in cancer progression, metastasis, and further adaptive chemoresistance, GRP78 has been identified as a promising, druggable target in cancer formulation development.^46–52

For the first time here, we engineer an FKBP-ELP that is both molecularly targeted to GRP78 and phase-separates at physiological temperatures. To do so, we linked the L-peptide to 5FV and showed that it can entrap and release the drug for extended periods. It enhances the delivery of functional Rapa into a breast cancer cell line and a spheroid culture. We compare the physicochemical characterization, cellular association, and dose/time-dependent inhibition of p-rpS6, which suggests that L-5FV is active at lower concentrations and faster than for the free drug or 5FV-bound drug.

MATERIALS AND METHODS

ELP Cloning, Production, Purification, and Concentration Measurement

The genetically engineered protein polymer named 5FV was designed by restriction enzyme cloning in the pET-25b(+) vector as reported previously.²⁴ For ligating the L-peptide (RLLDTNRPLLPY) attached to the VPGA linker into the 5FV plasmid, complementary phosphorylated oligonucleotide strands (5′ forward sequence TATGCGCCTGCTGGATACCAACCGCCCGCTGCTGCCGTATGTTCCGGGCGCTGG) and (5′ reverse sequence TACCAGCGCCCGGAACATACGGCAGCAGCGGGCGGTTGGTATCCAGCAGGCGCA) were synthesized by Genewiz Inc. (South Plainfield, NJ, USA), mixed in a 1:1 molar ratio, heated at 95 °C for 2 min, and annealed at room temperature for 2 h. The 5FV vector was incubated with NdeI and alkaline phosphatase (New England Biolabs, Ipswich, MA, USA #R0111, #M0371) for digestion and self-ligation inhibition, respectively. The L-peptide oligo with sticky ends was then inserted using T4 DNA ligase (New England Biolabs #M0202) into the 5FV linear plasmid at a 3:1 molar ratio, and this insertion was incubated overnight at 4 °C with gentle shaking. This ligation reaction was transformed by heat shock at 42.5 °C for 45 s in chemically competent TOP10 cells (Thermo Fisher Scientific, Waltham, MA, USA #C404010) which were selected on Miller’s LB agar plate (MilliporeSigma, Burlington, MA, USA #L3522), supplemented with 1× carbenicillin (GoldBio, Olivette, MO, USA #C-103). Colonies were picked up individually in 6 mL of Miller’s LB medium and grown overnight on a shaker at 37 °C, and DNA was extracted by a QIAprep Spin Miniprep kit (Qiagen, Germantown, MD, USA #27106). The purity of the extracted DNA was confirmed by a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific) where the 260/280 absorbance ratio was observed between 1.8 and 1.9. The insertion of the L-encoding oligonucleotide into the 5FV vector was confirmed by Sanger sequencing (Genewiz Inc.) using the T7 promoter and terminator primers. The plasmid size was confirmed by agarose gel electrophoresis using a 1 kb DNA ladder (New England Biolabs #N0550). The L-5FV vector was digested by BciVI (New England Biolabs #R0596) and the separated bands were matched with simulated digestion bands by SnapGene software (version 4.0.8).

To produce batches of this L-5FV sequence, the ligated DNA was transformed in ClearColi BL21 (DE3) cells (Lucigen, Middleton, WI, USA #60810) by electroporation. Stocks of verified transformed colonies were incubated in 100 mL of the autoclaved LB medium, spiked by 1× carbenicillin overnight at 37 °C on a shaker incubator at 250 rpm. The next day, this culture was diluted to 6 × 1 L flasks of the medium and incubated similarly. When the optical density at 600 nm (OD₆₀₀) reached 0.6 after few hours, flasks were supplemented by 1 M in 500 μL of isopropyl β-d-1-thiogalactopyranoside (GoldBio #I2481) overnight at 20 °C under shaking. Cells were collected under 4000 rpm for 15 min at 4 °C and resuspended in 15 mL of cold Dulbecco’s phosphate-buffered saline (PBS, Genesee Scientific, El Cajon, CA, USA #25–508) for protein solubilization through probe-tip sonication. Soluble ELPs were purified by a phase separation. The lysate was centrifuged at 10,000 rpm for 20 min at 4 °C (first cold cycle). The supernatant was isolated and incubated at 37 °C and spiked to 2 M sodium chloride (NaCl, VWR, Radnor, PA, USA #470302) to induce coacervation. ELPs were collected at 4000 rpm for 20 min at 37 °C as pellets (first warm cycle). At least 3 subsequent cold–warm cycles were performed until ELPs were sufficiently purified.

To calculate the ELP concentration (C) in molarity (M), ELPs were diluted at 6 M guanidine-HCl solution (Thermo Fisher Scientific #24115) and incubated for 5 min, and the NanoDrop 2000 spectrophotometer was used at 280 and 350 nm. The Beer–Lambert law was employed using the equation as follows

$$C_{\text{ELP}}=\frac{A_{280}-A_{350}}{\epsilon \times l}$$ (1)

where ε, the molar extinction coefficient (50,475 M⁻¹cm⁻¹), was estimated by the ExPASy ProtParam tool.⁵³ A is the measured absorbance (280 and 350 nm) and l is the light path length (0.1 cm).

To maintain ELP sterility, purified proteins were filtered by an Acrodisc Mustang E membrane filter (Cytiva Life Sciences, Marlborough, MA, USA #MSTG25E3) and subsequently tested for TLR4-activating contaminants, including residual lipopolysaccharide (LPS) from the bacterial expression process, using the HEK-Blue assay (InvivoGen, San Diego, CA, USA #rep-lps2), following the supplier’s instructions. If the endotoxin level was detected above 0.1 EU/mL, formulations were passed through the Detoxi-Gel endotoxin removing gel (Thermo Fisher Scientific #20339, #20344), as specified by the manufacturer’s guidance.

Characterization of ELP Purity and Identity

Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was used to assess ELP purity. 0.8 mg/mL was denatured in 1:9 2-mercaptoethanol (Thermo Fisher Scientific #31350010) to the Laemmli buffer (Bio-Rad Laboratories, Hercules, CA, USA #1610747) at 95 °C for 5 min, loaded onto a 4–20% SDS-PAGE gel (Bio-Rad Laboratories #4561095) along with a protein ladder (Bio-Rad Laboratories #1610375), stained with a G-250 Coomassie stain (Bio-Rad Laboratories #1610786), and visualized by a ChemiDocTM imaging system (Bio-Rad Laboratories).

The exact molecular weight of purified ELPs was confirmed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS, Bruker Inc., Billerica, MA, USA). 5 mg/mL ELPs were prepared in a 1:1 volume ratio to 2-mercaptoethanol, denatured at 95 °C for 10 min, and diluted 1:10 in 10 mg/mL 2,6 dihydroxyacetophenone solution that includes 50% acetonitrile and 0.1% formic acid (Thermo Fisher Scientific #022927). Less than 1 μL of this denatured mixture was spotted onto a MALDI-TOF target plate and air-dried. Crystallized samples were then processed by a RapifleX MALDI-TOF system (Bruker Inc., version 4.0).

Characterization of ELP Thermoresponsibility

A UV–vis spectrophotometer (Beckman Coulter, Brea, CA, USA) was used to ensure the ELP thermoresponsive behavior. Purified proteins were diluted in PBS and loaded into quartz cuvettes. Settings were adjusted to an analytical wavelength of 350 nm and a gradient of 1 °C of heat starting from ~15 to 85 °C. Readings were recorded every 0.3 °C using DU800 spectrophotometer software (version 3.0). ELP T_t was defined at the maximum first derivative of each curve and then fitted using the following equation

$$T_t=b-m\left[\text{lo}{\text{g}}_{10}\left(C_{\text{ELP}}\right)\right]$$ (2)

where b is the y-intercept corresponding to 1 μM ELPs in PBS, and m is the slope representing the T_t per 10-fold change in the ELP concentration (Table 1).

Table 1.

Summary of the Biophysical Characterization of Expressed 5FV Carriers

label	molecular sequencea	expected MWb [kDa]	observed MW by MALDI-TOFc [kDa]	observed MW by SEC-MALSd [kDa] (mean ± SD)	Rapa per ELP, SEC-MALSe (mean ± SD)	Rapa per ELP, HPLC (mean ± SD)	Rhf at 10 °C [nm] (mean ± SD)	Rh at 37 °C [nm] (mean ± SD)	Tt,bg [°C] (Cl)	Tt,m [°C/log(μM)] (Cl)	Tt at 150 μM [°C]
5FV	MG-[FKBP-(VPGVG)24]4-FKBP	98.4	98.3	96.1 ± 1.2	N/A	N/A	4.6 ± 0.3	1384.9 ± 22.9	50.0 [46.8–53.2]	14.3 [11.1–17.5]	18.9
L-5FV	MRLLDTNRPLLPYVPGAGMG-[FKBP-(VPGVG)24]4-FKBP	100.5	100.4	102.9 ± 0.8	N/A	N/A	4.7 ± 0.7	1867.1 ± 320.1	47.2 [34.8–59.5]	18.3 [6.1–30.6]	7.4
5FV-Rapa	MG-[FKBP-(VPGVG)24]4-FKBP/(Rapa)n	98.4	98.1	99.6 ± 1.8	3.8 ± 2.4	4 ± 0.9	4.8 ± 1.3	886.9 ± 784.7	43.9 [40.1–47.8]	7.6 [3.8–11.5]	27.4
L-5FV-Rapa	MRLLDTNRPLLPYVPGAGMG-[FKBP-(VPGVG)24]4-FKBP/(Rapa)n	100.5	100.3	107.8 ± 2.7	5.4 ± 3.1	4.5 ± 0.2	5.6 ± 1.9	1655.1 ± 1084.7	44.2 [43–45.4]	7.8 [6.6–9.1]	27.2

^aFKBP encodes for VQVETISPGDGRTFPKRGQTCVVHYTGMLEDGKKFDSSRDRNKPFKFMLGKQEVIRGWEEGVAQMSVGQRAKLTISPDYAYGATGHPGIIPPHATLVFDVELLKLEG. L-peptide shown in bold, RLLDTNRPLLPY.

^bThe expected MW is calculated based on the sequence-confirmed open reading frame, with N-terminal methionine excluded for 5FV.

^cThe MW was quantified using MALDI-TOF showing the mass of the free fusion protein.

^dThe MW was quantified using SEC-MALS showing solution complexation with Rapa.

^eRapa encapsulation was quantified by SEC-MALS (Rapa MW = 0.91 kDa) and HPLC, respectively.

^fR_h refers to the hydrodynamic radius determined by DLS.

^gThe intercept b and slope m for T_t fit parameters were estimated from the concentration–temperature phase diagram (Figure 2), (eq 2), [95% confidence interval (CI)].

Characterization of ELP Particles

DLS (DynaPro Plate Reader, Wyatt Technology, Santa Barbara, CA, USA) was used to verify the association between the ELP mass–radius distribution and its thermosensitivity under different temperatures (10 and 37 °C), and size exclusion chromatography with multiangle light scattering (SEC-MALS) was used to confirm the absolute molar mass of ELP/drug complexes and observe formulation aggregation.

For DLS, 60 μL of 25 μM ELPs were diluted in PBS, filtered by a sterile 0.02 μm Whatman Anotop filter (MilliporeSigma #WHA68091002), plated on a 384-well black plate with a clear bottom, and covered by 15 μL of mineral oil to prevent evaporation at 37 °C. A 1 mg/mL sample of filtered bovine serum albumin (BSA, MilliporeSigma #A3294) was used as a positive control. The plate was centrifuged for 1 min at 1000 rpm to eliminate bubbles, which may scatter light. Readings were processed by Dynamics V7 software (Wyatt Technology, version 7.1.9.3).

For SEC-MALS, 10 μM filtered ELPs (Cytiva Life Sciences #4612) were prepared in 100 μL of PBS, injected onto a size exclusion Shodex protein KW-803 column, 8.0 mm ID × 300 mm, (Showa Denko America, New York, NY, USA), which was pre-equilibrated with PBS at 0.5 mL/min. ELP elution was detected by three different instruments, a UV detector “SYC- LC-1200” (Agilent Technologies, Santa Clara, CA, USA), a multiangle light-scattering detector “DAWN HELEOS” (Wyatt Technology), and a differential refractometer “Optilab rEX” (Wyatt Technology). Filtered 5 mg/mL BSA was used as a protein control. Data were analyzed by ASTRA 6.1 software (Wyatt Technology).

Characterization of the Binding Affinity between FKBP Sites and Rapa

SPR (Biacore T200 instrument, Cytiva Life Sciences) was used to study the comparative binding affinity between FKBP domains and Rapa of both 5FV and L-5FV. Experiments were conducted using a 0.22 μm PES membrane-filtered (Genesee Scientific #25–227) 1× running buffer of the following 10× composition: 100 mM 4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid (HEPES, MilliporeSigma #H3375) buffer, supplemented by 1.5 M NaCl, 30 mM ethylenediaminetetraacetic acid (EDTA, MilliporeSigma #E6758), and 0.5% Tween-20 (Santa Cruz Biotechnology, Santa Cruz, Burlington, MA, USA #sc-29113), adjusted to pH 7.4. The buffer was further supplemented by 2% dimethyl sulfoxide (DMSO, MilliporeSigma #472301). 5FV and L-5FV (10 μg/mL) were immobilized at pH 4.5 on the surface of a Series S Sensor Chip CM5 chip (Cytiva Life Sciences #BR100530) by amine coupling chemistry using 1-ethyl-(3-(dimethylamino)propyl)-carbodiimide hydrochloride (EDC) and N-hydroxysuccinimide (NHS) (Cytiva Life Sciences #BR100050). Residual sites of dextran on these immobilized channels and the control flow channel were blocked with 1 M ethanolamine hydrochloride (Cytiva Life Sciences #BR100050). The control flow cell was used as a reference subtraction. Dilutions of Rapa were injected over the chip in 1× running buffer. Remaining Rapa was removed from the chip by 0.01% sodium dodecyl sulfate (SDS, Thermo Fisher Scientific #15525017) at a flow rate of 30 μL/min for a contact time of 30 s, followed by chip regeneration using 1 M sodium hydroxide (NaOH, MilliporeSigma #221465) at the same flow and contact rate. Sensograms were analyzed by Biacore T200 evaluation software (Cytiva Life Sciences, version 3.2.1). Response units (RUs) for the Rapa concentration were plotted against time. The RU for bound Rapa was normalized to the reference control. The association rate constant (K_on), dissociation rate constant (K_off), and equilibrium dissociation constant (K_d) were fitted. K_d was calculated as the ratio of the dissociation to association rate constants as follows

$$K_d=\frac{k_{off}}{k_{on}}$$ (3)

Characterization of the Binding Affinity between the L-Peptide and GRP78

SPR was also used to evaluate the affinity dynamics between L-5FV and GRP78 using different experimental settings.³² For studying the GRP78/5FV carriers’ molecular interaction, a sensor chip NTA (Cytiva Life Sciences #28994951) was activated by the NHS and EDC coupling kit and a 1 min pulse of 500 μM NiCl₂ solution (Cytiva Life Sciences #28995043), and then 30 μg/mL histidine-tagged recombinant human GRP78 BiP protein (Abcam, Cambridge, MA, USA #ab78432) was immobilized on the surface by chelated nickel. The chip surface was subsequently deactivated with 1 M ethanolamine and washed with 350 mM EDTA (pH 8.3). 1× of the filtered HBS-P buffer with 150 mM NaCl, 0.05% Tween-20, and 1 mM glycine (Thermo Fisher Scientific #A13816) was used as a system running buffer (pH 7.4). ELP constructs were diluted in the running buffer. 20 mM NaOH was used for chip regeneration. Sensograms were subtracted from the reference control cell and fitted using Biacore T200 software.

Characterization of Rapa Loading and Release

For encapsulating Rapa in 5FV and L-5FV carriers, Rapa (LC Laboratories, Woburn, MA, USA #R-5000) was dissolved in pure ethanol (Fisher Scientific, Pittsburgh, Pennsylvania, USA #BP2818), and 10× molar excess of Rapa (less than 5% v/v) was loaded dropwise at 4 °C to 5FV and L-5FV formulations with continuous stirring. The reaction was continued for 3 h. Following time, the resulting mixture was centrifuged at 13,000g for 15 min at 4 °C for pelleting nonreactant Rapa. For eliminating residual ethanol and free Rapa, the supernatant was obtained and dialyzed against sterile 1× PBS at 4 °C at a 1:350 ratio of the sample to dialysate for at least 5 rounds of buffer exchange over 48–60 h. For quantifying Rapa loaded in formulations, the standard of free Rapa and aliquots of loaded carriers were injected to reverse-phase high-performance liquid chromatography (RP-HPLC, Agilent 1260, Agilent Technologies Inc.) on a C-18 column (Agilent Technologies Inc. #693970–902T) and quantified at an optical density of 280 nm. Samples were diluted in a mixture of 60:40 methanol:ultrapure water (Fisher Scientific #A452–4), while samples were eluted under a mobile phase gradient ranging from 40 to 90% of methanol to water, supplemented with 0.1% trifluoroacetic acid (MilliporeSigma #T6508). The area under the curve (AUC) of standard dilutions of known Rapa concentrations was used to generate a calibration curve from which the concentrations of encapsulated Rapa in samples were quantified. Encapsulated Rapa was also confirmed by SEC-MALS, which reflected an increase in the MW, compared to free constructs, as stated above.

For studying Rapa release, ~350–400 μM Rapa encapsulated in 5FV carriers was placed in 10 K Slide-A-Lyzer G3 dialysis cassettes (Thermo Fisher Scientific #A52971) and dialyzed at 4 and 37 °C against 1× sterile PBS (1:650 volume), supplemented by a 1× penicillin–streptomycin (Pen/Strep) solution (Corning, Corning, NY, USA #30–002-CI). Three independent aliquots were taken at 0, 7, 14, and 28 days for Rapa retention analysis using the same HPLC method. The dialysis PBS buffer was changed after the first 48 h, followed by a weekly change. Data were fitted as shown in the following equation

$$C_{\mathit{\text{Rapa}}}=C_{\mathit{\text{Rapa}}}\left(0\right)e^{-\text{ln}\left(2\right)t/t_{1/2}}$$ (4)

by which, C_Rapa, C_Rapa (0), t, and t_1/2 indicate the Rapa concentration at a given time point, the Rapa starting concentration, the time in days, and the time when Rapa releases at half initial amount (half-life), respectively.

Cell Culture

BT-474 human epithelial breast ductal carcinoma cells (American Type Culture Collection ‘ATCC’, Manassas, VA, USA #HTB-20) were cultured in a Hybri-Care medium (ATCC #46-X), reconstituted in cell-culture-grade water and supplemented with 1.5 g/L sodium bicarbonate (MilliporeSigma #SX0320) and 10% fetal bovine serum (Corning #35–011-CV). Cells were humidified under 5% CO₂ at a 37 °C incubation condition and replaced with new medium until 70–80% confluency. When confluent, they were detached by 0.25% trypsin-EDTA (Thermo Fisher Scientific #25200–056) and centrifuged, and the pellet was resuspended in a new growth medium, counted, and seeded for assaying or frozen in a complete growth medium, containing 5% DMSO, as recommended by the manufacturer.

For a three-dimensional (3D) culture or spheroids, 1 × 10⁴ of BT-474 cells were seeded in ultralow attachment 96-well plates (MilliporeSigma #CLS4515), centrifuged at 1000 rpm for 5 min, incubated immediately (5% CO₂, 37 °C), and tracked until reaching ~350 μm in diameter. The growth medium was replenished every 48 h. When treated, they were exposed to ELP treatment conditions for 2 h, washed in triplicate with PBS, fixed by 3% paraformaldehyde (Thermo Fisher Scientific #043368), stained by Hoechst (Thermo Fisher Scientific #62249), and permeabilized in 0.1% Triton X-100 (MilliporeSigma #T8787) for further ELP detection by immunofluorescence. For ELP staining, cells were blocked using 0.2 μm Acrodisc Supor PES membrane-filtered 1% BSA in PBS for 1 h at room temperature to prevent nonspecific binding, incubated with mouse IgM anti-ELP AK-1 primary antibody (1:60, Cancer Therapeutics Laboratories, Los Angeles, CA, USA) overnight at 4 °C with gentle shaking, washed three times with PBS for 15 min each, and then incubated with Alexa Fluor 488-conjugated antimouse secondary antibody (1:250, Thermo Fisher Scientific #A-11001) for 1 h at room temperature with gentle shaking, followed by triplicate washes with PBS. Immunofluorescence staining for GRP78 was performed following the same blocking and permeabilization steps, ensuring the detection of total GRP78 (1:60 for GRP78 polyclonal primary antibody from Thermo Fisher #PA1–014A and Alexa Fluor 488-conjugated antirabbit secondary antibody from Thermo Fisher #A-21206 at 1:250), followed by the same antibody incubation times and PBS washing steps as above. Stained cells were imaged by confocal microscopy (Zeiss, Thornwood, NJ, USA). Collected images were processed by FIJI software (version 2.16.0/1.54p), by which the normalized background-subtracted integrated fluorescence density per μm² of the signal in each spheroid (F_sample) was quantified as follows⁵⁴

$$F_{\mathit{\text{sample}}}=\frac{I_{\mathit{\text{spheroid}}}}{A_{\mathit{\text{spheroid}}}}-M_{\mathit{\text{background}}}$$ (5)

where I_spheroid is the integrated density within the section of the spheroid, A_spheroid is the area of the spheroid, and M_background is the mean intensity of the background per unit area (I μm⁻²). Two sections near the middle of each spheroid were averaged, and those numbers were then averaged over spheroids with each treatment.

Western Blot

5 × 10⁵ of BT-474 cells were cultured in a 6-well plate. When confluent, they were washed by fresh media and then exposed to ELPs or Rapa in a dose- and time-dependent manner as specified in assays. At the end of the assay, cells were washed three times with PBS, containing calcium and magnesium (Corning #21–030-CV) for avoiding cell detachment, spiked by a 1× protease/phosphatase inhibitor cocktail (Cell Signaling Technology, Danvers, MA, USA #5872) in RIPA buffer (Thermo Fisher Scientific #89901), incubated on ice for 1 h, and vortexed every 10 min for an additional 1 h, and the supernatant was collected following centrifugation at 13,000g for 15 min at 4 °C. Protein concentration was quantified using a BCA assay kit (Thermo Fisher Scientific #23225), by which the absorbance of the sample supernatant at 562 nm was calculated against the absorbance of known albumin concentrations. Equal amounts of protein (1–2 mg/mL in 15 μL) were denatured in 5 μL of 1:9 parts of 2-mercaptoethanol to Laemmli buffer and loaded onto a 4–20% precast SDS-PAGE gel. A Precision Plus Protein Kaleidoscope ladder was run in parallel as a molecular weight reference. Following protein size separation, the gel was transferred to a nitrocellulose membrane (Thermo Fisher Scientific #IB23001) using an iBlot2 Dry Blotting system (Fisher Scientific, Pittsburgh, Pennsylvania, USA #IB21001), blocked by 5% BSA in 1× tris-buffered saline (TBS, Bioland Scientific LLC, Paramount, California, USA #TBS01) supplemented with 0.1% Tween-20 for 1 h at room temperature on a shaker. Primary antibodies (rabbit IgG anti-GRP78-BiP from Abcam #ab21685 at 1:1000, mouse IgM anti-ELP AK-1 at 1:2000, and rabbit IgG antiphospho-S6 ribosomal protein ‘p-rpS6’ Ser235/236 from Cell Signaling Technology #2211 at 1:1000) were added to the membrane for overnight incubation at 4 °C on a shaker. Following incubation, membranes were washed three times for 15 min, incubated with secondary antibodies (antirabbit IgG HRP-linked antibody for GRP78 and p-rpS6 from Cell Signaling Technology #7074 at 1:3000 and antimouse IgM HRP-linked antibody at 1:2000 for ELPs, Fortis Life Sciences, Waltham, MA, USA #A90–101P) for 1 h at room temperature, rewashed in triplicates for 15 min, incubated for 1 min in a PierceTM ECL Western blot substrate (1:1 of the peroxide solution to the luminol enhancer solution, Thermo Fisher Scientific #32106), and imaged for chemiluminescence using an iBright FL1000 system (Thermo Fisher Scientific). The membrane was further stripped, blocked again, and plotted against GAPDH (housekeeping control protein). Primary mouse IgG anti-GAPDH (1:1000, Cell Signaling Technology #97166) and secondary antimouse IgG HRP-linked antibody (1:3000, Cell Signaling Technology #7076) were similarly applied as target protein steps. All Western blot data were obtained at least in triplicate from independent samples. All blots were imaged simultaneously under the same exposure conditions to ensure consistency and allow a direct comparison between groups. For normalized chemiluminescence quantification per replicate (N), a constant size region of interest (ROI) was quantified using ImageJ software (version 1.54f) to obtain a background-subtracted integrated density (I) for each sample. These values were normalized to the background-subtracted GAPDH integrated density as follows

$$N_{\mathit{\text{sample}}}=\frac{(I_{\mathit{\text{sample}}}-I_{\mathit{\text{background}},\mathit{\text{sample}}})}{(I_{\mathit{\text{GAPDH}}}-I_{\mathit{\text{background}},\mathit{\text{GAPDH}}})}$$ (6)

The above signal as a function of the concentration was converted to AUCs for further statistical comparison.

For comparing the change of signal intensity over time-dependent study (S_sample), the normalized signals (N_sample) were further adjusted relative to the corresponding measurement at time zero (T₀) prior to AUC calculation as follows

$$S_{\mathit{\text{sample}}}=\frac{N_{\mathit{\text{sample}}}}{N_{\mathit{\text{sample}},T0}}$$ (7)

Statistical Analysis

The statistical significance between two groups was assessed using Student’s t-test, while comparisons among three groups were performed using one-way ANOVA followed by Tukey’s post hoc test using GraphPad Prism software (versions 9.2.0 and 9.5.1). Plots shown on log-transformed axes were used to improve the data representation. Error bar representations are detailed in each figure legend.

RESULTS

Characterization of Protein Cloning and Production

A gene encoding the L-peptide linked to the amino terminus of the 5FV carrier was engineered in a modified pET-25b(+) vector and verified by Sanger DNA sequencing (Figure 1A) and restriction endonuclease digestion (Figure 1B,C). The expected restriction digestion pattern was predicted using SnapGene (Figure 1B), and the experimentally observed bands matched the predicted fragments (Figure 1C). The gene product for L-5FV was expressed in ClearColi bacteria and purified using its thermoresponsive behavior over 3–4 rounds of cooling/heating and centrifugation. The purified yield was 75 mg/L, which is like that observed for 5FV. The product purity and exact MW were verified by G-250 Coomassie SDS-PAGE (Figure 1D) and MALDI-TOF (Figure 1E,F), which was confirmed in the Rapa-loaded formulation (Supplementary Figure S1).

Figure 1.

Molecular engineering of a GRP78-targeted ELP carrier for Rapa. An oligonucleotide cassette encoding for the GRP78-binding ‘L-peptide’ was ligated into a plasmid encoding the FKBP-ELP gene known as 5FV. FKBP domains are included for their ability to noncovalently bind Rapa. (A) Sanger sequencing confirmed the in-frame ligation after the ribosome binding site (RBS) at the amino terminus of the 5FV gene, as well at the stop codon of the carboxy-terminal stop FKBP. (B) Diagnostic restriction enzyme digestion predicted the proper DNA fragments (C) upon digestion with BciVI, which recognizes 3 or 4 sites within 5FV or L-5FV, respectively. The predicted bands for 5FV are at 5.153, 1.847, and 1.527 kilobase (kb), compared to L-5FV at 3.920, 1.847, 1.527, and 1.287 kb. (D) Following purification, fusions were analyzed by SDS-PAGE, stained by Coomassie blue. The major band is located at the expected band size (Table 1), near a 100 kDa band in the ladder. Loaded carriers appear at the same MW and relative purity compared to free carriers. (E, F) The exact MW was verified by MALDI-TOF, consistent with the expected values (Table 1).

Evaluation of ELP Thermosensitivity

5FV was selected because it phase-separates between room temperature and body temperature, making it ideal for depot formation. To test whether the L-peptide or drug association affected the ELP phase behavior, concentration–temperature phase diagrams were quantified using UV–vis spectrophotometry for free and encapsulated carriers (Figure 2A–D). While the inclusion of the L-peptide lowered the T_t compared to 5FV (Figure 2E), this was eliminated upon complexation with Rapa. The slope of these lines (eq 2) enables the prediction of T_t for the depot (~150 μM), which occurs above 27 °C (Table 1). Therefore, both formulations remain in solution at room temperature, awaiting injection; however, they form a depot upon administration to the body. Below 4 μM, all formulations have T_t above 37 °C, meaning they could circulate freely in biological solutions up to this concentration and in great excess to the inhibitory concentration (IC₅₀) expected for Rapa (Figure 7B). Despite minor differences, L-peptide fusion and Rapa encapsulation appear to minimally affect T_t.

Figure 2.

Rapa association minimally affects the ELP transition temperature. Optical density (350 nm) was used to monitor formulation phase separation as a function of concentration and temperature with and without drug. (A) 5FV, (B) 5FV-Rapa, (C) L-5FV, (D) and L-5FV-Rapa, all showed a concentration-dependent cloud point. (E) The maximum first derivative with respect to temperature was defined as T_t and fitted by eq 2. Rapa entrapment slightly decreased the concentration dependence (slope) of phase separation for both 5FV and L-5FV (Table 1). Dashed lines indicate 95% CI.

Figure 7.

L-5FV-Rapa suppresses p-rpS6 more effectively than does 5FV-Rapa. Western blot was used to monitor phosphorylation of rpS6 in BT-474 exposed to free or ELP-bound Rapa. (A) A monolayer of cells was incubated for 2 h with increasing concentrations (0.1–100 nM) of Rapa in 5FV, in L-5FV, or free. Cellular lysates were assessed for p-rpS6 and GAPDH. (B) The normalized p-rpS6 signal to GAPDH was plotted (eq 6) and fitted by nonlinear regression to estimate the inhibitory concentration at 50%. The IC₅₀ values for 5FV-Rapa, L-5FV-Rapa, and free Rapa are 2.1 nM [95% CI: 0.9–5.1 nM], 0.5 nM [95% CI: 0.1–2.2 nM], and 3.3 nM [95% CI: 1.1–10.7 nM], respectively. (C) AUC across concentrations was compared by ANOVA (n = 3–4, independent samples), which demonstrates the greatest suppression of the p-rpS6 signal for L-5FV-encapsulated Rapa over the concentration (p-values shown). (D) Based on the concentration study, a 1 nM Rapa dose was selected to evaluate the kinetics of p-rpS6 suppression using the same assay (eq 7). (E) Compared to 5FV-Rapa and free Rapa, L-5FV-Rapa more rapidly blocked the phosphorylation of rpS6. (F) The AUC over all time points was compared by ANOVA (n = 3–4, independent samples), which revealed that L-5FV-Rapa gave the most potent suppression of mTORC1 (p-values shown). At this concentration, there was no significant difference between the concentration of 5FV-Rapa and that of free Rapa. This is consistent with more effective delivery by the L-5FV-Rapa carrier, which utilizes cell-surface GRP78 (Figure 8). Images used to generate this figure are in Supplementary Figure S4. Values indicate the mean ± SD.

Assessment of the Hydrodynamic Radius and Peptide Association by Light Scattering

Light scattering was used to assess particle assembly and association with the cargo in solution. DLS determines the particle size distribution based on their apparent Brownian diffusion, by which larger particles more slowly induce fluctuations in constructive and deconstructive interference of scattered light.^55,56 Thus, DLS easily detects the significant increase in size from monomeric ELPs into micrometer-sized depots in response to heating. Guided by UV–vis spectrophotometry (Figure 2), a fixed concentration (25 μM) was screened below T_t (Figure 3A) (10 °C) and above T_t (Figure 3B) at 37 °C. All 5FV-based formulations transitioned from a ~5 nm size to a ~1 to 2 μm size upon heating (Table 1), and assembly was unaffected by loading with Rapa. When assessed by phase contrast microscopy imaging, all formulations assembled rounded, amorphous droplets consistent with light-scattering data (Supplementary Figure S3). To characterize the formulation solution MW below T_t, SEC-MALS was used. Static, multiangle light scattering was able to both estimate the fraction of the formulation that was present as a monomeric species; furthermore, it also detects the association of Rapa with the complex.^57,58 Low-concentration samples (10 μM) were fractionated by size exclusion chromatography to identify the monomeric species by light scattering, the UV absorbance at 230 nm, and the refractive index. BSA was evaluated as a control, which revealed that the major species has a monomeric mass (Table 2). Likewise, 5FV and 5FV-Rapa revealed that the major species was close to the expected MW for a monomer; furthermore, an increase in the MW between 5FV and 5FV-Rapa implied the association of ~4 Rapa per molecule of 5FV, which was similar to that estimated by a calibrated RP-HPLC assay (Table 1). The addition of the L-peptide to 5FV increased the prevalence of multimeric, higher-MW aggregates, which elute at earlier time points; however, based on the refractive index and UV profiles, these remain below 20% of the formulation (Table 2). Similarly, the monomeric fraction of L-5FV revealed a Rapa-dependent increase in the MW, consistent with that identified by RP-HPLC (Table 1). From these light-scattering studies, we conclude that the L-peptide has a detectable effect on the aggregation state of 5FV below T_t; however, it does not prevent drug binding (Figure 3F) or temperature-dependent phase separation (Figure 2D,E).

Figure 3.

Light scattering reveals the association of Rapa with FKBP domains in solution. DLS was used to evaluate the formulation R_h below and above T_t (25 μM) at (A) 10 and (B) 37 °C, respectively. Below T_t, the R_h for all samples was explained by a single peak below the 10 nm range, which is consistent with their monomeric MWs (Table 1).⁵⁹ Above T_t, the ELP-based formulations assembled larger droplets (500–3000 nm range), while a BSA control did not. SEC-MALS was used to evaluate the formulations below T_t, at room temperature (10 μM). For unlabeled (C) 5FV and (D) L-5FV, these closely matched their expected MWs. As confirmation of drug entrapment, a noncovalent association with Rapa was detected via an increase in the MW for (E) 5FV and (F) L-5FV of ~3.5 and 5 kDa, respectively. (G) Trimeric, dimeric, and monomeric BSA were detectable. For 5FV, a dimeric peak was identifiable. For L-5FV, there was evidence of higher-MW aggregates; however, they accounted for less than 20% by mass. Detailed mean ± SD for the sample MW, including the detected populations, is noted below (Table 2). Population 1 is the dominant, monomeric population, while populations 2 and 3 represent multimeric species.

Table 2.

List of Population Sizes and Percentages Eluted in SEC-MALS

label	population 1 (mean ± SD) kDa, mass fraction%	population 2 (mean ± SD) kDa, mass fraction%	population 3 (mean ± SD) kDa, mass fraction%
BSA	65.8 ± 0.8 (91.3%, monomer)	140 ± 2.1 (7.7%, dimer)	252.4 ± 13.9 (1.1%, multimer)
5FV	96.1 ± 1.2 (95.4%, monomer)	221.7 ± 17.8 (4.6%, dimer)	N/A
L-5FV	102.9 ± 0.8 (97.6%, monomer)	617.8 ± 31.3 (2.1%, multimer)	1.5 × 105 ± 4.8 × 102 (0.3%, multimer)
5FV-Rapa	99.6 ± 1.8 (92%, monomer)	221.6 ± 8.9 (8%, dimer)	N/A
L-5FV-Rapa	107.8 ± 2.7 (83.7%, monomer)	266.7 ± 11.5 (9.6%, dimer)	914.6 ± 47.8 (6.7%, multimer)

Quantification of Rapa Binding and Retention

After quantifying the thermoresponsive particle size of 5FV (Figure 2, 3A,B) and their particle formation (Figure 3C–F), their affinity and rate of release were quantified by SPR (Figure 4A,B) and retention under sink dialysis (Figure 4C,D). SPR is a label-free technique for studying the molecular interactions between ligands and analytes on a gold sensor chip surface covalently coupled with FKBP proteins. The rate of increase RU under different analyte concentrations enables the estimation of the kinetic binding constant (k_on). During each dissociation step, the kinetic rate constant (k_off) is fitted. The chip is regenerated between each analyte concentration, which enables the calculation of the equilibrium binding constant (K_d).^60–62 To solubilize Rapa, 2% DMSO was necessary in the binding buffer. To evaluate the release rate as relevant to in vivo use, loaded carriers were dialyzed for 28 days at 4 or 37 °C in PBS (without DMSO). This allows us to make two important observations: (i) the FKBP ELPs retain Rapa with a half-life more than one month and (ii) at body temperature, which induces phase separation, Rapa is retained for slightly longer. Thus, these data suggest that an FKBP/Rapa depot can retain and thus release significant amounts of the drug for at least 1 half-life, which would be 40 days.

Figure 4.

Both 5FV carriers exhibit high binding affinity and prolonged release of Rapa when induced to coacervate. SPR was used to quantify FKBP-Rapa affinity using a CM5 sensor chip conjugated with 5FV carriers. Rapa was loaded over a range of concentrations (75–625 nM) in HEPES. Fitting sensorgrams demonstrated a strong binding affinity with the following kinetics (eq 3): (A) (k_on = 9.1 × 10⁴ M⁻¹ s⁻¹, k_off = 2.9 × 10⁻⁴ s⁻¹, K_d = 3.2 nM) and (B) (k_on = 3.5 × 10⁵ M⁻¹ s⁻¹, k_off = 7.1 × 10⁻⁴ s⁻¹, K_d = 2.0 nM) for 5FV and L-5FV, respectively. To evaluate the rate of Rapa release from the loaded formulation in the absence of a solvent, between 350 and 400 μM Rapa loaded with 80 μM (C) 5FV or (D) L-5FV was dialyzed against PBS (1× Pen/Strep) under sink conditions for one month either below (cold, 4 °C) or above (hot, 37 °C) T_t. The drug remaining in the dialysis bag was quantified using RP-HPLC. The apparent half-life for Rapa loss from the dialysis bag was ~28 [CI: 21–41] and 40 [CI: 33–52] days at 4 °C for 5FV and L-5FV, respectively. Above the formulation T_t, at 37 °C, these half-lives were ~46 [CI: 32–74] and 49 [CI: 42–59] days for 5FV and L-5FV, respectively. Each sample was independently heated once until collection. Mean ± SD (n = 3). Best-fit curve (eq 4) (solid) with 95% CI (dotted).

L-Peptide Facilitates GRP78-Targeted Binding and Cellular Association of the 5FV Carrier

Many studies have linked GRP78 upregulation to breast cancer invasion, suggesting its utility as a molecular target of cancer.^63–67 The discovery that a subfraction of GRP78 translocates from the ER to the cell surface has made it an excellent target for enhancing receptor-mediated uptake into cancer cells.^41,44,68,69 To leverage cell-surface GRP78 for cancer targeting,^39,70 we engineered the L-peptide to 5FV.

To validate the effect of the L-peptide on interactions with GRP78, the binding affinity of both 5FV and L-5FV was assessed by using SPR (Figure 5A,B). Having confirmed the presence of GRP78 in BT-474 (Supplementary Figure S2), the cellular internalization of L-5FV was monitored using Western blot (Figure 5C), which significantly demonstrated a dose-dependent cellular association (Figure 5D,E).

Figure 5.

L-5FV selectively binds GRP78 and exhibits a dose-dependent cell association in the GRP78-expressing cell line. (A) SPR showed a nonspecific association of 5FV to histidine-tagged GRP78 (30 μg/mL) nickel chelated to a NTA chip. (B) In contrast, L-5FV showed a strong concentration-dependent association (eq 3) (k_on = 704.8 M⁻¹ s⁻¹, k_off = 3.1 × 10⁻⁴ s⁻¹, K_d = 442 nM). While the increased 5FV concentration yielded a nonspecific association, increasing L-5FV concentrations elicited proportionally greater responses. Even dilute L-5FV (6.25 μM) had a ~2-fold greater response than that of concentrated 5FV (30 μM). (C) Western blot was used to detect cellular uptake following a 2 h treatment of BT-474 cells using anti-ELP and anti-GAPDH antibodies. (D) Anti-ELP signal was normalized to GAPDH over increasing concentrations (eq 6) and (E) the AUC for both treatments was compared by the t-test (n = 3, independent replicates). L-5FV associated significantly more with cells than 5FV (p-values shown). Values indicate the mean ± SD.

L-Peptide Increases the GRP78-Mediated Surface Association of 5FV to a Cultured Spheroid

We further compared the differential association between GRP78, 5FV, and L-5FV using microscopy in a 3D spheroid culture (Figure 6A). This confirmed the higher affinity of L-5FV for a GRP78-positive culture (Figure 6B). 3D culture models are useful bridges between traditional monolayer cells and in vivo models. This approach may mimic the complex tumor microenvironment, including factors like drug and formulation penetration, cell migration, and the hypoxic conditions often present in solid tumors. By providing 3D geometry, spheroids enable the evaluation of therapies in culture that more closely resemble physiology.^71–73

Figure 6.

L-5FV associates more with the spheroid surface. (A) BT-474 cells were cultured as 3D spheroids prior to treatment with ELPs. Spheroids were fixed and stained by secondary immunofluorescence using anti-ELP (green) and anti-GRP78 (red) primary antibodies. Nuclei were stained with Hoechst (blue). Z-stacks through the spheroid were used to reconstruct an image of the entire spheroid (scale bar = 300 μm). GRP78 staining was evident with or without ELP; furthermore, L-5FV showed more spheroid association than 5FV. When only secondary antibodies were used, a minimal nonspecific binding/background was observed. (B) Compared by the t-test (n = 4, independent replicates) (eq 5), L-5FV indicates a stronger cellular association at the spheroid surface (p-values shown), where it colocalizes with GRP78. Values indicate the mean ± SD.

L-Peptide Enhances the Modulation of p-rpS6 Signaling

If the L-peptide enhanced the cellular uptake of the carrier, then it was expected to inhibit mTORC1 at lower concentrations compared to that of the untargeted carrier. BT-474 cells treated with increasing concentrations of Rapa in L-5FV led to a dramatic reduction in rpS6 phosphorylation at a concentration of 1 nM (Figure 7A); furthermore, 10-fold more concentrated free drug was required to observe the same suppression. This inhibition was quantified and fitted to determine the IC₅₀ value (Figure 7B), and the AUCs across all concentrations were compared. L-5FV-Rapa suppressed significantly more p-rpS6 signal than either free or untargeted formulations (Figure 7C). To explore the kinetics of inhibition, 1 nM Rapa was used to compare the targeted and untargeted carriers in the suppression of p-rpS6 (Figure 7D,E). Quantification showed that L-5FV-Rapa inhibited half of the activation in just 60 min, whereas untargeted 5FV-Rapa required 120 min (Figure 7F). Showing the power of both formulations, free Rapa was unable to inhibit the activity at this low concentration.

DISCUSSION

According to the 2024 American Cancer Society surveillance, breast cancer remains the second cause of cancer-related deaths in the United States of estimated 310,720 and 56,500 diagnosed cases with invasive breast cancer and ductal carcinoma in situ, respectively.^74,75 The Rapa-related pharmacophore everolimus is widely indicated for controlling hormone receptor-positive breast cancer.^76–79 In a randomized double-blinded phase III study in postmenopausal women with hormone receptor-positive,^80,81 advanced breast cancer, the addition of everolimus to exemestane led to a median progression-free survival of 6.9 months, compared to 2.8 months for placebo and exemestane. However, the combination treatment potentiated the occurrence of symptoms, including pneumonitis and stomatitis, leading to therapy discontinuation. Later on, 18-month follow-up data confirmed consistent findings of both prolonged progression-free survival and toxicity. In addition to toxicity, the FDA-approved oral dosage form suffers from limited and inconsistent bioavailability.^1,82 This issue can be linked to fluctuating therapeutic drug levels and the further development of drug resistance. Targeting an injectable therapeutic is one possible solution to stabilize plasma drug concentrations, which could minimize side effects driven by peak concentrations for oral formulations.^35,83 Herein, we used genetic engineering to construct a sequence of five human cognate receptors for Rapa ‘FKBP’, linked by four hydrophobic ELPs. Further, this construct was modified with a GRP78-targeting ligand ‘L-peptide’ to promote cancer cell-specific drug uptake. UV–vis spectroscopy demonstrated a similar thermosensitive behavior for both formulations (Figure 2). In response to the physiological temperature, both formulations form droplets on the 1 μm size scale (Figure 3B), which was minimally affected by drug entrapment. Upon injection, these “coacervates” form in situ depots with potential extended release.^24,84,85 To confirm the association of Rapa with FKBP, Rapa loading was detected by SEC-MALS (Figure 3C–F). This revealed an increase of 5FV and L-5FV-Rapa carriers’ MWs by about 3.5 and 5 kDa, which accounts for nearly the maximum number of Rapa per carrier. This finding was independently confirmed by using RP-HPLC (Table 1). We also characterized the binding affinity between Rapa and the 5FV carriers by SPR (Figure 4A,B). In the presence of 2% DMSO necessary to solubilize the free drug, the K_d values were ~3 and 2 nM for 5FV and L-5FV, respectively. This confirms that the L-peptide does not hinder Rapa binding. To evaluate release in the absence of the solvent, long-term dialysis was performed under sink conditions (Figure 4C,D), which revealed that coacervation slightly increased the half-life of release to greater than one month. Based on this study, Rapa depots may retain drugs for 1 month or longer. Thus, a carrier such as L-5FV has the potential to overcome the limitations of oral delivery of rapalogues with a subcutaneous injection given less frequently than once a month. Through combined sustained delivery and molecular targeting, this strategy has the potential to enhance therapeutic outcomes in cancers that depend on mTORC1 signaling.

Targeting GRP78 has emerged as a critical, widely applicable strategy due to its multifunctional chaperone role in ER-related UPR survival mechanisms. In fact, GRP78 is upregulated in many cancers, where it supports their survival and invasion.^86–88 Effective targeting to GRP78 can reverse these cancer protective mechanisms and restore cellular immunity.⁸⁹ The L-peptide “RLLDTNRPLLPY” is a cell-surface GRP78-targeting moiety identified via phage display to improve therapeutic outcomes, when decorating various cancer delivery systems.^41,90–92 Lee and co-workers reported that the L-peptide-modified liposome, loaded by doxorubicin, significantly reduced nasopharyngeal carcinoma in tumor-bearing mice, compared to a nontargeted liposome.⁹⁰ Moreover, the group confirmed that therapy binding was halted in the tumor tissue section when a competitive synthetic L-peptide was administered. Later, Wang and co-workers developed an L-peptide-mediated construct for theranostic purpose via the detection of the rhenium-188 (¹⁸⁸Re) radioisotope-decorated liposome. When loaded with doxorubicin, they noted enhanced tumor selectivity, uptake, and inhibition.⁹¹ More recently, Niu and co-workers fused the L-peptide to dual acidic pH and thermal-activated copolymerized chitosan and poly(N-isopropylacrylamide) nanoparticles that encapsulated paclitaxel.⁹² This systemic therapy allowed GRP78 active targeting and complete inhibition of tumor growth over a 60-day experimental period. Both formulations required intravenous administration of these therapeutics once every 3 days for 30 days, starting when the tumor volume reached 100 mm³. To study the biophysics of L-peptide/GRP78 affinity, Wang and collaborators performed in silico studies that revealed binding interactions of the L-peptide sequence within leucine 9 and 10 and proline 11 associated with three different pockets of GRP78.³⁴ In another study, the same group performed SPR on GRP78 immobilized on an NTA chip to control the geometrical orientation by Ni-GRP78 chelation chemistry under binding with the L-peptide at different concentrations.³² The K_d was estimated at 10 μM, which was defined by hydrogen-bonding interactions. Similarly (Figure 5A,B), we quantified the K_d for both 5FV and L-5FV, which revealed a slightly stronger K_d for L-5FV of ~0.4 μM, while 5FV demonstrated minimal dose-dependent binding. Consistent with cellular studies, both Western blot and 3D microscopy studies confirmed that L-5FV interacts more strongly with GRP78-expressing BT-474 cells (Figure 5E, 6B). Recently, our group reported a significantly lower cellular association and subsequent activity in BT-474 cells for 5FA, compared with L-5FA.³¹ In a relative research from Hasani and co-workers,⁹³ they demonstrated efficient cellular uptake and activity for a GRP78 cell-penetrating peptide, Pep42, engineered with iron oxide magnetic and β-cyclodextrin nanoparticles loaded with doxorubicin in a BT-474 cell model. This finding further supports the potential of GRP78 as a therapeutic target for cancer suppression, as well as the use of the BT-474 in vitro model for GRP78-targeting therapy screening.

One prerequisite for achieving therapy with any drug carrier is its uptake mechanism.^94,95 For ELP-mediated constructs, the main endocytic uptake routes are clathrin-, caveolin-, and macropinocytosis-mediated pathways.^96–99 Our prior study with a soluble FKBP-ELP, known as FAF, demonstrated macropinocytosis, which was inhibited by amiloride.⁹⁸ Relative to Liu and co-workers,^100,101 ELPs linked with the L-peptide are expected to bind GRP78 and then get endocytosed. When quantum dots were conjugated with the GRP78-targeting cyclic peptide ‘Pep42’, they were explored for colocalization with GRP78 on the A549 lung adenocarcinoma cells and subsequent cellular fate using transferrin and cholera toxin subunit B to differentiate between clathrin- and caveolin-mediated endocytosis. Upon inhibiting clathrin and caveolin uptake by chlorpromazine and nystatin or methyl-β-cyclodextrin, respectively, Pep42-Qdots internalization was only impacted by the clathrin inhibitor. This confirms that clathrin-mediated endocytosis is central to their cellular delivery. Another group identified a monoclonal antibody against GRP78 ‘MAb159’ and reported their clathrin-mediated uptake.⁴⁶ In general, target–ligand complexes enhance cellular selectivity as well as trigger endocytic uptake.^102–104 Following the cellular entry of L-5FV, Rapa delivery is facilitated by its interaction with FKBP12.¹⁰⁵ To confirm this, we explored the functional inhibition of mTORC1 signaling under low concentrations of the drug. When BT-474 was treated by both formulations (Figure 7A), 100 and 10 nM resulted in complete p-rpS6 inhibition; however, only 1 nM showed comparable suppression of p-rpS6 within 2 h. For further understanding p-rpS6 regulation over 2 h, we tracked the signal every 30 min (Figure 7D). As shown, L-5FV-Rapa complex-treated groups inhibited p-rpS6 more completely. However, the noticeable inhibition occurred following 1 h, which correlates with the endocytic uptake timing and Rapa-mTORC1 interaction.^{31,101,104–106} Compared with free Rapa at these low doses, both 5FV and L-5FV-Rapa complexes had more significant inhibition on p-rpS6.

Overall, this paper highlights the successful cloning, thermoenvironmental response, and enhanced cellular delivery using 5FV carriers with superior potency to free Rapa. This level of activity has not been easy to achieve using other drug carriers. For example, a study from Nandi and co-workers¹⁰⁷ encapsulated concentrated Rapa in liposomal nanoparticles. They compared conventional and poly(ethylene glycol) (2000 MW)-shielded liposomes composed of dipalmitoylphosphatidyl choline (DPPC) with distearoylphosphatidyl choline (DSPC) cholesterol liposomes. A slight difference in the release rate was observed for DPPC, which was attributed to its lower melting temperature. When cytotoxic effects were tested in BT-474 cells over 72 h, neither formulation reduced the viability more than ~50%, despite incubation at concentrations ~900,000 nM. Minor differences between their antiproliferative effect were attributed to the slower release of the DSPC composition, which possesses a higher membrane phase transition temperature (T_m) that extends release. In comparison, both 5FV and L-5FV formulations retain Rapa for ~1 month (Figure 4C,D), internalize into BT-474 cells (Figure 5C, 6A), and inhibit p-rpS6 in less than 2 h (Figure 7A,D). Like 5FV-Rapa, their liposomal drug release data reflected ~1% of Rapa per day; however, L-5FV-Rapa reached an IC₅₀ value for mTORC1 inhibition of only 0.5 nM (Figure 7B). Thus, in the cell culture, L-5FV-Rapa is almost 2 million times as potent as their Rapa liposomes. In addition to liposomal carriers, multiple nanoplatforms have been explored for Rapa encapsulation, including lipid–polymer complexes, nanostructured lipid carriers, lipid nanocapsules, and biodegradable polymeric systems.^108–112

Compared to clinically approved depot-forming polymers like PLGA,^113,114 ELPs offer thermosensitive advantages of injection and extended release, which are influenced by physiological factors including temperature, concentration, ionic strength, and protein–protein interactions.¹¹⁵ High depot concentrations after administration may minimize systemic exposure^116,117 while allowing low-level absorption at concentrations of 5FV-Rapa that remain soluble during circulation and may redirect the distribution properties of the drug. To maximize potency, ligands like the L-peptide motif could facilitate faster tissue uptake, mitigate off-target interactions, and increase the duration of a dose.³¹ While this manuscript has not investigated Rapa release in vivo, our recent report of a localized 32 μL injection of Rapa encapsulated in 5FV to the lacrimal gland reduced inflammation at a lower dose than required for systemic Rapa.²⁴ That small dose achieved a 3 day mean residence time (MRT), which was significantly greater than that of a soluble control. In another study using a different cognate receptor–drug complex,^23,118 we engineered cyclophilin, the intracellular cognate receptor of cyclosporine A linked by a backbone ELP similar to 5FV, called CVC. Interestingly, the subcutaneous administration for CVC demonstrated an MRT of ~52 h, a 4-fold increase, compared to both the subcutaneous and intravenous administration of a soluble control ELP (CAC). Both of those in vivo studies monitored free drug concentrations and provided evidence of superior activity and lower toxicity based on an ELP depot. Consistent with those findings, the Rapa dialysis release data herein confirms that Rapa can be contained within a depot for at least one month (Figure 4C,D). ELP phase separation itself does not cause burst release; furthermore, the shorter MRT observed in vivo suggests that ELP depot dissolution, not the dissociation of Rapa from 5FV, is likely to determine the duration of a dose. Moreover, this manuscript advances this approach by incorporating a targeting ligand capable of engaging cells with the UPR, which are hallmarks of cancer and inflammation.^119,120 When faced with the tumor microenvironment and extracellular and intracellular proteolytic ELP degradation,^121–123 carrier drug release may be impacted. Future studies must clearly determine how to extend the duration of dissolution using ELP depots and how that affects the relative bioavailability of free and bound drugs. While these data suggest that 5FV can retain Rapa longer than the mean absorption time of the depot, additional investigation is required to confirm whether free Rapa or 5FV-bound Rapa contributes greatest to therapy. For a ligand-mediated carrier like L-5FV to be more effective than 5FV in vivo, Rapa must remain bound until cellular uptake occurs. In the case of L-5FV-Rapa, circulating Rapa concentrations above its K_d (2 nM) and/or IC₅₀ (0.5 nM) (Figure 4B, 7B) will be necessary to see benefits of targeting the UPR. Achieving this balance in vivo requires fine-tuning the balance of the absorption, distribution, biodegradation, and clearance of the carrier. In contrast to many other platforms,¹²⁴ ELP biopolymers are defined by their tunable amino acid blocks, which enable controllable molecular engineering, tailored temperature responsiveness, scalable protein expression, dynamic cellular uptake and trafficking, and biodegradability (Figure 8). These aspects raise their potential as next-generation delivery systems.^{121,125–130}

Figure 8.

Temperature-responsive carrier of Rapa, L-5FV, targets GRP78 and promotes the inhibition of mTORC1. (A) Below T_t, L-5FV and bound Rapa remain freely dissolved in solution (Figure 3A); however, above T_t, the formulation phase-separates into micron-scale particles (Figure 2, 3B, and Supplementary Figure S3). Meanwhile, encapsulated Rapa is retained for long durations within the formulation (Figure 4C,D). (B) At concentrations of 1 μM or below, where L-5FV remains soluble (Figure 2E, Table 1), the L-peptide facilitates recognition by a fraction of GRP78 translocated to the cell surface during the UPR prevalent in cancer stress, which leads to internalization (Figure 5C, 6A). This endocytosis-mediated uptake triggers Rapa transfer to endogenous FKBP12, which inhibits mTORC1, blocks p-rpS6 phosphorylation (Figure 7), and inhibits cell growth.

CONCLUSION

Biomolecular engineering of cognate human receptor proteins, such as FKBP, offers a precise and tunable approach to design drug-specific carriers. To maximize the duration of each dose, they should retain the drug for long durations and leverage receptor-mediated cellular internalization. To achieve this, herein, we report a new generation of carriers complexed with Rapa that target cells under ER stress, which is a hallmark of cancer. Both 5FV and L-5FV fusions contain five FKBP domains to bind Rapa, which associate with nanomolar affinity. To form a long-release depot upon injection, they are linked through an ELP selected to phase-separate at body temperature. In the depot phase, they retain the drug with a half-life greater than 40 days. After absorption, they might be expected to retain the drug until taken up by target cells. As they noncovalently bind the drug, cellular uptake and lysosomal proteolysis are expected to degrade the carrier, leading to cargo release. To maximize low circulating concentrations of the formulation, the NPs may benefit from target-mediated uptake. To demonstrate this, the L-5FV formulation was modified with a GRP78-targeting peptide. This enhanced the cellular association and functional inhibition of mTORC1, as detected through the suppression of p-rpS6. Biomolecular engineering of proteins such as L-5FV-Rapa affords deliberate control over drug loading, retention, cellular uptake, and release. For the first time, this article implements this strategy to deliver Rapa from an ELP depot to cancer cells under ER stress. Such a strategy could be broadly implemented with other drugs and molecular signatures of disease.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.biomac.5c02348.

Molecular weight characterization of rapamycin-encapsulated 5FV formulations; confirmation of GRP78 expression in BT-474 cells; phase contrast microscopic evaluation of ELP coacervation; and raw Western blot data from mTORC inhibition study (PDF)