Industry

Process Development Intern @ Bristol Myers Squibb (Summer 2026)

This summer, I had the unique opportunity to intern in Bristol Myers Squibb's Seattle location. I was an intern in their Gene Delivery and Editing Process Development Department, specifically part of the Gene Editing Raw Materials (GERMs) Team.

I got to perform studies on optimizing lipid nanoparticle (LNP) formulation parameters for the delivery of a DNA cargo. Through these studies, I gained exposure to numerous LNP formulation and characterization techniques such as using DLS, Quant-iT, liquid handlers, etc.

I describe the various studies I took part in the sections below.

Lipid Composition Study

The goal of this study was to determine the optimal LNP lipid composition for DNA delivery. LNPs are composed of four main components, each with a specific function. Broadly, these are the ionizable lipid (cargo encapsulation and endosomal escape), helper lipid (particle structure), cholesterol (membrane rigidity), and the PEGylated lipid (prevents aggregation).

In this study, LNPs were formulated with varying amounts of each lipid component. The ionizable lipid, helper lipid, and cholesterol were manipulated within specific molar percentage ranges. The PEGylated lipid remained at a constant molar percentage for each formulation.

Due to the three lipid components having wide ranges of molar percentages, a software called Design of Experiments was utilized to determine the best runs (or formulations) to experiment on within those ranges. Design of Experiments (DoE) is a statistical tool that takes in an input (the variables being manipulated and the constraints tied to each variable) and generates an output (the optimal runs that would yield the most useful data). Using DoE, 16 different LNP formulations with differing lipid compositions were generated and used in the subsequent experiment. The LNPs were then formulated using microfluidics mixing techniques. The cargo was a GFP encoding npDNA.

This is an example of a DoE Ternary Plot. In this plot, the mass fractions of three ingredients in a cookie are being varied in order to optimize cookie texture. The constraints fed to the DoE are as follows: 0.4 < Flour < 0.6; 0.2 < Butter < 0.4; 0.1 < Egg < 0.2. The white region represents the design space. The dots represent a cookie with a different composition of eggs, flour, and butter. A similar approach was utilized for this lipid composition study. Although, a mixture design DoE using Space Filling was utilized. This gives preferences to formulations in the middle of the design space. This cookie plot is a mixture design using D-Optimal, giving preference to edge cases instead.

The workflow consisted of using donor T cells to test the efficacy of the LNP formulations as well as electroporation (EP). The LNPs were tested on two CD4+:CD8+ donor T cells at a npDNA dosage of 2 ug/mL. Additionally, EP was tested with a npDNA dosage ranging from 0-157 ug/mL in order to obtain a full titration curve. The EP controls served as a benchmark to compare LNP transfection efficiency to an already well established non-viral modality. Flow cytometry was used as the main data readout 24 hours after transfection.

The results indicated a general region within the design space that yielded lipid compositions with greater transfection efficiency. It was concluded that this portion of the design space could be a hotspot for DNA delivery. Future formulations should use this optimal region of lipid composition as a guide when delivering DNA in vitro.

Donor Study

While a very broad trend was identified in the lipid composition study, the results weren’t conclusive. The goal of this study was to expand the data set by increasing the number of donors tested. This was done by transfecting the same LNPs from the prior study on five donors (two repeat donors and three new donors).

The data suggested a stronger trend towards a specific corner of the design space. Additionally, multiple lipid trends were observed with higher or lower amounts of some lipid components yielding formulations with greater transfection efficiency. A better guideline for DNA-LNP formulations were determined from this study.

The workflow remained largely identical. The same npDNA dosage of 2 ug/mL was used, and flow cytometry was the main data readout. The main differences being that there were no electroporated cells and only 15 LNPs were used due to low volume in one formulation.

Lipid Ratio and Screening Study

The first objective of this study was to determine an optimal lipid ratio for DNA-LNPs. The lipid ratio being tested was the N:P ratio. “N” refers to ionizable amines. These functional groups are abundant in ionizable lipids and are responsible for protonation during encapsulation and endosomal escape. The “P” refers to phosphate groups which are prevalent in the DNA and RNA cargoes that LNPs encapsulate.

Having a higher N:P ratio would indicate an abundance of ionizable lipids. This is beneficial for encapsulation and endosomal escape. Additionally, more ionizable lipid is correlated with smaller LNP size which can aid with cellular uptake. Although, an overabundance of ionizable lipid can be toxic to cells. A lower N:P ratio means less ionizable lipid and therefore worse encapsulation and cellular uptake but less toxicities.

The second objective was to screen various ionizable lipids to determine which is optimal for LNP DNA delivery. Four ionizable lipids were tested. Their main differences include pKa, chemical structure, and molecular weight. Generally, ionizable lipids with a lower pKa more readily protonate in acidic conditions.

Numerous LNPs with differing ionizable lipid and N:P ratios were tested. An optimal ionizable lipid and N:P ratio was determined and used for subsequent DNA-LNP studies.

Chemical structures of some of the ionizable lipids that were screened. The amine circled in green represents the ionizable functional group within each lipid.

Presenting my work at the end of summer intern presentations.

Key Takeaways

Over these last two months, I gained invaluable experience in the PD side of industry. Out of everything I learned, I feel there are two lessons that will stick with me in the future. The first is if you want to pursue an alternative therapy, simply showing better safety is not enough. There needs to be better (or equivalent) efficacy. The second is to get used to failure and troubleshooting. Very commonly in science, things will not go as expected. You can be doing everything right, but still end up with bad data due to factors outside of your control. And often times, these big issues stem from very minor, overlooked details.

Huge thank you to Colleen, Michael, and Doug for the mentorship this summer!

Page updated

Google Sites

Report abuse