bkoep Staff Lv 1
Starting in summer 2021, we ran 6 rounds of IL-2R binder design puzzles. Now we’ve selected 115 promising Foldit designs to test for binding in the lab! In early 2022, we will conduct a FACS experiment to test if these protein designs successfully bind to the IL-2R target.
Targeting IL-2R for cancer treatment
IL-2R is a protein found on the surface of human immune cells, and is composed of three different protein chains (α, β, and γ). Due to its role in regulating the immune system, IL-2R is an important target in cancer immunotherapy. However, IL-2R drugs are associated with severe side effects that seem to arise from over-activation of the α chain.
The goal of the Foldit IL-2R puzzles is to design a protein binder for the α chain of IL-2R, to block over-activation by immunotherapy drugs. The designed protein could potentially be given to cancer patients alongside normal immunotherapy to reduce its side effects. (In a different approach, other researchers have recently designed a protein binder for IL-2R β/γ that avoids the α chain altogether; that protein is currently in Phase I clinical trials.)
Selecting Foldit solutions for testing
We analyzed the solutions from all 6 IL-2R Foldit puzzles and selected 115 solutions based on their Foldit score and the binder metrics DDG, Contact Surface, and BUNS.
For this experiment, we did not factor in AlphaFold confidence when selecting designs to test, even though the AlphaFold tool was available in some of the IL-2R binder puzzles. However, most of the selected designs have AlphaFold confidence greater than 75%, so we think they have a good chance of folding.
* {box-sizing: border-box;}.column {float: left;width: 50%;padding: 10px;}.row:after {content: "";display: table;clear: both;}<div class="row"><div class="column" ><p>
2011731_c0001 Crossed Sticks
2011731_c0007 grogar7
2011731_c0008 Galaxie
2011731_c0013 stomjoh
2011731_c0018 ucad
2011731_c0019 fpc,frood66
2011731_c0024 Mike Cassidy
2011731_c0026 jobo0502
2011731_c0029 robgee
2011731_c0032 fiendish_ghoul
2011731_c0038 Bruno Kestemont
2011731_c0063 mirp
2011731_c0145 alcor29
2011731_c0151 LociOiling
2011852_c0001 Bletchley Park,BootsMcGraw
2011852_c0003 dcrwheeler
2011852_c0005 sallallami
2011852_c0009 alcor29
2011852_c0010 grogar7
2011852_c0012 ZeroLeak7
2011852_c0013 OWM3
2011852_c0017 NeLikomSheet
2011852_c0020 stomjoh
2011852_c0028 Mike Cassidy
2011852_c0029 Bruno Kestemont
2011852_c0032 Bruno Kestemont
2011852_c0035 dcrwheeler
2011852_c0044 silent gene
2011852_c0047 spvincent
2011852_c0048 robgee
2011852_c0082 dcrwheeler
2011852_c0128 ShadowTactics
2011852_c0164 BootsMcGraw
2011926_c0001 Crossed Sticks
2011926_c0002 Galaxie,grogar7
2011926_c0003 toshiue,Bruno Kestemont,Phyx
2011926_c0004 Bletchley Park,spvincent
2011926_c0007 Bruno Kestemont
2011926_c0009 Mike Cassidy
2011926_c0011 Bletchley Park
2011926_c0012 dcrwheeler
2011926_c0014 LociOiling
2011926_c0015 fpc,jausmh
2011926_c0016 robgee
2011926_c0018 HuubR,Keresto,ManVsYard
2011926_c0022 ShadowTactics
2011926_c0025 Phyx
2011926_c0028 spvincent
2011926_c0038 NinjaGreg,toshiue
2011926_c0041 fiendish_ghoul
2011926_c0051 dcrwheeler
2011926_c0061 Bletchley Park,spvincent
2011926_c0068 dcrwheeler
2011926_c0073 dcrwheeler
2011926_c0102 dcrwheeler
2011926_c0152 LociOiling
2011926_c0172 LociOiling
2011926_c0233 ShadowTactics
</p>
</div>
2011958_c0002 sgeldhof 2011958_c0005 Enzyme 2011958_c0011 Crossed Sticks 2011958_c0039 blazegeek 2011958_c0043 robgee 2011958_c0048 Enzyme 2011958_c0049 silent gene 2011958_c0051 Enzyme 2011958_c0055 Enzyme 2011958_c0056 Enzyme 2011958_c0062 LociOiling 2011958_c0063 Phyx 2011958_c0077 Enzyme 2011958_c0083 blazegeek 2011958_c0085 Todd6485577 2011958_c0094 mirp 2011958_c0097 alcor29 2011958_c0106 robgee 2011958_c0117 dcrwheeler 2011958_c0122 akaaka 2011958_c0165 Galaxie 2011958_c0176 fiendish_ghoul 2011958_c0249 mirp 2012023_c0003 sgeldhof 2012023_c0004 silent gene 2012023_c0005 nspc 2012023_c0007 Bruno Kestemont 2012023_c0020 Galaxie,grogar7 2012023_c0022 LociOiling 2012023_c0023 Galaxie 2012023_c0025 akaaka 2012023_c0029 alcor29 2012023_c0035 Bruno Kestemont 2012023_c0036 ichwilldiesennamen 2012023_c0038 Bruno Kestemont 2012023_c0071 BootsMcGraw 2012023_c0079 PLAYER_1 2012023_c0110 Galaxie 2012023_c0122 LociOiling 2012150_c0001 nspc 2012150_c0002 spdenne 2012150_c0027 LociOiling 2012150_c0032 Crossed Sticks 2012150_c0034 sgeldhof 2012150_c0043 Enzyme 2012150_c0046 Bruno Kestemont 2012150_c0048 nspc 2012150_c0053 gmn 2012150_c0054 robgee 2012150_c0055 alcor29 2012150_c0057 dcrwheeler 2012150_c0059 alcor29 2012150_c0070 Bruno Kestemont 2012150_c0093 Bruno Kestemont 2012150_c0113 zippyc137 2012150_c0133 Galaxie 2012150_c0147 LociOiling
</div>
The binding experiment will follow the same template that we’ve used before to test for binding against SARS-CoV-2 spike, MERS-CoV spike, and IL-6R.
We start with the amino acid sequence of each protein design and reverse-translate it into a DNA sequence. This custom DNA is ordered from a specialized company, and the DNA is inserted into yeast so that the yeast cells can produce our proteins and display them on the cell surface. Finally, we use fluorescent tags and microfluidics technology to sort out the yeast cells that can bind to our target protein–in this case, the α chain of IL-2R. See this blog post for a full description of the experiment.
Diversifying the testing pool
To increase our chances of success and make the most of Foldit players’ work, we again used a grafting technique to expand the diversity of the 115 Foldit designs.
We combine the binding interfaces from Foldit solutions with a large library of stable scaffold proteins to make variations of the original Foldit designs. This boosts the number of proteins in our testing pool, and allows us to more thoroughly test the binding interfaces designed by Foldit players. Our grafting method is described in more detail in this previous blog post about binders for MERS-CoV spike.
Diversifying the testing pool helps to bank against cases where the interface looks good but the binder protein fails to fold correctly. If your binder protein misfolds, then the interface residues will not be correctly positioned to bind the target. It doesn’t matter how good your DDG or Contact Surface metrics look if your protein design doesn’t fold!
By grafting the interface residues of each Foldit design onto a diversity of protein scaffolds, we generate multiple designs with the same interface, but with (potentially) very different folding behavior. This maximizes the chances that at least one of these proteins will fold correctly and present the designed interface to bind the target as intended.
We used this grafting method to generate over 1700 variations of the original 115 Foldit designs. Together with a third set of experimental re-designs (described in a future blog post), we’ll be testing a total of 1997 IL-2R binder designs from the work of Foldit players. These will be tested at the Institute for Protein Design alongside 30,000 additional designs from IPD researchers.
A big thank you goes to all Foldit players who participated in our IL-2R binder puzzles! We are very excited to get some experimental data about how these Foldit solutions behave in the real world. Keep an eye out for experiment results in the early months of 2022. In the meantime, happy folding!
