Oral delivery of siRNA lipid nanoparticles: Fate in the GI tract

Oral delivery, a patient-friendly means of drug delivery, is preferred for local administration of intestinal therapeutics. Lipidoid nanoparticles, which have been previously shown to deliver siRNA to intestinal epithelial cells, have potential to treat intestinal disease. It is unknown, however, whether the oral delivery of these particles is possible. To better understand the fate of lipid nanoparticles in the gastrointestinal (GI) tract, we studied delivery under deconstructed stomach and intestinal conditions in vitro. Lipid nanoparticles remained potent and stable following exposure to solutions with pH values as low as 1.2. Efficacy decreased following exposure to “fed”, but not “fasting” concentrations of pepsin and bile salts. The presence of mucin on Caco-2 cells also reduced potency, although this effect was mitigated slightly by increasing the percentage of PEG in the lipid nanoparticle. Mouse biodistribution studies indicated that siRNA-loaded nanoparticles were retained in the GI tract for at least 8 hours. Although gene silencing was not initially observed following oral LNP delivery, confocal microscopy confirmed that nanoparticles entered the epithelial cells of the mouse small intestine and colon. Together, these data suggest that orally-delivered LNPs should be protected in the stomach and upper intestine to promote siRNA delivery to intestinal epithelial cells.


Results
LNPs maintain potency in a wide range of pH solutions. The pH of the GI tract ranges from 1-2 in the stomach to 8 in the large intestine 31 . A concern for oral siRNA LNP delivery is that LNPs may not be stable under such a broad range of pH conditions. Therefore, LNP stability as a function of pH was investigated by formulating the LNPs in PBS at a pH of 7.4 and then diluting them into PBS buffered to pH values of 1-9. After 30 minutes of incubation, LNPs were characterized and evaluated for their ability to induce gene silencing. Gene silencing was assessed by transfecting HeLa cells that stably express firefly and Renilla luciferase with LNPs containing siRNA against the firefly luciferase gene. Renilla luciferase served as a negative control, and reductions in Renilla luciferase would be suggestive of cytotoxicity or off-target effects.
As seen in Fig. 1a, LNPs induced ~80% luciferase gene silencing in HeLa cells at an siRNA dose of 2 nM, regardless of buffer pH (blue circles). Controls were conducted in which HeLa cells received the same volume of titrated PBS as the transfected samples. Since changes in buffer pH did not significantly alter luciferase gene expression (black squares), we concluded that LNP efficacy is not a function of pH. LNP siRNA entrapment efficiency, shown in Fig. 1b, remained around 85% for all pH values up to 8, at which point entrapment dropped to ~50%. Interestingly, this decrease in measured siRNA entrapment did not correlate with reduced efficacy. Similarly, as seen in Fig. 1d, buffer pHs between 1 to 7 did not affect the z-average diameter or polydispersity index (PDI) of the LNPs, with values of approximately 140 nm and 0.12, respectively. LNPs incubated in pH 9 buffer showed a slight increase in size and PDI at 150 nm and 0.17, respectively. As expected, surface charge, measured by zeta potential, varied with buffer pH. The LNP surface charge became slightly positive in lower pH PBS and slightly negative at pHs of 7-9 (Fig. 1c). The change in zeta potential can be attributed to protonation of lipidoid amines under low pH conditions. On the whole, these experiments suggested that the changes in pH that LNPs would experience as they move through the GI tract will not adversely affect their functionality.
Pepsin and bile salts reduce LNP efficacy. In addition to drastic pH changes, the GI tract contains digestive enzymes that break down proteins, nucleic acids, sugars and other nutrients during digestion. Therefore, we next asked whether or not these enzymes impact LNP delivery efficacy. We tested LNPs after a simulated in vitro digestion in which they were exposed to the enzymes pepsin and pancreatin, as well as bile salts. Pepsin is found in the stomach and degrades proteins. Pancreatin, a secretion of the pancreas, is a mixture of enzymes that include trypsin, amylase, lipase, ribonuclease, and proteases 32 . Along with bile salts, which emulsify lipids, pancreatin and pepsin digest macromolecules into their building blocks for absorption or clearance from the body [33][34][35] .
LNP "digestion" was performed in two steps. First, formulated LNPs were incubated in simulated gastric fluid that contained pepsin (pH 1-2). After 30 minutes, they were exposed to a simulated intestinal fluid containing pancreatin and incubated for 30 minutes. LNPs were then delivered to Caco-2 cells, which are an established in vitro model of intestinal epithelial cells 36,37 . LNPs contained siRNA specific against the housekeeping gene, GAPDH, which was dosed at 10 nM total siRNA. Gene silencing was assessed 24 hours later by quantitative PCR. As seen in Fig. 2a, the digested LNPs were ineffective, while the undigested LNPs achieved ~70% gene silencing. Potency may have been reduced by aggregation of the LNPs, as seen in the significant increase in the z-average diameter and PDI of the digested LNPs compared to the original nanoparticles (Fig. 2b,c). To identify the cause of the efficacy reduction, we deconstructed the in vitro digestion process. LNPs were examined after exposure to either pH 1.2 buffer, pepsin, pancreatin, or bile salts at concentrations representative of "fed" conditions in humans. Figure 2d shows that pepsin and bile salts both negatively impacted LNP efficacy.
Subsequently, we asked how pepsin and bile salt concentration influenced gene silencing, because digestion enzyme concentration can vary with the timing and nutritional content of meals. The concentrations applied to LNPs in Fig. 2d are representative of higher, fed state concentration of pepsin and bile salts. In this experiment, LNPs loaded with siGAPDH were mixed with increasing concentrations of pepsin (0-3.2 mg/mL) and delivered to Caco-2 cells at an siRNA dose of 10 nM. Although LNP efficacy declined with increasing pepsin concentration (Fig. 3a), potency was partially retained at "fasting" pepsin concentrations (0.25 mg/mL). LNP diameter and PDI increased with pepsin concentration (Fig. 3b,c) while the DLS particle count rate remained constant, suggesting that the enzyme caused LNP aggregation which prevented cell uptake and/or efficient gene silencing in Caco-2 cells.
A similar trend was seen when the LNPs were mixed with increasing concentrations of bile salts (3-15 mM) (Fig. 3d). LNPs exposed to 3-10 mM (fasting concentrations) of bile salts retained most of their gene silencing ability in Caco-2 cells. However, the 15 mM bile salt concentration completely prevented gene silencing. LNPs incubated with bile salts maintained their original size at bile salt concentrations up to 10 mM (Fig. 3e). The PDI (Fig. 3f) and the particle count rate (data not shown) did not change with concentration. Together, these data indicate that oral LNP delivery should be conducted under fasting conditions to facilitate LNP stability and retain the majority of LNP gene silencing ability.

Mucin prevents LNP gene silencing in vitro.
Mucus, which coats the intestinal epithelium, serves as a physical barrier that hinders the transport of macromolecules. While this physiological barrier prevents bacteria and other toxins from accessing epithelial tissue, it also inhibits the delivery of lipid nanoparticles to their cellular targets. The mucus barrier of the intestines is 200-800 µm thick and has a short turnover time of several hours that facilitates the rapid clearance of undigested compounds 38,39 . To test LNP diffusion through mucus, siGAPDH-loaded LNPs were delivered to Caco-2 cells that were incubating in culture media containing 2-5% (w/v) mucin type II. These mucin concentrations are representative of those observed in human intestines 40 . As seen in Fig. 4a, mucin affected LNP gene silencing in a concentration dependent manner. Even low concentrations of mucin (2% w/v) significantly reduced gene silencing from ~90% to ~40% at an siRNA dose of 100 nM. These experiments suggest that LNPs do not readily diffuse through mucin to enter the Caco-2 cells.
Previous studies have reported that an increase in PEG density on the outside of a particle can improve diffusion through mucus 41 . Therefore, we asked whether increasing the PEG mol% in the LNP formulation above the standard 1.5 mol% would improve gene silencing in the presence of mucin. LNPs were formulated with 1.5-10 mol% PEG and pipetted on top of the mucin-covered Caco-2 cells. Twenty-four hours post-transfection, nanoparticles containing 2.5 mol% PEG resulted in the most gene silencing, as seen in Fig. 4b. Although this value was not significant compared to 1.5 and 5 mol% PEG, there is likely a trade-off between improving diffusion and inhibiting target cell uptake 42,43 . LNPs formulated with higher mol% PEG had lower siRNA entrapments and z-average diameters (Fig. 4c,d).

LNPs do not affect tight junction integrity.
We also evaluated LNPs for their effect on intestinal epithelial tight junctions. Tight junctions are protein complexes that form between intestinal cells that restrict transport of macromolecules from the apical to the basolateral side of the intestinal epithelium 44 . In several intestinal diseases, including inflammatory bowel disease and colon cancer, persistent inflammation results in the loosening of tight junction, adversely affecting the integrity of the intestinal barrier. To avoid further exacerbation of this condition, it is important that intestinal therapeutics do not affect tight junctions. To determine LNP influence on tight junction integrity, we cultured Caco-2 cells into monolayers that recapitulate intestinal cell arrangement and tight junction formulation in vivo. One to three hours after delivery of LNPs to Caco-2 monolayers, tight junction integrity was assessed by transepithelial electrical resistance (TEER). Reductions in resistance would be indicative of loosening of tight junctions and decrease in barrier function. As seen in Fig. 5a, there was no significant change in TEER, even at very high LNP-siRNA doses, suggesting that LNPs do not affect barrier function. After three hours, some of the monolayers were fixed and stained to visualize the arrangement of the tight junction protein, zonula occludens-1 (ZO-1). The localization of ZO-1 did not change in the presence of 100 nM siRNA LNPs (Fig. 5b).
LNPs remain in the mouse GI tract for at least 8 hours post-delivery. We have previously shown that LNPs enter Caco-2 intestinal cells and elicit potent gene silencing in these cells in vitro 28 . The next question is, what will happen to the nanoparticles after oral delivery in vivo? First, we studied LNP biodistribution after a single oral dose of LNPs containing Cy5.5-labeled siRNA in C57BL/6 mice. We also conducted PBS and naked Cy5.5-labeled siRNA controls. The mice were sacrificed 30 minutes to 8 hours after the oral dose, and their excised organs were imaged for fluorescent signal using an in vivo imaging system (IVIS). Figure 6a shows this quantified signal as a function of time for the stomach, small intestine, and colon. Both naked siRNA and LNPs moved through the stomach and into the small intestine by 30 minutes post-administration. Over the next several hours, fluorescent signal increased in the colon as the samples moved through the GI tract. Total signal decreased over time as siRNA was eliminated from the GI tract. Figure 6b shows representative fluorescent images at 4 hours post-gavage. There was no significant siRNA signal in the kidneys, heart, liver, spleen, or pancreas, suggesting that a negligible amount of naked siRNA or the LNPs crossed the epithelial barrier to enter the blood stream.
After confirming LNP biodistribution after oral gavage in mice, we tested gene silencing efficacy in mouse colon cells following a single oral or rectal dose of LNPs. In these experiments, LNPs loaded with siGAPDH were  delivered by oral gavage or rectal administration to mice at an siRNA dose of 5 mg/kg. Viewing Fig. 7, the average GAPDH gene expression was lower in mice that received LNPs; however, average expression was not statistically significant from expression in the untreated mice GAPDH levels. Obtaining statistical significance was hindered, at least in part, by the substantial variability in GAPDH expression levels in untreated mouse colon cell samples. Based on our mucin data, we suspect that an insufficient number of LNPs are able to diffuse to the vicinity of the intestinal epithelial cells to significantly knockdown gene expression.

LNPs enter the cells of the small intestine and colon.
To investigate the lack of gene silencing further, confocal microscopy was used to visualize intestinal cell entry in mice. Tissue sections from mice that received an oral gavage of LNPs loaded with Cy5.5-labeled siRNA were washed, fixed and stained for DNA and actin. Control mice received either PBS or naked Cy5.5-labeled siRNA. As shown in Fig. 8, only the mice that received LNP-formulated siRNA had significant Cy5.5 fluorescence signal in the small intestinal and colon epithelial cells. Surprisingly, some LNPs penetrated to the base of the villi in the small intestine and colon, indicating that at least a subset of LNPs successfully diffuse to the intestinal tissue. LNP uptake across samples and within the samples shown was not consistent, however, and it is likely that a greater percentage of intestinal cells must receive LNPs to result in gene silencing.

Discussion
RNAi therapeutics have the potential to offer a mechanistically distinct form of therapy for several intestinal diseases, including inflammatory bowel disease, colon cancer, and irritable bowel syndrome 2,4,45 . Unfortunately, the oral delivery of nucleic acids is challenging due to the harsh environment of the GI tract 14 . Specifically, nucleic acids like siRNA must survive low pH conditions, avoid nucleases, and penetrate the mucosal barrier to reach the target intestinal epithelial cells 14 . Most siRNA delivery systems, including lipidoid nanoparticles, were developed with systemic delivery in mind. Understanding their stability and transport properties in the GI tract is important if their use is to be extended for oral delivery and the treatment of intestinal disorders.
Transit through the GI tract exposes a therapeutic to a wide range of pH values, from 1-2 in the stomach to approximately 8 in the large intestine 31 . Our experiments (Fig. 1) demonstrated that LNPs maintained stability and gene silencing efficacy in HeLa cells following exposure to a wide range of buffer pHs (1-9). Similar results had been found previously for long term LNP storage in solutions of different pH 46 . These data suggest that pH instability in the GI tract will not adversely affect lipid nanoparticle mediated gene silencing.
Digestive enzymes in the GI tract may pose a greater challenge. We selected enzyme concentrations based on human studies. The fed concentration of pepsin has been found to range from 0.26-1.72 mg/mL, and the fasted concentrations can range from 0.11-0.87 mg/mL 34 . The wide range in reported literature values is due to differences in experimental methods for measuring pepsin concentrations in humans.
Unfortunately, pepsin at high physiological concentrations prevented LNP gene silencing in Caco-2 cells. Based on the data in Fig. 3a, LNPs would need to be administered to the patient under fasting conditions to minimize the aggregation effect of pepsin. A similar study investigating nanoparticle stability in the presence of pepsin reported that solid lipid nanoparticles, nanostructured lipid carriers, and lipid nanoemulsions did not change particle size in the presence of pepsin at a concentration representative of a fasted stomach (0.32% w/v) 47 . However, at the same pepsin concentration, a study of miltefosine-loaded lipid nanoparticles reported a three-fold increase in particle size 48 . The mechanism of pepsin-induced LNP aggregation is unclear. Previous studies have reported that PEG and pepsin may interact electrostatically through the protonated carboxyl groups of the pepsin and the ether of the PEG by hydrogen bond formation 49 . Spelzini and colleagues observed that lower molecular weight PEGs, like the one used on LNPs, are more hydrophobic in an extended conformation with a higher capacity to interact with the pepsin protein domain 50 . The pepsin and PEG2000 on the outer surface of the LNPs could be favorably interacting to cause the formation of aggregates that prevent the LNPs from entering the cell. Bile salts, another class of molecule that aids digestion, also prevented gene silencing at high concentrations. In the small intestines of humans, the fasted bile salt concentration is reported to be 2.6-6.4 mM and the fed concentration is 11-14.5 mM 33,34,51 . There is a possibility the LNPs would still result in potent intestinal gene silencing when the patient has fasted beforehand. Interestingly, the z-average diameter of the LNPs drastically decreased at a bile salt concentration of 15 mM. It is unclear why the high concentration reduced LNP size, however, it could be due to an interaction between the lipidoids in the LNP and bile salts. Bile salts are amphiphilic and self-assemble into micelles in an aqueous environment 33 . It is possible that at high concentrations, bile salts (15 mM) insert themselves into the lipid bilayer of the LNPs or solubilize some of the lipophilic components, leading to a change in particle formation. The size reduction could lead to less siRNA entrapment within the particles and therefore a loss of gene silencing efficacy.
The presence of mucin in Caco-2 cell culture also inhibited LNP gene silencing in a concentration dependent manner. The size, surface chemistry, and charge of LNPs play likely roles in their interactions with mucus. Another class of nanoparticles known as mucus penetrating particles (MPPs) have been shown to diffuse through mucus at speeds comparable to diffusion in water. These MPPs are ~100 nm in diameter, neutrally charged, and have a high density of low molecular weight PEG on the surface 38,52 . LNPs have similar structures to MPPs, except LNPs have much lower PEG surface density. In addition to aiding in mucus diffusion, nanoparticle PEGylation can also reduce non-specific protein aggregation, and immune cell uptake that hinders efficacy [53][54][55][56] . Unfortunately, we found that a large increase in PEG mol% in the LNPs reduced efficacy, likely because of shielding effects and decrease in siRNA entrapment. It is not clear if there is a way to formulate LNPs to facilitate their ready diffusion through mucus while retaining efficacy.
Fortunately, even though pepsin, bile salts, and mucin reduced gene silencing in vitro, the LNPs entered mouse small intestine and colon cells after oral delivery. The fluorescent siRNA signal penetrated down to the base of the intestinal villi, close to the lamina propria area of the mouse intestines. It is possible LNPs became more concentrated in certain areas of the mouse intestines than others, and this led to diffusion into the intestinal crypts because of a larger concentration gradient. This could be beneficial for delivering siRNA to immune cells that are present in the area. Unfortunately, even though signal from LNPs with siRNA was seen within small intestine and colon cells of mice, there was no significant gene silencing attained after oral or rectal delivery. There is a possibility the dose delivered (5 mg/kg) isn't sufficient to achieve full coverage of LNPs over the majority of epithelial cells. Furthermore, when excising intestinal sections for mRNA extraction, the small tissue sample that is taken may not be representative of the remaining length of the intestine. In other words, it is possible that there are localized sections of intestine where epithelial silencing is occurring. We found that different sections of the colon experienced varying degrees of gene silencing with no apparent pattern. Uniform delivery throughout the intestines may be the greatest remaining challenge to achieving siRNA-mediated gene modulation in the intestinal epithelium.
Based on this work, it is recommended that the particles be protected, at the very least, in the stomach to prevent pepsin inactivation. One option is particle encapsulation with a pH responsive polymer (e.g. Eudragit ® ) that will not dissolve until a specific pH is reached within the GI tract 57,58 . Another option would involve loading lyophilized LNPs 46 into a capsule coated with a pH responsive polymer. Alternatively, it may be possible to decorate LNPs with a high density of responsive polymer chains that provide steric hindrance to enzymatic degradation and are programmed to be released prior to cell entry. LNP formulation. LNPs were prepared as previously described 30   Animal Studies. Animal protocols were approved by the institutional animal care and use committee at Carnegie Mellon University (Pittsburgh, PA). All animal experiments were conducted in accordance with approved protocols. Female C57BL/6 mice of at least 6 weeks of age were purchased from Charles River Laboratories. Mice were housed under controlled temperature (25 °C) in 12 hour light-dark cycles. Animals were given access to standard diet and water.

Methods
In vivo oral and rectal LNP gene silencing. Mice were orally or rectally gavaged with LNPs loaded with siGAPDH at a dose of 5 mg/kg. After 24 hours, mice were sacrificed and the colons were removed. Feces was flushed from the mouse colon using PBS in a syringe. The colons were carefully cut open to reveal the intestinal epithelium and intestinal cells were collected by scraping the insides of the colon with a spatula. For this experiment intestinal cells were collected from a 3 cm section above the mouse rectum. The intestinal epithelium was then placed in TRIzol ® Reagent (Invitrogen) for mRNA extraction according to the manufacture's protocol.
Quantitative PCR was carried out as described above.
In vivo oral delivery imaging studies. LNPs with Cy5.5-labeled siRNA were formulated as previously described at 0.05 mg/mL siRNA. Female C57BL/6 mice were orally gavaged with either PBS, Cy5.5-labeled siRNA, or LNPs containing Cy5.5-labeled siRNA at an siRNA dose of 0.5 mg/kg. The mice were fasted 4 hours prior to oral gavage. At specific time points after oral gavage (0.5, 1, 2, 4, 6, or 8 hours), the mice were sacrificed and organs removed for fluorescence imaging on an IVIS spectrum imaging system (Perkin Elmer, MA) at excitation and emission wavelengths of 675 and 720 nm. Images from the organs of each mouse were taken and the fluorescence was analyzed using Living Image ® software. Data availability. Raw data will be made available upon reasonable request to the corresponding author.