An LNP (lipid nanoparticle) formulation can pass every quality check performed in phosphate-buffered saline and still fail the moment it enters circulation. Buffer measurements describe a particle in isolation; serum describes the same particle in a crowded, competitive, enzyme-rich fluid that actively reshapes its surface. Serum stability testing exists to close that gap before animal studies begin, so that in vivo results reflect the biology of the delivery system rather than an unmeasured formulation defect.
When an LNP is diluted into serum, several processes start at the same time. Plasma proteins adsorb onto the lipid surface within seconds to minutes, forming a biomolecular corona that becomes the particle's effective biological identity. Apolipoproteins, immunoglobulins, and complement factors compete for binding sites. Lipids exchange with circulating high-density and low-density lipoproteins, which can destabilize the particle shell and open a path for cargo release. Polyethylene glycol (PEG) chains can desorb, exposing the lipid core and eroding colloidal protection. Each event can change size, surface charge, morphology, and payload retention via pathways that are invisible in a simple buffer measurement.
Running in vivo work without prior serum characterization creates a diagnostic blind spot. If a formulation underperforms in an animal model, the result is hard to interpret: poor expression or rapid clearance could reflect a genuine design limitation or simply particles that aggregated, leaked, or were opsonized within minutes of administration. Because the adsorbed protein layer, not the pristine particle surface, drives biodistribution and cell-level recognition, characterizing stability in serum first converts an in vivo experiment from a guess into a hypothesis test.
A well-designed serum panel delivers four practical outcomes:
Early go/no-go decisions: formulations that aggregate or leak in serum are deprioritized before animal resources are committed.
Mechanistic diagnosis: separating surface-driven failures (corona, charge) from core-driven failures (cargo leakage, lipid exchange) shows which component to change.
Comparable ranking: testing all candidates under identical serum conditions produces a defensible ranking that is closer to biological behavior than buffer data alone.
Dose and route rationale: retention and stability profiles support the exposure assumptions behind a dosing plan.
Serum is also an analytically hostile matrix. It scatters light, contributes its own proteins to any size or charge measurement, and contains enzymes that degrade unprotected cargo. No single technique answers the full question, which is why serum stability evaluation is organized as an orthogonal panel: physicochemical monitoring, payload integrity, corona analysis, and functional biology. The four test groups below map onto that panel and together define what a formulation must survive before it reaches an animal.
Table 1. Common serum-induced failure modes and their downstream effects.
| Failure Mode | Underlying Cause | Typical Readout | Downstream Effect |
| Aggregation and size growth | Loss of steric or electrostatic protection, PEG desorption | Z-average rise > 20%, PDI increase, count-rate drop | Capillary entrapment, reduced target exposure |
| Surface charge shift | Protein adsorption masking the original surface | Zeta potential drifting toward a less negative value | Altered cell interaction and clearance rate |
| Payload leakage | Lipid exchange with lipoproteins, shell destabilization | Rising free RNA fraction by dye exclusion or chromatography | Lower delivered dose, flattened dose response |
| Payload degradation | Nucleases gaining access after a shell breach | Low-molecular-weight smear instead of an intact band | Loss of functional transcript irrespective of retention |
| Corona-driven tropism | Apolipoprotein enrichment on the surface | ApoE-rich corona by LC-MS/MS | Redirected uptake away from the intended tissue |
| Innate immune recognition | Complement and immunoglobulin adsorption | Complement activation products, cytokine release | Accelerated clearance and unexpected reactogenicity |
The first group of tests asks whether the particle itself remains physically intact once it meets serum. These measurements are fast, require little material, and are usually the earliest signal that a formulation is in trouble. The guiding principle is to compare a serum-incubated sample against an identical sample diluted into buffer, so that any difference can be attributed to the biological matrix rather than to handling.
Dynamic light scattering (DLS) is the workhorse of serum stability monitoring because it reports hydrodynamic size, size distribution, and scattering intensity without labeling or consuming the sample. A standard design incubates LNPs in 10–50% serum at 37 °C and measures at defined intervals, commonly 0, 1, 6, 24, and 48 hours. The most informative readouts are not the raw mean size alone:
Z-average (mean hydrodynamic diameter): a rise beyond roughly 10–20% of the baseline value signals aggregation or swelling, and the rate of change matters more than the absolute number.
Polydispersity index (PDI): values climbing above about 0.2–0.3 indicate a widening population, frequently the first detectable sign of instability.
Count rate (scattering intensity): a falling count rate at fixed laser power reflects fewer or smaller scatterers, which often means the intact particle population is disappearing even when the apparent size is unchanged.
Intensity distribution: a second peak at several hundred nanometers exposes a small aggregate fraction that PDI alone can hide.
Serum complicates interpretation because its own proteins scatter light and can associate weakly with particles. Practical safeguards include running serum-only blanks, matching the dilution medium between the serum and buffer arms, and treating the count rate as a co-primary readout. Teams that need particle-level confidence can pair DLS with a dedicated particle size analysis workflow that cross-checks the scattering result against an orthogonal principle.
Table 2. DLS readouts and how to interpret them during serum incubation.
| Parameter | What It Reflects | Stability Signal | Practical Threshold |
| Z-average | Mean hydrodynamic diameter | Aggregation or swelling | Stay within 10–20% of baseline |
| PDI | Width of the size distribution | Population broadening | Remain below 0.2; investigate above 0.3 |
| Count rate | Number and size of scatterers | Loss of intact particles | Stable versus baseline; a large drop is a red flag |
| Intensity distribution | Presence of subpopulations | Emerging aggregate fraction | A single dominant peak |
Zeta potential reports the effective surface charge of the particle and, by extension, the electrostatic component of its colloidal protection. In serum, a drift toward a less negative or less positive value is expected as proteins adsorb and screen the original surface, so the informative signal is the magnitude and speed of that drift rather than the absolute number.
Two cautions matter. First, zeta potential is highly sensitive to pH, ionic strength, and protein content, so a value measured in serum cannot be compared directly with one measured in water or dilute buffer, and the surrounding conditions should always be reported alongside the result. Second, sterically stabilized systems such as PEGylated LNPs can remain fully colloidally stable even when the zeta potential approaches zero, because their protection comes from the hydrated PEG layer rather than from charge repulsion. For these formulations, a zeta potential analysis is best read as an indicator of surface remodeling, such as PEG loss or protein binding, rather than as a standalone verdict on stability.
Cryo-transmission electron microscopy (cryo-TEM) provides the structural ground truth that scattering methods cannot. Samples are vitrified so that particles are captured in a near-native, hydrated state, allowing direct inspection of shape, internal structure, and population homogeneity. In serum stability work, imaging at time zero and at the end of the incubation answers questions that DLS only hints at:
Structural integrity: are particles still intact and discrete, or has a fraction collapsed into amorphous debris?
Population homogeneity: is the sample monodisperse, or has a second species of multilamellar or fused particles appeared?
Core organization: has the interior acquired structure that would suggest lipid reorganization or cargo release?
Because cryo-TEM is low-throughput, it is typically reserved for leading candidates and for confirming a suspected instability rather than for broad screening. A focused morphology characterization pass at the start and end of incubation is usually enough to decide whether a scattering signal reflects real structural change.
Table 3. Physicochemical serum stability test matrix.
| Method | Question Answered | Typical Condition | Time Points | Key Output |
| DLS | Size, distribution, aggregation kinetics | 10–50% serum, 37 °C | 0, 1, 6, 24, 48 h | Z-average, PDI, count rate |
| Zeta potential | Surface charge remodeling | 10–50% serum, 37 °C | 0, 24 h | Zeta drift versus baseline |
| Cryo-TEM | Structural and morphological integrity | 50% serum, 37 °C | 0, 24 h | Intact versus disrupted fraction |
| Turbidity and visual check | Gross aggregation | 50% serum, 37 °C | 0–48 h | Opacity change over time |
BOC Sciences runs serum-incubated DLS, zeta potential, and cryo-TEM panels with matched buffer controls, so you can see exactly what serum does to your formulation before you commit to animal work.
Physical integrity does not guarantee cargo integrity. A particle can keep its size and shape while slowly releasing its payload, or it can stay tightly sealed while the payload inside degrades. Because these two failures look identical in a simple potency readout but require opposite fixes, the second test group measures both how much cargo remains inside the particle and what condition that cargo is in. The methods below are complementary, and the strongest programs combine at least one retention assay with at least one integrity assay.
The RiboGreen fluorescence-based dye exclusion assay remains the most widely used retention method because it is fast, sensitive, and requires very little material. The fluorophore intercalates into accessible RNA but is physically excluded from RNA sequestered inside an intact particle. The standard workflow is straightforward:
Serum incubation: LNPs are mixed with 10–50% serum or human plasma and held at 37 °C for 0–72 hours.
Free RNA measurement: RiboGreen working solution (about 1:200 in TE buffer) is added and fluorescence is read near 485 nm excitation and 530 nm emission to quantify leaked or surface-associated RNA.
Total RNA measurement: a surfactant such as Tween 20 or Triton X-100 disrupts the particles, after which RiboGreen quantifies total RNA.
Retention calculation: retention (%) equals (total − free) divided by total, multiplied by 100.
Two limitations deserve attention. Fluorescence quenching by serum components requires serum-only blank correction, and the assay cannot separate truly encapsulated RNA from RNA merely adsorbed to the surface, so dye exclusion tends to underestimate encapsulation relative to charge-based separation. It is an excellent comparative screening tool, but it should not stand alone. A dedicated payload retention testing program pairs it with orthogonal methods for this reason.
Table 4. RiboGreen dye exclusion workflow for serum-exposed LNPs.
| Step | Operation | Purpose |
| 1 | Incubate LNPs with 10–50% serum or plasma at 37 °C for 0–72 h | Simulate exposure to a biological matrix |
| 2 | Add RiboGreen working solution in TE buffer, read Ex 485 nm / Em 530 nm | Quantify leaked and surface-associated RNA |
| 3 | Disrupt particles with Tween 20 0.5% or Triton X-100 1%, then add RiboGreen | Quantify total RNA in the sample |
| 4 | Retention (%) = (Total − Free) / Total × 100 | Derive the encapsulated fraction |
Where dye exclusion measures how much cargo remains inside the particle, denaturing agarose gel electrophoresis measures what that cargo looks like. In a typical stability experiment, LNPs are incubated with serum, the payload is recovered by extraction, and the recovered RNA is resolved on a denaturing MOPS-formaldehyde gel alongside untreated controls.
The readout is qualitative but highly diagnostic. An intact, full-length band indicates preserved cargo, whereas a low-molecular-weight smear indicates nuclease-mediated degradation that has breached the particle shell. Because degradation and leakage are chemically distinct events, running the gel alongside a retention assay is what allows the two to be told apart, and that distinction determines whether the fix is a tighter shell or a faster release profile. Gel analysis is also a natural companion to nucleic acid encapsulation work, where cargo state matters as much as cargo quantity.
Chromatographic separation resolves the ambiguity that dye-based assays cannot. Anion exchange chromatography (AEX) exploits the charge difference between anionic free nucleic acid and near-neutral intact LNPs, and uniquely distinguishes three payload states: truly free, surface-bound, and fully encapsulated. Deformulating size exclusion chromatography (SEC) takes the opposite approach, using a mobile phase that disrupts particles online so that total payload can be quantified in roughly five minutes without offline preparation. Ion-pair reversed-phase chromatography (IP-RP) adds value for multi-payload systems by resolving individual cargo species and detecting lipid-payload adducts.
These methods are complementary rather than competing: dye exclusion for fast comparative screening, gel electrophoresis for integrity, and chromatography for a definitive free-versus-encapsulated partition. Teams that need quantitative, method-validated encapsulation data can build on efficiency testing for LNP encapsulation and, where a clean baseline is required, on free payload removal before measurement.
Table 5. Comparison of payload integrity and retention methods.
| Method | What It Measures | Principle | Strength | Limitation |
| RiboGreen dye exclusion | Free versus total RNA | Dye intercalation excluded from intact particles | Fast, high throughput, minimal sample use | Cannot resolve surface-bound from encapsulated; quenching effects |
| Denaturing gel electrophoresis | Cargo integrity | Size separation of extracted RNA | Detects degradation directly | Semi-quantitative; requires extraction |
| Anion exchange chromatography | Free, surface-bound, encapsulated | Charge-based separation | Resolves three distinct payload states | Pressure-sensitive; method tuning per formulation |
| Deformulating SEC | Total payload | Online particle disruption | About five-minute run; minimal preparation | Destructive; no particle integrity information |
| Ion-pair reversed phase | Individual cargo species | Hydrophobicity-based separation | Multi-payload quantification; detects adducts | Offline preparation; detergent-dependent |
BOC Sciences combines dye exclusion, denaturing gel electrophoresis, and chromatographic separation to tell leakage apart from degradation, so you know which part of the formulation to change.
The particle that circulates is not the particle that was manufactured. Within minutes of contact with serum, a layer of adsorbed proteins forms on the lipid surface and becomes the interface that cells and clearance pathways actually see. Two formulations with identical size, charge, and encapsulation can therefore behave completely differently in vivo because their coronas differ. Corona analysis turns surface adsorption from an assumption into a measurement, and it is what connects a physicochemical stability profile to an expected biological fate.
Corona analysis begins with a separation problem: the proteins that matter are those adsorbed to the particle, not the far more abundant proteins in the surrounding medium. Common approaches include ultracentrifugation, size exclusion separation, and magnetic capture for magnetically responsive systems. Each must be performed under conditions that limit post-isolation exchange, because the loosely bound soft corona can be lost during washing while the tightly bound hard corona persists. Total adsorbed protein is then quantified with a colorimetric protein assay and profiled by SDS-PAGE before deeper analysis.
Table 6. Approaches for isolating and quantifying the biomolecular corona.
| Approach | Separation Basis | Best Suited For | Consideration |
| Ultracentrifugation | Density | Broad applicability and high protein load | Can shear loosely bound proteins |
| Size exclusion | Size | Gentle separation that preserves the soft corona | Dilutes the sample; limited capacity |
| Magnetic capture | Magnetic response | Magnetic or magnetically labeled carriers | Requires a magnetic handle |
| Affinity capture | Specific binding interactions | Focused analysis of selected corona subsets | Narrow scope by design |
Mass spectrometry converts the corona from an abstract concept into a protein list. After enzymatic digestion of the isolated corona, nano-liquid chromatography–tandem mass spectrometry (nanoLC-MS/MS) with label-free quantification identifies the adsorbed proteins and ranks them by abundance. Composition, not just total amount, carries the biological meaning: recurring themes in LNP corona work include enrichment of apolipoproteins, immunoglobulins, complement components, and acute-phase proteins, each with different implications for where a particle goes and how the body responds to it.
The value of corona profiling is that it links a measurable surface property to a predictable biological outcome. An apolipoprotein-enriched corona, particularly one rich in apolipoprotein E, is associated with hepatocyte uptake through the LDL receptor pathway, which helps explain the pronounced liver tropism of many systemically administered LNPs. Complement and immunoglobulin adsorption, by contrast, points toward faster clearance and potential immune recognition. When corona signatures are collected alongside distribution data, they become a predictive tool: a team can flag a formulation likely to be redirected away from its intended tissue before committing to a full animal study, and confirm the change with a biodistribution analysis.
Table 7. Corona features and their likely delivery implications.
| Corona Feature | Mechanism | Predicted Behavior | Design Implication |
| Apolipoprotein E enrichment | LDL receptor recognition | Hepatocyte uptake | Expected for liver-tropic LNP; engineer away for other targets |
| Apolipoprotein A-I and HDL | Lipid exchange | Shell remodeling and leakage risk | Tune lipid composition and PEG shielding |
| Complement components | Opsonization | Faster clearance, immune activation | Consider a complement-sparing surface design |
| Immunoglobulin adsorption | Fc recognition | Immune cell engagement and clearance | Evaluate together with immune readouts |
| High total protein load | Nonspecific adsorption | Reduced targeting specificity | Improve stealth through PEG density and architecture |
BOC Sciences isolates and profiles the biomolecular corona by nanoLC-MS/MS, then interprets the result against expected biodistribution, so surface adsorption becomes a design input instead of a surprise.
A particle can be physically intact, fully loaded, and still fail to deliver. Functional stability testing closes the loop by asking whether a formulation that survives serum can still enter cells, release its cargo, and avoid triggering unwanted biological responses. These assays are the closest in vitro proxy for the question that matters most: does serum exposure cost the formulation its function?
The most direct functional test pre-incubates LNPs in serum for defined periods, then applies the aged particles to cells and measures uptake and expression. Dual-label readouts are especially informative: labeling the cargo, for example a fluorescently tagged mRNA, and the lipid simultaneously allows a single flow cytometry experiment to separate particles that were taken up from cells that actually translated the payload. Comparing the serum-aged sample against a freshly diluted control isolates the functional cost of serum exposure, a cost that physicochemical data alone cannot quantify. A focused cellular uptake testing design with intracellular localization readouts shows whether the payload also reaches the compartment where it can act, and a broader in vitro evaluation can extend the same logic across cell models.
Because most systemically administered LNPs meet blood directly, hemocompatibility is a central readout. A hemolysis assay exposes red blood cells to the formulation and quantifies hemoglobin release as a percentage of total lysis, with low values indicating membrane compatibility. Complementing this with an assessment of particle behavior in whole blood, including aggregation, protein binding, and cell association, gives a fuller picture of how the formulation behaves in the compartment it must survive before reaching any tissue. This work sits naturally alongside a broader safety assessment of the carrier.
Serum stability and immune recognition are intertwined, because the same surface that adsorbs targeting-relevant proteins can also trigger the complement cascade. Measuring complement activation products and cytokine release after serum or whole-blood exposure reveals whether a formulation provokes an innate response, while repeat-exposure designs can surface anti-PEG antibody formation that alters clearance on later doses. These readouts matter for interpretation: an unexplained loss of efficacy on repeat dosing is frequently a stability-and-recognition problem rather than a dosing problem. They also connect directly to cellular and in vivo evaluation, where the functional and immune consequences of serum exposure are ultimately confirmed.
Table 8. Functional and biological stability assay panel.
| Assay | Question Answered | Readout | Typical Design |
| Uptake and transfection | Does a serum-aged LNP still deliver? | Percent positive cells, expression level, dual-label | Pre-incubate in serum, then transfect |
| Hemolysis | Is the carrier membrane compatible? | Hemoglobin release percentage | Incubate with red blood cells, measure lysis |
| Whole-blood behavior | Does the carrier aggregate in blood? | Size drift and blood cell association | Spike into fresh whole blood |
| Complement activation | Does the surface trigger innate immunity? | Complement activation products | Serum or whole-blood exposure |
| Cytokine release | Is there an unwanted immune response? | Cytokine levels | Exposure and repeat-dose design |
BOC Sciences pairs serum-aged uptake and transfection readouts with hemocompatibility and innate immune profiling, so functional stability is demonstrated rather than assumed.
Serum stability testing is rarely a single clean experiment. The matrix interferes with instruments, different failure mechanisms produce the same symptom, and results shift when conditions change. Most of the frustration in this area traces back to four recurring problems, and each has a practical workaround.
Serum proteins scatter light and can associate weakly with particles, inflating the apparent size and broadening the distribution. The fix is straightforward: run serum-only blanks to subtract the matrix contribution, dilute into the identical medium used for the buffer control, treat the count rate as a co-primary readout so that a stable count rate argues against true aggregation, and confirm any suspected aggregation with a count-based or imaging method. Reporting the medium composition alongside every measurement keeps the comparison honest.
Leakage and degradation produce the same symptom, less functional cargo, from different causes, and they require different fixes. Resolve them by pairing a retention assay with an integrity assay. Cargo recovered intact but found outside the particle points to a permeability problem; cargo recovered degraded points to a barrier problem that let nucleases in. The practical guidance for building this comparison is summarized in this overview of LNP payload retention during serum exposure.
Serum is a biological reagent with real lot-to-lot variability, and species, source, and concentration all shift the outcome. Standardize and report the serum type, species, percentage, temperature, and time points; use pooled serum or a qualified lot for cross-study comparisons; and keep the dilution factor constant, because serum concentration itself changes protein competition and lipoprotein availability.
Aggregation and leakage often begin long before a size or retention metric crosses a conventional threshold. Improve sensitivity by monitoring the count rate and intensity distribution alongside the Z-average, using dual-label functional readouts that detect small populations of failed particles, running forced-stress conditions such as elevated temperature, freeze-thaw, and agitation to amplify early weaknesses, and selecting stability-indicating methods that respond to the specific degradation pathway you expect. Methods that probe environmental triggers, such as stimuli-responsive testing, can reveal latent instability that standard incubation misses.
Table 9. Common challenges, root causes, and practical solutions.
| Challenge | Root Cause | Practical Solution | Confirming Readout |
| Protein interference with sizing | Serum light scattering and weak adsorption | Matrix blanks, matched dilution, count-rate co-monitoring | Agreement between scattering and an orthogonal method |
| Leakage versus degradation | Two mechanisms, one symptom | Combine a retention assay with an integrity assay | Free-intact versus free-degraded partition |
| Cross-study inconsistency | Serum lot, species, and concentration variability | Standardized and reported conditions; qualified pooled lot | Reproducible baseline curves |
| Low early-detection sensitivity | Threshold-based endpoints and single-method reporting | Count-rate and distribution monitoring; forced stress | A signal that appears before a threshold is breached |
| Corona-driven mistargeting | Apolipoprotein enrichment | Corona profiling linked to distribution data | Corona signature matched to tissue distribution |
Serum stability evaluation works best when the four test groups are designed around a specific question rather than run as a generic checklist. BOC Sciences builds serum stability programs that start from the payload, the intended route, and the decision the data must support, then selects the methods that answer that decision most efficiently.
Panel design begins with the study objective. BOC Sciences assembles physicochemical monitoring, payload integrity, corona analysis, and functional readouts into a coherent workflow, and develops the underlying analytical methods, from DLS and zeta protocols to dye-exclusion, chromatographic, and mass spectrometry workflows, so that results are comparable within a program and interpretable across programs. This method work is grounded in established LNP characterization and stability capability, and can include formal method development for encapsulation analysis when a defensible quantitative readout is required.
mRNA, siRNA, plasmid DNA, and protein or peptide cargoes each fail differently in serum. Nucleic acid payloads face nuclease attack and leakage through lipid exchange, while protein payloads face conformational and proteolytic stress. BOC Sciences tailors the panel to the payload, tuning retention and integrity assays to the cargo chemistry, adding corona profiling where tropism is a concern, and choosing functional readouts appropriate to the delivery goal, including workflows built around mRNA delivery and antigen delivery.
When a formulation fails in serum, the productive question is which component to change. Stability profiles can point to PEG architecture, lipid composition, or buffer and excipient choices as the lever. BOC Sciences supports optimization across these axes, from buffer screening and cryoprotectant screening to ionizable lipid optimization and PEG-lipid optimization, and pairs it with encapsulation troubleshooting workflows that trace a symptom back to its formulation cause. Because that failure mode can also depend on where the particle is meant to go, optimization can be carried through targeted LNP development so that serum stability and delivery intent are improved together.
Table 10. BOC Sciences services supporting LNP serum stability evaluation.
| Service | Scope of Service | Key Deliverables | Inquiry |
| Critical Quality Attributes and QC Testing | Definition of CQA panels and QC-style testing for serum-exposed LNPs | CQA report with acceptance ranges and batch comparison | Inquiry |
| Surface Charge Analysis | Surface charge measurement across buffer and serum conditions | Charge profiles with matched-condition comparison | Inquiry |
| Structural Characterization | Structural and internal organization assessment of intact and serum-aged LNPs | Structural integrity report with supporting micrographs | Inquiry |
| Encapsulation Efficiency Optimization | Formulation and process levers to improve and preserve encapsulation under serum stress | Optimized encapsulation efficiency with supporting data | Inquiry |
| Lipid Ratio and Composition Optimization | Lipid molar ratio screening for improved shell stability in serum | Ranked lipid ratio candidates with stability data | Inquiry |
| Protein Corona Proteomic Profiling | Corona isolation followed by nanoLC-MS/MS composition analysis | Corona protein list with relative abundance and interpretation | Inquiry |
| Transfection Troubleshooting and Functional Evaluation | Diagnosis of functional loss after serum exposure and corrective recommendations | Root-cause report with functional performance data | Inquiry |
| Excipient and Buffer Optimization | Excipient and buffer screening to improve serum stability and handling | Optimized excipient and buffer composition with stability data | Inquiry |
Serum stability testing is the bridge between a formulation that looks good in a buffer and one that behaves predictably in a living system. By working through the four test groups in sequence, physicochemical integrity, payload integrity, corona composition, and functional biology, research teams can identify why a formulation fails and, just as importantly, which component to change. Each group answers a different question, and together they turn an in vivo study from an expensive gamble into a test of a specific, well-characterized hypothesis. Adopting an orthogonal, question-driven panel early is what allows preclinical programs to spend their animal resources on candidates that have already demonstrated, in the most biologically relevant matrix available, that they can survive, deliver, and remain tolerated.
