SCHEMA: Designing a Gene Delivery Vector to Neural Stem Cells

Neural stem cells (NSC) are multipotent stem cells with the ability to differentiate into neurons and glia. Since NSCs are responsible for creating new oligodendrocytes and neurons throughout an adult’s lifetime, this is a particularly attractive field of study. This occurs in the subventricular zone (SVZ), a region of the brain where some neural progenitors reside (CNS). Unfortunately, in vitro manipulation of these cells has failed to capture the signal rich environment of in vivo testing. Adeno-associated viruses (AAV) are one way in which we can test in vivo manipulation, so it has become particularly necessary to optimize regulatory mechanisms in the virions to target stem cells in the SVZ.

Although the natural evolutionary forces acting on AAVs do not drive them to transduce neural stem cells, we can engineer these non-pathogenic viruses. Through this process, we can learn more about different cells in vitro, or in a test tube, which has implications for neurogenesis, aging, development and disease. The problem, however, arises in trying to engineer novel AAV capsids to transduce specific cells in the central nervous system in vivo. In particular, natural AAV serotypes do not infect NSC of the SVZ very well. Thus, Ojala et. al found it necessary to find a variant that would act as a suitable vector.

Screen Shot 2017-11-23 at 1.05.58 AM
Figure 1: RASPP library with different <E> values

SCHEMA is an algorithm that evaluates proteins to find optimal crossing over points for mutations by providing a score based on the structural integrity of the resulting molecule. This is done by measuring the number of interactions (<E>) that are broken in the reshuffled chimeric proteins. It’s then possible to maintain a level of diversity by engineering an average level of mutations desired. Although it is possible to do this with various techniques: think DNA shuffling, Stemmer shuffling and various in vivo methods; these have trouble targeting specific cell types as well as generating diverse chimeric libraries from distant parents. In general, this also gets around the problem of random mutagenesis techniques that introduce a multitude of deleterious gene exchanges. Although DNA shuffling creates greater diversity in its resulting mutations by swapping functional groups, it can be difficult to predict the least disruptive crossover points to the protein folding. Often, such mechanisms place these crossover points in “areas of high homology,” which leads to more stable recombinants. [1] With SCHEMA, we can optimize the crossover points for DNA shuffling, and evaluate the resulting library to find chimeras with some defined fraction of folded sequences. This allows us to perform site-directed recombination with a better idea of how we can change our target sequences.

Screen Shot 2017-11-23 at 1.09.21 AM.png
Figure 2. Final selected library out of the 160 least disruptive variants is show in red.

This week’s article aimed to evaluate SCHEMA’s AAV library, and the supporting selection strategy Ojala et. al designed. A SCHEMA library with six parent serotypes (microorganisms distinguishable by their antigens) and seven crossover positions theoretically contains 1.6 million variants. After applying design specifications, Ojala et. al then applied Recombination as a Shortest Paths Problem (RASPP) in order to select the 160 least disruptive variants. 

EXTRA: RASPP is library design algorithm to optimize between the conflicting interests of <E> stability of the library energy and <m> the number of mutations. The <E> or average energy of a protein library generally increases with diversity, just as the greater <m> or number of mutations tends to lower stability in protein folding, even though it increases diversity. The algorithm minimizes the average energy of all chimeric proteins, given some peptide length constraint, and is show to be equivalent to finding the shortest path between two nodes in a network here.

Although the selected library is not the least disruptive, it shuffled key structural features in an evolutionary component with high divergence. As is visible in Figure 3 below, Ojala et. al then packaged this library into AAV virions, placing loxP on either side of the cap gene. LoxP is a DNA recombination site specifically catalysed by Cre recombinase, which is expressed by NSCs when genetically modified to do so. Thus, by packaging this library and then performing a intracerebroventricular injection, we ensure that we can isolate the genomic data from the NSC of the SVZ. I really appreciated their explanation of the loxP site placements: the first attempt placed the loxP sites around the cap, but recombinant levels were low. The loxP sites were then replaced to ensure that after bacterial propagation of the vector plasmid library, no viral proteins would be encoded. As a result, only neural stem cells expressing Cre will bind the loxP sites flanking the cap and knock out the gene.

Screen Shot 2017-11-23 at 1.11.20 AM.png
Figure 3.

By using a SCHEMA designed library and a Cre-dependent selection strategy, the AAV evolution converged upon a variant called SCH9. This variant was chosen for its morphology- loops inherited from AAV2, AAV8 and AAV9, as well as its apparent dominance of the recovered clones after Sanger sequencing analysis of the recovered clones. To further prove that SCH9 had better success at transduction, Ojala et. al compared it to its parent serotypes. Among the cell types in the SVZ, vascular cell adhesion molecule 1 (VCAM1) acts as a marker specifically for NSCs. The following figure shows the VCAM1-positive NSCs in the SVZ.

Screen Shot 2017-11-23 at 1.15.45 AM.png
Figure 4.

Previous experiments have shown that migrating neuroblasts at late time points are derived from NSCs. Ojala et. al then tested SCH9 transduction into NSCs by determining the number of neuroblasts expressing tdTomato expression 30 days after injection.

Screen Shot 2017-11-23 at 1.14.36 AM.png
Figure 5.

As you can see with the in vivo characterization of SCH9, there exists vivid improvement in NSC transduction when SCH9 is compared to the parent serotype AAV9.

The reasons for this improvement also include a selective advantage for SCH9. Whereas AAV2 used heparan sulfate proteoglycans (HSPG) on the cell surface and AAV9 used galactose, the mutations resulted in SCH9 requiring both to be blocked in order to prevent cell transduction. In addition, since DNA shuffling has been shown to disrupt neutralizing antibody epitopes, Ojala et al. investigated SCH9’s antibody resistance. To human intravenous immunoglobulin (IVIG), which is a mix of antibodies against the AAV serotypes, SCH9 required 2 to 10 fold higher antibodies before seeing an immune reaction, when compared against its parent serotype. Thus, the antibody reaction to the virus is less likely to trigger an immune response.

As a result, Ojala et. al demonstrated the success of SCHEMA-guided recombination. Using Cre-selection on the largest SCHEMA library as of the paper, they were able to present SCH9: a better way to study the regulatory mechanisms of the subventricular NSCs by gene transduction.


  1. Ojala et al., In Vivo Selection of a Computationally Designed SCHEMA AAV Library Yields a Novel Variant for Infection of Adult Neural Stem Cells in the SVZ, Molecular Therapy (2017), [2]




Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s