bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022.

The copyright holder for this preprint (which

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A flexible, efficient, and scalable platform to produce circular RNAs as

new therapeutics

Chuyun Chen 1, *, Huanhuan Wei 1, *, Kai Zhang 2,*, Zeyang Li 2, Tong Wei 1, Chenxiang
Tang 2, Yun Yang 2, #, Zefeng Wang 1,3, #

1
CAS Key Laboratory of Computational Biology, Shanghai Institute of Nutrition and
Health, University of Chinese Academy of Sciences, Chinese Academy of Sciences,
Shanghai 200031, China
2
CirCode Biomedicine Inc. Shanghai, China
3
Bio-Med Big Data Center, Shanghai Institute of Nutrition and Health, University of
Chinese Academy of Sciences, Chinese Academy of Sciences, Shanghai, 200031,
China

* These authors contributed equally to this study

# Corresponding to: wangzefeng@picb.ac.cn, yangyun@circodebio.com

Competing Interest Statement: CirCode has filed a patent application related to the
technology described in this paper.
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Abstract:
Messenger RNA (mRNA) has recently emerged as a new drug modality with great
therapeutic potential. However, linear mRNAs are relatively unstable and also require
base modification to reduce their immunogenicity, imposing a limitation to the broad
application. With improved stability, the circular RNA (circRNA) presents a better
alternative for prolonged expression of the proteins, however the in vitro circularization of
RNA at a large scale is technically challenging. Here we developed a new self-catalyzed
system to efficiently produce circRNAs in a co-transcriptional fashion. By rational
sequence design, we can efficiently produce scarless circRNAs that do not contain
foreign sequences. The resulting circRNAs are very stable and have low immunogenicity,
enabling prolonged protein translation in different cells without cellular toxicity. The
circRNAs generated from this platform can be encapsulated in lipid nanoparticles and
efficiently delivered into mice to direct robust protein expression. Finally, the circRNAs
encoding RBD of SARS-CoV-2 S protein induced strong antibody productions, with
neutralization antibody titers higher than the preclinical data from the linear mRNAs.
Collectively, this study provided a general platform for efficient production of circRNAs,
demonstrating the potential of circRNAs as the new generation of mRNA therapy.

Keywords: Circular RNA, mRNA therapy, Cap-independent translation, Self-splicing

intron, In vitro RNA synthesis

2
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Introduction:
mRNA has recently come into focus as a new drug modality with great therapeutic
potential, as mRNA-based therapy can theoretically use any protein as the active
pharmaceutical ingredient 1. Despite the clear demonstration of efficacy for infectious
disease and cancer vaccines, application of mRNA for non-vaccine therapeutics has been
limited by the duration of expression, stability, immunogenicity, and the ability to control
cell type-specific expression. In addition, there are technical challenges associated with
mRNA production, modification, and extra-hepatic delivery. Circular RNAs (circRNAs)
are covalently closed RNA molecules generated mainly from pre-mRNA back-splicing 2,
and have also been found to function as mRNAs to direct protein synthesis inside cells 3.
As a new generation of therapeutic mRNAs, circRNAs may address some of the
4,5
limitations of linear mRNA including increased stability and reduced immunogenicity .
Therefore, translatable circRNAs have the potential to expand the application of mRNA-
based therapy, especially when a prolonged expression of the target protein is required.
A major limitation for the borad application of circRNAs is the in vitro production of
circRNAs, especially in the step of RNA circularization after the in vitro transcription (IVT)
6
. The most commonly used methods for RNA circularization include the end-ligation of
linear RNA with T4 RNA ligase 7, or the permuted intron-exon (PIE) splicing strategy using
the self-catalyzed group I introns 8,9. However the direct ligation with RNA ligases usually
needs to use a “splint” sequence to connect the free ends, and thus is technically
challenging to scale up the production due to the formation of concatemers 6. In addition,
introducing the ligase protein will also increase the complexity in the product purification.
On the other hand, the PIE method will introduce a foreign fragment into the circRNA as
9,10
a “scar” sequence , which may distort the structures of circRNAs to provoke innate
11
immune responses and complicate the future approval for therapeutic applications.
Therefore reliable circRNA production methods are needed for the therapeutic application
of circRNAs on a large scale.
In this study, we developed a new and scalable technology to engineer and
produce circRNAs that efficiently direct protein translation. Our design is based on the
self-splicing group II intron to achieve co-transcriptional circularization of scarless
circRNAs with high efficiency. The resulting circRNAs can be purified and transfected

3
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

into cells to direct robust cap-independent translation from the internal ribosomal entry
sites (IRESs). Using this platform, we generated a series of circRNAs containing different
genes to direct the translation in cultured cells and in mice, and the purified circRNAs did
not cause deleterious innate immune responses. The therapeutic application of this
system was exemplified by engineering circRNAs that encode the RBD of the SARS-
CoV-2 S protein to induce robust antibody production in mice. Collectively, this study
provided an efficient and scalable platform for circRNA engineering and production, which
has broad application in the new generation of mRNA therapy.

Results:
Design and production of circular RNAs in vitro.
To efficiently produce circular RNAs, we take advantage of the self-catalyzed
splicing reaction by group II introns, which are mobile genetic elements found mainly in
bacterial and organellar genomes 12. All group II introns have six structural domains (D1
to D6), of which the domain 1 (D1) is the largest domain and contains several short exon
binding sites (EBS) to determine the splicing specificity (Fig. 1A). Specifically, the EBS1
in D1 can pair with the intron binding site (IBS1) at the 5’- splicing junction to determine
the exact position of splicing, whereas the 3’- splicing junction is determined by the paring
of IBS3 with the EBS3 in D1 (for group IIB, IIC and IIE/F introns) or the δ base of D1 (for

group IIA introns). The domains 2 and 3 play key roles in assembling the active intron
structure and stimulating splicing reaction, whereas the D4 is a stem-loop structure with
a long loop containing the ORF of maturase. The highly conserved D5 is the heart of
13
active site for self-splicing , whereas the D6 contains a bulged adenosine as the
branching site (Fig. 1A). The in vitro self-splicing of group II intron requires only correct
folding of intronic RNA structure and Mg2+ 14, however the in vivo splicing requires the
assistance of the maturase 15.
Based on this domain configuration, we first split the group II self-splicing intron
from the surface layer protein of Clostridium tetani (ctSLP) 16 at the loop region of different
domains (loops of D1, D3, D4), generating a series of split-intron systems that contain a
customized exon flanked by the two half-introns (Fig. S1 and Fig. 1A). A fraction of the
stem region was separated and placed into each end of the resulting RNA, thus forming

4
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

a complementary structure that helps the folding of the active intron (Fig. S1 and Fig. 1A,
indicated with green lines). Two short 6-nt sequences in the exons, IBS3 (intron binding
site 3) and IBS1, were included at each splicing junction to provide long-range interactions
with the EBS3 and EBS1 of the intron. Upon the in vitro run-off transcription, the resulting
RNA precursor could be self-spliced by the ctSLP group II intron to produce a circular
RNA of the customized exon and a branched intron RNA. By including an IRES sequence
at upstream of a gene of interest, the resulting circular RNA can function as an mRNA to
direct protein synthesis through cap-independent translation (Fig 1A). We named this
system as the circular coding RNA (CirCode), which could serve as a general platform to
produce any given circRNA for protein translation.
To validate this design, we included the ORF of the Renilla luciferase gene
together with an IRES into the customized exon, and used IVT to generate this intron-
exon-intron RNA precursor. We observed that all the resulting RNA precursors (D1, D3,
or D4 split intron) can be self-spliced in vitro to produce an extra band corresponding to
circRNAs, with the D4 split intron being the most active (Fig. S1). This design was
therefore selected for further study. We first mutated the splice junction (at the IBS1) of
the D4 split intron, and found that the mutated RNA precursor failed to produce the
circRNAs (Fig. 1B). Next, we used two different methods to confirm that the additional
band below the precursor RNA is indeed a circRNA, which cannot be extended by tailing
reaction by poly-A polymerase and is resistant to the digestion by RNase R treatment
(Fig. 1B). In addition, we gel-purified the circRNAs, validated its identity using RNase H
digestion (Fig. 1C) and direct sequencing across the junction (Fig. 1D). Finally, we
optimized the reaction conditions using different concentrations of MgCl2 and NaCl in the
circularization buffer following the IVT step (see method), and found that using 10-20 mM
Mg2+ with 50-100 mM NaCl is the optimal reaction condition for the RNA circularization
(Fig, 1E). We also observed certain degree of RNA circularization (~30%) even when the
circularization buffer does not contain any Mg2+, presumably because the RNA
circularization happened co-transcriptionally in the IVT buffer containing 24mM MgCl2
(see methods).

Platform optimization of circRNA production and translation

5
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

The PIE methods using group I introns for circRNA production also introduced an
extraneous sequence from T4 bacteriophage or Anabaena into the final products 9,10. This
“scar” sequence is usually around 80-180 nt long, which limits the design flexibility of
target circRNAs and may introduce some unwanted side-effect during drug development.
Our initial design used two short (6-nt) sequences, IBS1 and IBS3, for the intron-exon
recognition, which leaves a shorter “scar” of 12-nt. To reduce the potential interference
by the scar sequence, we further modified the design by changing the exon binding sites
in the D1 domain, making the EBS1 and sequence upstream of EBS1 to respectively form
base pairs with the 3’ and 5’ end of the circular exon (Fig. 2A, left). This new design
enabled the generation of “scarless circRNAs” without extraneous sequences. We
validated this design by generating two different circRNAs encoding the EGFP and Rluc-
P2A (Fig. 2A, right), and observed the efficient circularization of scarless circRNA in
different ion conditions. The scarless self-splicing of circRNAs was further confirmed by
the sequencing of final circRNA products (Fig. 2A, bottom of the right). In summary, we
have engineered the CirCode system that can efficiently produce scarless circRNAs of
any sequences, providing a general platform for using circRNA as different therapeutic
tools.
The previous study using the PIE method suggested that addition of a short spacer
region before the IRES may assist the correct folding of the IRESs and the active introns
9
, and thus we introduced several versions of spacer sequences containing IRES-like
17
short elements at each end of the circular exon to optimize their circularization and
translation efficiency. We found that different spacers indeed affected the RNA
circularization efficiency, which ranged from 47% (for SP5) to 83% (for SP4) (Fig. 2B). As
expected, different spacers containing short IRES-like elements also affected the
translation activities of circRNAs (Fig. S2). We further transfected two circRNAs with
different spacers into three cell lines at different doses, and observed dose dependent
increase in protein production as judged by the luciferase activity assay (Fig. 2C). Finally,
we examined the time course of protein production upon circRNA transfection, and found
that the active protein can be detected in six hours and reach the expression peak of
expression at 48 hours after transfection (Fig. 2D).

6
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

circRNAs mediate prolonged protein production

A major advantage of circular mRNAs is their superb stability because of lacking
the free ends, therefore the circRNA should have a good shelf life for protein expression
compared to the linear counterparts. To directly test this, we synthesized both linear and
circular mRNA encoding the Gaussia luciferase (Gluc), and stored them parallelly in
nuclease-free water at room temperature for different days before transfecting them into
293T cells. We found that the activity of the circRNA to direct protein translation is
essentially unchanged during the two weeks, whereas the linear mRNA lost about half of
its activity by day 3 of the storage (Fig. 3A), suggesting that circRNAs can be stably stored
in room temperature. In addition, we found that once transfected into cells, the protein
production from linear mRNA decreased rapidly (with a >80% reduction on day 2),
whereas the translation from circRNAs lasted 5-7 days (Fig. 3B). The prolonged protein
translation is also consistent with our previous results using back-spliced circRNA
reporter 3. It is also worth noting that, in this experiment, the protein production from
circRNAs may also be suppressed because the cells reached the stationary phase at the
end (Fig. 2B), as we only replaced the culture medium without splitting the cells during
the entire week of the culture.
We further compared the protein production from the linear mRNAs and the
circRNAs produced using the PIE method or the new CirCode system. As a control, the
unmodified linear mRNAs were generated using IVT with the same coding sequences of
Gluc, and transfected into 293 T cells in parallel with the two types of circRNAs. We found
a more robust expression of proteins from both circRNAs compared to the linear mRNA
(Fig. 3C, top), supporting the previous reports using PIE circRNAs 9. An even larger
increase in protein production was observed in circRNAs when we measured the
accumulated luciferase activity over the span of 6 days (Fig. 3C, bottom), presumably
because of the superior stability of circRNAs. In addition, our data suggested that the
circRNAs generated from two different methods showed similar ability in the direct protein
translation.

Purified circRNAs can direct robust translation of target proteins

7
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

The mRNA purity was found to be a key factor for the protein production and
induction of innate immunity, as the removal of dsRNA by HPLC can eliminate immune
18
activation and improves translation of linear nucleoside-modified mRNA . However,
there are some debates on the immunogenicity of circRNAs. While an early report
suggested that in vitro synthesized circRNAs are more prone to induce cellular immune
19
response than the linear RNA , it was later reported that purification of circRNAs from
byproducts of IVT and circularization reactions, including dsRNA, linear RNA fragments
and triphosphate-RNAs, can eliminate the cellular toxicity and immunogenicity of the
circRNAs 4,5. A recent study also suggested that the sequence identity and structure are
the main determinant of cellular immunity of circRNAs, as the circRNAs produced by
11
different methods showed different immunogenicity . To examine if the circRNAs
produced using CirCode platform can induce innate immune response and cell toxicity,
we purified the circRNAs with gel purification or HPLC (Fig. 3D), and measured if the
circRNAs can induce cellular immune response upon transfection of circRNAs. We found
that transfection of the unpurified circRNAs cause a significant amount of cell death,
whereas the purified circRNAs did not show detectable cell toxicity compared to mock
transfection (Fig. 3E). In addition, compared to the unpurified circRNAs that stimulated
innate immune response by inducing RIG-I and IFN-B1 (interferon-β1), the purified
circRNAs showed minimal immunogenicity (Fig. 3F), supporting the previous observation
that the immunogenicity of circRNAs was mainly caused by the by-products from IVT
reaction 4,5.

LNP encapsulated circRNAs direct robust protein production in mouse

An important question for the therapeutic application of circRNAs is whether the
circRNA production can be scaled up reliably and how is the reproducibility between
different batches of production. Because the CirCode platform uses self-splicing intron
for RNA circularization without the involvement of RNA ligase or additional co-activators,
the procedure is relatively simple to scale up. To test the scalability of this system, we
proportionally expanded the IVT and circularization reaction for 50 fold (from 20 μl to 1
ml), and found that the high circularization efficiency (~70%) stayed essentially
unchanged while the total amount of RNA products reached 7.5 mg in a single reaction

8
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

(Fig. 4A). In addition, we found that the liquid chromatography production and purification
of the circRNAs were scalable and highly reproducible across different batches (Fig. S3
and Fig. 4B), laying the ground for the in vivo application of the circRNAs.
We further generated circRNAs encapsulated with lipid nanoparticles (LNP) for
their in vivo delivery (Fig. 4C). The circRNA in the aqueous solution was packed by
ionizable cationic lipids, which formed a nanoparticle with other lipid components such as
DMG-PEG2000 and cholesterol (see methods), resulting in a ~95% encapsulate
efficiency with the effective diameter at ~80 nm. We have tested three different ionizable
cationic lipids in our formulation to encapsulate the circRNAs encoding Gluc, and the
resulting LNP-circRNAs were injected into BALB/c mice through intramuscular (IM) or
intraperitoneal (IP) injection (n=3 for each experimental group). Three formulations using
different ionizable cationic lipids (MC3, SM-102 and ALC-0315) were tested in this
experiment. The expression of luciferase was assayed either using the luciferase
luminescence assay with serum (Fig. 4D) or bioluminescence imaging of the animals (Fig.
4E). We found a robust expression of luciferase from two different formulations of LNP-
circRNAs, suggesting that the circRNAs produced through CirCode system can reliably
induce in vivo protein expression.

Generation of a potential circRNA vaccine for SARS-CoV-2

We further tested the application of circRNAs in mRNA therapy by engineering the
circRNAs encoding the receptor binding domain (RBD) from the S protein of SARS-CoV-
2, which can potentially be used to produce mRNA vaccine. Based on previous reports,
20
two different antigen designs were constructed into the circRNAs (Fig. 5A top) . The
first one used a single RBD fused with the foldon from T4 fibritin that enables trimerization
21,22
of RBD . The other used a RBD dimmer that was shown to induce strong antibody
production in pilot study 20,23. We have designed the coding sequences of both proteins
with an engineered IRES, and constructed the circRNA vectors. The purified circRNAs
were validated with capillary electrophoresis (Fig. 5A, left), and translation of the circRNA-
encoded proteins was further validated by transfecting into 293 cells and detected by
western blot antigen production (Fig. 5A, right).

9
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

We used two different formulations to produce the LNP-circRNA particles, and

achieved high encapsulation efficiency (>90%) with typical nanoparticle size at 90-100
nm (Fig. 5B showing the representative result using SM-102 formulation). The LNP-
circRNAs were inoculated into the BALB/c mice with two separated IM injections, and the
blood samples were collected at 2 weeks after each inoculation for further analysis (Fig.
5C). We tested several formulations and circRNA designs, however the results from the
LNP-circRNA-RBP using SM-102 formulation were further examined for a better
comparison with published results. A flow cytometry antibody panel was designed to
identify naïve B cells (CD19+/IgD+/CD27−), total memory B cells (CD19+/CD27+),
including an unswitched IgD+ population and a switched IgD− population, plasma cells
(CD19+/IgD−/CD38+/CD27+), and transitional B cell (CD19+/IgDdim/CD38+). To
determine whether LNP-circRNA-RBP vaccination induced the activation and expansion
of antigen-specific B cells, we measured the frequency of RBD-binding B cells using
Alexa 647 labeled RBD (RBD-Alexa 647). A large fraction of RBD-specific lymphocytes
were detected in the memory B-cells (CD19+/CD27+), including an RBD specific switched
B-cell population (CD19+CD27+ IgD- RBD+) and an RBD specific unswitched memory B-
cell population (CD19+CD27+ IgD+ RBD+) (Fig. 5D, 5E), indicative of a long-lasting
immune response induced by antigen. In particular, >70% of CD19+CD27+ B-cells
expressed RBD-specific antibody (i.e., RBD+), and ~50% of the CD19+CD27+ B-cells were
also IgD+/RBD+ (Fig. 5E), suggesting a strong RBD-specific memory B cell response
elicited by the LNP-circRNA vaccination. As a control, the injection of LNP alone did not
induce activation of RBD-specific memory B cells, confirming that the observed response
was specifically induced by the circRNA.
To test the activity of the RBD antibody in mouse serum, we next performed an
antibody blocking assay to examine if the mouse serum can block the binding of Alexa
647-labeled RDB to the 293 T cells stably expressing human angiotensin I-converting
enzyme 2 (hACE2) (Fig. 6A). We found that the serum can effectively block the binding
of three different RBD variants (RBDwild type, RBDdelta, and RBDomicron) to the cell surface
at a modest dilution (1:20). Given that we used a pretty high concentration of RBD (0.5
μg/ml), this protection is quite impressive. Even in a very high dilution ratio (1:200), the
serum still completely blocked the binding of the wild-type RBD and significantly reduced

10
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

the binding of the RBD of the Delta variant. However, the serum did not block the binding
of the omicron RBD at the same dilution ratio, suggesting a diminished blocking activity
for the omicron variant. This result is consistent with the finding that the current mRNA
vaccines based on the wild type S protein are weak in protecting against the omicron
variant 24.
We further measured the antibody titers after each inoculation, and found a robust
production of IgG production against RBD (Fig. 6B-6C). The binding antibody titer was
close to 105 at two weeks after the first shot and reached near 108 after the second shot,
which is ~10 times higher than the reported mouse results using the LNP encapsulated
linear mRNAs with a similar formulation 25 or using another circRNA design 26. In addition,
the ratio of IgG2/IgG1 is around 0.75, suggesting a balanced Th1/Th2 immune response.
Similar immune response was previously reported using linear mRNA vaccine mRNA-
25
1273 , implying circRNA vaccine may have a low risk in vaccine associated enhanced
disease and a potential good safety profile. Finally, we measured the pseudo-neutralizing
antibody titer using the protection assay for SARS-CoV-2 pseudovirus, and found a strong
protection with a pseudo-neutralizing titer close to 105 (Fig. 6D). Collectively, our
preliminary data in the mouse model suggested that the CirCode platform had a strong
performance in the design and generation of circRNA vaccines against SARS-CoV-2,
which can potentially be expanded to the vaccine development for other viral pathogens.

11
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Discussion:
The approval of mRNA vaccines against Covid-19 pandemics has pushed the
mRNA therapy into the front stage. In addition to the vaccine, mRNA drugs had shown
potential in the treatment of a variety of diseases, including cancers, rare genetic
disorders, and neuronal diseases 1. However, linear mRNAs are intrinsically unstable,
which limited their application. Several features of circRNA make them an appealing
candidate as the new generation of mRNA drugs. First, circRNAs have increased stability
both in vitro and in vivo, which enables a prolonged expression of target genes. Second,
the chemical modification of circRNAs seems to be unnecessary, as the unmodified
circRNAs can be reliably translated without inducing the innate immune response (Fig.
4,5
3E) . In addition, circRNAs use IRESs or IRES-like sequences to initiate translation in
3,17
a cap-independent fashion , which may allow cell-type specific translation through
cellular trans-acting protein factors 4,17,27. Despite these advantages, there are technical
challenges to overcome before circRNAs become a general gene expression platform,
including circRNA manufacturing, purification and quality control, optimization of IRES
mediated translation, as well as in vivo delivery. We have developed a new general
platform for circRNA design and production, which can be integrated with the sequence
design and IRES optimization to achieve a high efficiency of protein translation from
circRNAs.
The inverse splicing of both group I and group II self-catalyzed introns have been
8,10,28
discovered more than two decades ago to produce a small amount of circRNAs .
However, the circularization efficiency is very low in these early studies, and the circRNAs
can only be produced from their intrinsic exon sequences. Recently the PIE method has
been optimized to produce circRNAs containing customized target sequences 9, however
a large fragment of the exons adjacent to the group I intron have to be included in the
final circRNA product as a scar sequence. In this study, by screening a series of group
II introns and analyzing the intron structures, we were able to engineer a new split-intron
system to produce circRNAs containing a short foreign fragment of 12 nt or do not contain
any foreign fragment (i.e., scarless). The main reason for the scarless inverse self-
splicing is that we were able to rearrange the structure of group II intron in a modular
fashion to make it accommodate different exon sequences at the splice junction. In

12
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

addition, the target sequence seems to play a less role in affecting the folding of group II
intron, because our method can achieve decent circularization efficiency (>50%) for most
genes. With additional optimization, we can routinely obtain a circularization efficiency of
60-90% with different designs of the junction sequences.
Because of superb stability and low immunogenicity, the base modification seems
to be unnecessary for circRNAs. In addition, circRNA production does not require the step
of 5’ capping, which can significantly reduce the complexity and cost for circRNA
production. The new method in this study has enabled co-transcriptional circularization
without the supplement of extra GTP, which further streamlines the production process.
The emerging new variants of SASR-coV-2 presented a challenge for new mRNA
vaccines. Because the levels of antibodies induced by mRNA were found to decrease
29
gradually, the repeat booster of vaccination is usually needed for effective protection .
Using circRNA as a new vaccination platform has been shown to be a promising
26
alternative to the linear mRNA vaccine , probably because of its superb stability.
Compared to the linear mRNA, the antigen production from cap-independent translation
of circRNA may peak slower but last longer, resulting in more antigen production by
accumulative protein expression. It remains unclear how this different dynamic of antigen
expression can affect the antibody response, however our early results showed a very
strong induction of memory B-cells. Additional experiments will be needed to optimize
the circRNAs for vaccine production, including the comparison of different antigen
designs and optimization for the formulation.
There are still many technical questions remain to be solved for circRNAs as a new
therapeutic reagent, including the sequence optimization for circRNA translation and the
specific methods for in vivo delivery. Although the canonical IRESs that drive circRNA
translation is mainly derived from viral sequences, recent studies found that many
additional sequence elements can function as IRES-like elements or regulatory elements
17,30
to promote IRES activity . Therefore the sequence optimization for circRNA
translation will likely be different from the optimization of UTR sequences in linear mRNAs.
In addition, the superior stability of circRNAs may make them more tolerable to additional
formulation methods of RNA delivery. We found that most formulations for mRNA
delivery can efficiently deliver circRNAs, however it is both intriguing and important to

13
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

study if certain methods unfit for linear mRNA delivery may actually work well for circRNAs.
Overcoming these technical hurdles will certainly improve our understanding on the
translation regulation of circRNAs, as well as the circRNA transport between or inside
cells.
In summary, we have developed a new system for efficient and scalable production
of circRNAs in a large scale, which is suitable for industry level applications. The resulting
circRNAs can be engineered to direct robust protein translation, providing a new platform
of mRNA therapy with improved stability and reduced immunogenicity. In addition, this
system can also be used to generate various circRNAs with different therapeutic activities,
such as miRNA sponge or RNA aptamers. The continuous improvement of this platform
should help to take circRNA technology into various clinic applications in the near future.

14
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Methods:

Plasmid construction
The fragments of the group II intron in Clostridium tetani (CTE) and IRES
sequences were chemically synthesized from GENEWIZ, and selected mutations were
introduced in the synthesis to make it recognize different exons. Different protein coding
fragments (ORF) were amplified by PCR and merged with the IRES fragments using PCR
reactions. These fragments were cloned into the NheI and XbaI digested backbone
containing T7 RNA polymerase promoter and terminator using Gibson assembly (see
supplementary data 1 for the sequences of these DNA segments and primers).

RNA synthesis and circularization

The plasmid DNAs were linearized with XbaI digestion and purified with Post PCR
cleanup kits from Qiagen. The linearized DNAs were used as a template for in vitro
transcription with T7 RiboMAX™ Large-Scale RNA Production System (Promega, P1320)
in the presence of unmodified NTPs. After DNase I treatment, the RNA products were
column purified with RNA Clean and Concentrator Kit (ZYMO research, R1013) to remove
excess NTP and other salts in IVT buffer, as well as the possible small RNA fragments
generated during IVT. In some experiments, the purified RNA was further circularized in
a new circularization buffer. The RNA was first heated to 75°C for 5 min and quickly
cooled down to 45°C, after which a buffer including indicated magnesium and sodium
was added to a final concentration: 50mM Tris-HCl at pH 7.5, 50/100mM NaCl, 0-40mM
MgCl2, and was then heated at 53°C for indicated time for circularization. The best-
optimized reaction condition including concentration of magnesium and sodium, and
incubation time at 53°C, was selected for further experiments.

CircRNA identification
For the poly A tailing and RNase R treatment, the total RNAs from IVT were purified
by RNA cleanup columns, and then treated with E. coli Poly A Polymerase (NEB, M0276S)
following the manufactory instruction. This step will add a poly-A tail to the free end of
the unspliced linear RNA precursor. After Poly A-tailing, the purified RNAs were digested

15
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

by RNase R exoribonuclease (Lucigen, RNR07520) following the manufacturer’s

instructions, and enriched circRNAs were purified by column.
For the RNase H nicking assay, the RNase R enriched circRNAs were incubated
with a 24-nt ssDNA probe at 1:20 ratio, The RNase H buffer was added to the DNA-RNA
mixture immediately. Subsequently, the mixture slowly cool to room temperature. After
annealing, RNase H (Thermo Scientific, EN0201) was added to the mixture for 20min at
37 °C. The sequence of the ssDNA probe is 5’- TGGTGCTCGTAGGAGTAGTGAAAG-3’.
For all the gel electrophoresis, the total RNA from IVT samples was separated on
a low melting point agarose gel (Sigma Aldrich, A4018) at 120V using ice-cold DEPC-
treated MOPS butter. To gel purify the circRNA, the circRNA band was cut and purified
with Zymoclean Gel RNA Recovery Kit (ZYMO research, R1011).
To analyze the circRNA with sequencing, the gel-purified RNA was reverse
transcribed into cDNA using a PrimeScript RT Reagent Kit with random primers (TAKARA,
RR037B), followed by PCR with primers that can amplify transcripts across the splice
junction. The PCR products were sequenced using Sanger sequencing to validate the
backsplice junction of the circular RNA.

HPLC purification and electrophoresis of RNA

To obtain the high-quality circRNA, spin column purified DNase I-treated RNA from
IVT was resolved with high performance liquid chromatography (HPLC). For a small-scale
preparation with SHIMADZU LC-20A (Kyoto, Japan), 40μg RNA was loaded onto a
4.6×300 mm size exclusion column (Waters XBridge, BEH450A, 450Å pore diameter,
3.5μm particle size) and eluted with mobile phase containing 10mM Tris, 1mM EDTA,
75mM PB, pH7.4 at 25°C with flow rate 0.5ml/min. For a large-scale preparation with
Sepure SDL-30 (Suzhou, China), 4mg RNA was loaded each run onto a 30×300mm SEC
column (Sepax, SRT SEC-1000A, 1000Å pore diameter, 5.0μm particle size, Suzhou,
China) with mobile phase containing 10mM Tris, 1mM EDTA, 75mM PB, pH7.4 at 25°C
with flow rate 10ml/min. Fractions were collected as indicated and testified with agarose
gel electrophoresis.
Circular RNAs purified from large-scale production were further analyzed with
capillary electrophoresis with Agilent 2100 Bioanalyzer in the RNA mode. Samples were

16
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

diluted to a propriate concentration and analyzed according to the manufactory’s

instructions.

Measurement of the translation products from circRNAs

Cells were seeded into 24-well plates one day before transfection. The purified
circRNAs are transfected into cells using Lipofectamine Messenger Max (Invitrogen,
LMRNA001) according to the manufacturer’s manual. After transfection, cells were
cultured at 37°C for 24h. The cell lysis and supernatant were collected for luminescence
assay using the Dual-Luciferase® Reporter Assay System (Promega, E1910).

LNP Production Process

The circular RNA was encapsulated in a lipid nanoparticle via the NanoAssemblr
Ignite system as previously described 25,31. In brief, an aqueous solution of circRNA at pH
4.0 is rapidly mixed with a lipid mixture dissolved in ethanol, which contains different
ionizable cationic lipid, distearoylphosphatidylcholine (DSPC), DMG-PEG2000, and
cholesterol. The ratios for the lipid mixture are MC3:DSPC:Cholesterol:PEG-
2000=50:10:38.5:1.5 for formulation 1, SM-102:DSPC:Cholesterol:PEG-2000 =
50:10:38.5:1.5 for formulation 2, and ALC-0315:DSPC:Cholesterol:ALC-0159 =
46.3:9.4:42.7:1.6 for formulation 3. The resulting LNP mixture was then dialyzed against
PBS and stored at -80 °C at a concentration of 0.5 µg/µl for further application.

Administration of LNP-circRNAs in mice

Female BALB/C mice aged 8 weeks were purchased from Shanghai Model
Organisms Center. 20 µg of LNP-circRNAs in PBS were administrated into mice
intramuscularly with 3/10 insulin syringes (BD biosciences). The serum was collected 24
hours after the administration of LNP, and 50 µl serum was used for Luciferase activity
assay in vitro. Bioluminescence imaging was performed with an IVIS Spectrum (Roper
Scientific). 24 hours after Gluc-LNP injection, 2 mg/kg of Coelenterazine
(MedChemExpress,MCE) was administrated to mice intraperitoneally. Mice were then
anesthetized after receiving the substrates in a chamber with 2.5% isoflurane (RWD Life
Science Co.) and placed on the imaging platform while being maintained on 2% isoflurane

17
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

via a nose cone. Mice were imaged 5 minutes post substrate injection with 30 seconds
exposure time to ensure the signal was effectively and sufficiently acquired.
For immunogenicity studies, 8-week-old female BALB/c mice (Shanghai Model
Organisms Center, lnc) were used. 10µg of CircRNA-RBD-LNP were diluted in 50µl
1XPBS and intramuscularly administrated into the mice’s same hind leg for both prime
and boost shots. Mice in the control groups received PBS and empty LNPs. The blood
samples were collected 2 weeks after prime and boost shots. Mice spleen were also
harvested at the endpoint (2 weeks after boost) for immunostaining and flow cytometry.

B cell surface staining for flow cytometry

The spleenocyte suspensions from whole spleens were generated using a tissue
dissociator (RWD Life Science) followed by 70-μm filtration. Cells were isolated and
resuspended in R10 media (RPMI 1640 (Gbico) supplemented with Pen-Strep antibiotic,
10% HI-FBS, 1%Glutamax, and1% HEPES) followed by density gradient centrifugation
using Fico/Lite-LM medium (R&D Systems). Cells were washed and stained with Fixable
Viability Stain 510 (BD Pharmingen) at 4℃ in the dark at least 5 min in brilliant stain buffer.

Cells were then washed with FC buffer (PBS supplemented with 2% HI-FBS and 0.05%
NaN3 ) and resuspended in Fixation Buffer (BD Pharmingen)for 5 min at 4℃ in the dark.

Cells were then blocked with Fc Block（BD, clone 2.4G2; 1:100）at 4℃ for 5 min, cells
were labeled with the following antibodies: CD19-PE (BD), CD38-BV421(BD), CD27-
BV786(BD), IgD-BV650(Biolegend), RBD-Alexa 647 in brilliant stain buffer (BD) for 20min.
Cells were then washed and resuspended in FC buffer before running on a LSRFortessa
flow cytometer (BD).and anlayzed using FlowJo software version 10.

ELISA
Elisa plates (Corning) were coated with 100ng per well of recombinant Spike RBD
protein in 1×coating buffer (Biolegend) overnight at 4°C. After standard washes and
blocks, plates were incubated with serial dilutions of heat-inactivated sera for 2 hours at
room temperature followed by standard washes. Anti-mouse IgG1, IgG2a, or IgG2c-
horseradish peroxidase conjugates (Abcam) were used as secondary antibodies. After 1

18
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

hour incubation, plates were washed and 3,5,3′5′-tetramethylbenzidine (TMB) substrate

(Beyotime) were used as the substrate to detect the antibody specific signal. The reaction
was stopped using stop solution (Beyotime) and the final result was read at 450 nm on
an Elisa plate reader (Biotek). End-point titers were calculated as the dilution that emitted
an optical density exceeding 4 times of background and calculated in Graphpad Prism
software version 8 and R-Studio.

Lentivirus-based pseudovirus-neutralization assay

SARS-CoV-2 lentivirus-based pseudovirus encoding a luciferase reporter (Genwiz) was
used for neutralization assay following a previously described method (cite). In brief, serial
dilutions of heat-inactivated sera were mixed with pseudoviruses, incubated, and then
added to hACE-2 and human transmembrane protease serine 2 (TMPRSS2) co-
expressing HEK293 cells followed by a period of incubation time. Cells in each well were
lysed and measured for luciferase activity (in relative light units (RLU)). Neutralization
was calculated considering uninfected cells as 100% neutralization and cells infected with
only pseudovirus as 0% neutralization. IC50 titers were determined using a log (agonist)
vs normalized-response (variable slope) nonlinear function in Graphpad Prism software
version 8 (GraphPad) and R-Studio.

Confocal Immunofluorescent analysis of RBD binding

Dilutions of heat-inactivated sera were mixed with Alexa-647 labeled RBD protein
followed by 5 minutes at room temperature. hACE-2-expressing HEK293 cells were
seeded on a Nunc Lab-Tek Chamber Slide system (Thermo Fisher) and cultured
overnight. Cells were washed and stained with Hoechst for nuclear staining followed by
staining with the mixture of sera and Alexa-647 labeled RBD. Cells were washed and
imaged using confocal microscopy (Sony).

19
bioRxiv preprint doi: https://doi.org/10.1101/2022.05.31.494115; this version posted June 1, 2022. The copyright holder for this preprint (which
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

ACKNOWLEDGEMENT

The authors want to thank Dr. Leaf Huang for his suggestions and comments on the

manuscript. This work is partially supported by the National Natural Science Foundation

of China to Z.W. (91940303 and 31730110) and Y.Y. (31870814). Z.W. is also sponsored

by the type A CAS Pioneer 100-Talent program, Y.Y. is also sponsored by the Youth

Innovation Promotion Association CAS, SA-SIBA Scholarship Program and Shanghai

Science and Technology Committee Rising-Star Program (19QA1410500). Z.W. is also

supported by the Starry Night Science Fund at Shanghai Institute for Advanced Study of

Zhejiang University (SN-ZJU-SIAS-009).

20
Figures and Legends:

A EBS2
Bulged
δ adenosine

EBS3 EBS1 A

IR
5’ Exon 3’ Exon

ES
IBS2 IBS1
D1 D2 D3 D4 D5 D6 IBS3

I
G
O
A

IBS 1
+

IB
lytic

S3
A
cata
Auto plicing
IRES GOI s
D5 D6 D1 D2 D3 self-
IBS3 IBS1

Poly(A) RNase R C
B polymerase tailing treatment
Linear RNA
IRES Luc
AAAAA 763nt 541nt
Probe (25nt)

IR
Circular RNA 1329nt

ES
c
wt S1

wt S1

S1
-IB

-IB

-IB

u
L
ut

Probe (25nt)
ut

ut
wt
m

Poly A tailing - - + + + +
Linear RNA
RNase R - - - - + + control circular RNA
RNase H + - + -
1329nt
precursor RNA
763nt
CircRNA 541nt

Free introns

E 15min 5min
D IR
PCR 50mM NaCl 100mM NaCl 50mM NaCl 100mM NaCl
ES

c
primer [Mg 2+ ] (mM) 0 2 5 10 20 2 5 10 20 2 5 10 20 2 5 10 20
u
L

Splice
Junction
IRES
Circularization

80
efficiency

60
40
20
0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Fig.1 Engineer of a new circRNA production system.

(A) Schematic diagram for the design of CirCode system. The autocatalytic self-splicing group II
intron was split into two fragments at the D4 domain, and a customized exons containing
IRES and coding region of a gene of interest (GOI) were inserted between the split intron.
(B) The circRNA is in vitro synthesized and analyzed with agarose gel. IBS1 was mutated to
disable self-splicing. After IVT, circularized RNAs were confirmed by Poly A tailing and RNase
R treatment.

21
(C) Further confirmation of RNA circularization by RNase H nicking assay. Linear RNA containing
the same sequence in circular RNA could be digested into 3 fragments by RNase H.
(D) Sanger sequencing output of RT-PCR across the splice junction of the CircRNA sample
depicted in lane 1 and lane 3 from (B).
(E) Agarose gel demonstrating the effect of cation concentration and reaction time on splicing.
Column purified RNA from IVT was incubated with buffer including indicated magnesium (50
mM Tris-HCl, 50 /100 mM NaCl, 0-20 mM MgCl2, pH=7.4) for 5 or 15 min.

22
 M
[Mg2+]

M
10 M

m
20 M

M
A [Mg2+]

0m

10
m
m

m
0m
40
δ
EBS1

EBS1 δ EBS1 δ
GCAACT TCTCGT
CGGGAT TCGTGT

Scarless-circRNA Scarless-circRNA

B spacers C V1
SP1 V3
SP2
30
293T
IRES Fluc

20

10
Precursor
circRNA

free introns 0
Luminescence signal (X104)

1 2 5 10 20 50 100 200 500

4
HepG2
Circularization

1 00 % 3
efficiency

5 0%
2
0%
V11 V32 V43 V54 V65 1
SP SP SP SP SP

0
D 1 2 5 10 20 50 100 200 500
Luminescence signal (X104)

50 20
HepA 1-6
40
15
30
10
20
10 5

0 0
6hrs 12hrs 24hrs 48 hrs 1 2 5 10 20 50 100 200 500

Hours after transfection RNA dose (ng)

Fig. 2. Optimization of circRNA production

(A) Design of scarless circRNA production. Left, the schematic diagram for modification of intronic
sequences, resulting in the specific recognition of the splice junction of custom exon and the
generation of a scarless circRNA. Right, two examples of scarless circRNAs produced using
the CirCode design. The circularization was confirmed by both gel electrophoresis and
sequencing of the junction region.
(B) Testing different spacer regions between the ORF and the IRES.

23
(C) Transfection and translation of circRNA in different doses. The circRNAs containing two
different spacers were gel purified and transfected into three cell lines cultured in 24-well plate
at different doses, the activity of firefly luciferase were measured 24 hours after transfection.
(D) Time course of circRNA translation. The circRNAs encoding the Rluc gene were transfected
into transfected into 293T cells (500 ng circRNAs were used in each transfection), and the
luciferase activity were measured at 6, 12 and 24 hours after transfection.

24
 B linear mRNA
A
circRNA

Normalized GLuc activity

Normalized GLuc activity

1 oRNA using PIE

1

(Norm. to Day 1)
(Norm. to Day 0)

circular RNA
0.5 linear mRNA 0.5

0 0
0 3 6 9 12 15 1 2 3 4 5 6

RT Shelf time before transfection (Days) Day after transfection

C Linear PIE CirCode D

mRNA circRNA circRNA

Peak 2
GLuc activity (x106 RLU)

8.0
* * Peak 1

6.0 Peak 1: precursor

Absorbance

Peak 2: circular
Peak 3: intron
4.0
Peak 3

2.0

0 m RNA o RNA Ci rc od e

3.0
* *
activity (x107 RLU)
Cumulative GLuc

2.0 Retention time (min)

Input Fraction: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

1.0

0 m RNA o RNA Ci rc od e

PIE CirCode
PIE PIE CirCode
E circRNA circRNA
F circRNA
CirCode
circRNA circRNA circRNA
IFN-B1 induction (x104 fold)

* *
2.0 * 1.5
3.0 * *
RIG-I induction (x00)

*
Cell number (x10 5)

2.0 1.0
1.0
0.5
1.0

0 0
0.0
ed

ed
d

ed

ed
d

d
ed

ed
k

e
oc

i
ifi

ifi

i
ifi

ifi
i

i
ifi

ifi

ri f

ri f

ri f

ri f
ri f

ri f
M

ur

ur

ur

ur
ur

ur

pu

pu

pu

pu
pu

pu

P
P

n
n

U
U

Fig. 3 Stability and immunogenicity of circRNAs

(A) Thermostability of circRNA. Purified unmodified CVB3-Gluc circRNA and Gluc mRNA were
diluted to the final concentration 100 ng/ul by nuclease-free water, and were stored at room
temperature (25 ℃). RNA samples were collected at day 0, 3, 7 and 14, and stored at -80℃
for further analysis. All collected RNA samples were then transfected into 293T cells (100ng
circRNA and mRNA were used per well in 96-well plate), and the luciferase activity in the
media were measured at 24 hours after transfection.

25
(B) Prolonged protein expression of circRNA in cells. Unmodified Gluc mRNA and CVB3-Gluc
circRNAs generated through group II intron or PIE method were transfected into 293T cells
(equimolar quantities of each RNA, equivalent to between 50 and 100 ng dependent on size,
were used per well in 96-well plate). The cell culture media were fully collected and replaced
at 1, 2, 3, 4, 5, and 6 days after transfection, and the luciferase activity in media were
measured.
(C) High protein expression of circRNA containing the CVB3 IRES and Gluc ORF. The luciferase
activity in medium of 293T cells 24 h after transfection with CVB3-GLuc circRNA or unmodified
GLuc mRNA (Top panel). Relative cumulative luciferase activity produced over 6 days by
293T cells transfected with CVB3-GLuc circRNA or unmodified GLuc mRNA (Bottom panel).
(D) HPLC purification of CVB3-Gluc circRNA from spin column purified sample after IVT. The top
panel is the HPLC chromatogram indicating the peak of precursor, circular and intron RNA,
respectively. The bottom panel demonstrates the agarose gel of input and collected fractions.
(E) Cell viability of A549 cells transfected with unpurified or purified CVB3 Gluc circRNAs
generated by group II intron or PIE method. The cell viability was mesure 2 days after
transfection.
(F) The induced RIG-I and IFN-b1 gene expression of A549 cells at 24 h after transfection with
unpurified or purified CVB3 Gluc circRNAs generated by group II intron or PIE method.

26
 h1
h2
h3
h4
circ c_Batc
circ c_Batc
circ c_Batc
atc
B

c_B
A IVT reaction

Glu
Glu
Glu
Glu
(37℃ Overnight)

circ
Scale up
Circularization efficiency

Total RN A amount (mg)

circRNA

Reaction volume

D IM injection IP injection

C 1000
Gluc activity

Circular RNA 100

Ionized lipids

DSPC 10
DMG-PEG2000

Cholesterol 1
MC3 SM-102 ALC-0315
Formulation
100
15
Eff.Diam= 80 nm
2

03
-10

80
Encap. Efficiency= 96%
Frequency (%)

C-

E
S

SM

AL
PB

60

40

20

0
10 100 1000

Particle size (nm)

Dose: 20ug/mouse

Fig. 4 In vivo delivery of circRNA for protein expression in mice.

(A) Scale up the production of circRNAs from 0.25 mg to ~7.5 mg.
(B) Reproducibility of the circRNA production. Four batches of Gluc circRNA purified from HPLC
were analyzed using capillary electrophoresis with Agilent 2100 Bioanalyzer.
(C) Schematic illustration of circRNA-LNP complex and particle size of CircRNAGluc-LNP.
(D) Gluc activity assayed from mice serum 24 hours post-injection of CircRNAGluc-LNP with
different formulation.
(E) Representative IVIS images of BALB/c mice administrated with 20ug CircRNAGluc-LNP with
two formualtion by the intramuscular(i.m.) routes. Relative luminescence plot is shown and
the scale of luminescence is indicated.

27
 A RBD SP RBD (319-537) foldon B 100

Eff.Diam= 87 nm
80

Frequency (%)
RBD-dimer SP RBD (319-537) RBD (319-537) Encap. Efficiency= 94%

60

d)
ifie
(p l)
d)
-d (tota
40

ur
-R -dm ifie
cR -R (p l)
cir NA BD (tota
NA BD ur

m
20
cR -R D

BD
cir NA B
cR A-R

-dm
0
cir RN

NA D
BD
10 100 1000

B
Particle size (nm)
c

cir NA-R
-R
cir

cR
cR
cir

C
75kD

50kD
1st shot 2nd shot (booster)
37kD
25kD Week: 1 2 3 4 5 6

D 10 µg LNP-circRNA

E
RBD+IgD+
RBD+IgD-
80
CD19

60
IgD

Percent of cells

CD27 RBD 40
Empty LNP
20

0
NA

S
LN

PB
cR
pty
cir
P-

Em
LN
CD19

IgD

µg
10

CD27 RBD

Fig. 5 Design circRNAs for Covid-19 vaccine

(A) circRNA RBD and circRNA RBD dimer purified from HPLC were analyzed using capillary
electrophoresis with Agilent 2100 Bioanalyzer (left bottom), the protein expression of cell
culture medium from circRNA-RBD or circRNA-RBD dimer was determined by western blot
at 24hr after transfection.
(B) The particle size and encapsulate efficiency of circRNA RBD-LNP complexes.
(C) Schematic diagram of the circRNA RBD-LNP vaccination process in BALB/c mice and serum
collection schedule for antibody analysis
(D) Represented unvaccinated and vaccinated cohorts are shown for RBD specific B cell
responses.
(E) FACS analysis results showing the percentages of RBD specific B cell.

28
SARS-Cov-2 Spike RBD Protein Binding Inhibition Assay

A Serum 1:20 Serum 1:200 PBS B

log10 (reciprocal of antibody titer)

log10 (reciprocal of antibody titer)

5
Strong protection against RBD
Hochest 8

Alex- RBDwild type (0.5 ug/ml)

binding of WT strain
4
6

Sera were collected

3 2 weeks post-
4
Alexa boost; RBD Alexa 647–serum
mixtures were incubated for 5 min
2
at room temperature 2

1
SARS-Cov-2 Spike RBD Protein Binding Inhibition AssayIgG1
0

Merge IgG2a IgG2b IgG1 IgG2a IgG2b

Weeks 3 (2 weeks Week 5 (2 weeks

after 1st shot) after the 2nd shot)

Confidential data from CirCode for internal use.

Weak protection against RBD
Hochest

Alex- RBD Delta (0.5 ug/ml)

binding of Delta strain
C

Ratio
Sera were collected 2 weeks
Alexa post-boost; RBD Alexa 647–
serum mixtures were incubated
for 5 min at room temperature

SARS-Cov-2
Merge Spike RBD Protein Binding Inhibition Assay
1 1
G G
/Ig c/Ig
2a G
2
G Ig
Ig
Confidential data from CirCode for internal use.
Alex- RBDOmicron (0.5 ug/ml)

Log10 (reciprocal of IC50 titer)

Very weak protection against 5
Percentage of Inhibition (%)

Hochest DRBD binding of Omicron strain Positive control

4

3
Sera were collected 2 weeks CRP01
Alexa post-boost; RBD Alexa 647– 2
serum mixtures were incubated
for 5 min at room temperature
1

Merge Log10(reciprocal dilution of serum)

Fig. 6. Analysis
Confidential of neutralizing
data from CirCode for internal use. antibody generated by circRNA vaccine.
(A) The competitive binding of RBD and immunized mouse serum to 293T cells expressing
hACE2 in vitro. Alex-RBD: Alexa 647 labeled RBD.
(B) Measurement of RBD-wt specific IgG1, IgG2a and IgG2c endpoint titers in mice at 2 weeks
after 1st shot and 2 weeks after 2nd shot.
(C) Measurement of RBD-wt specific IgG2a/IgG1 and IgG2c/IgG1 ratios.
(D) Neutralization assay of lentivirus-based SARS-CoV-2 pseudovirus with the sera from mice
immunized with 10ug circRNA RBD-LNP.

29
 A

IRES GOI
D2 D3 D4 D5 D6 D1
EBS2
Bulged
D1 IBS3 IBS1

δ adenosine

EBS3 EBS1 A A

5’ Exon 3’ Exon
D3 D4 D5 D6
IRES GOI
D1 D2
D1 D2 D3 D4 D5 D6 IBS3 IBS1
IBS2 IBS1 IBS3

D4 A

IRES GOI
D5 D6 D1 D2 D3
IBS3 IBS1
D1 D3 D4

* Linear
Circular
60
circRNA (%)
Percent of

40
20
0
1 2 3

Fig. S1. Analysis

Fig. of circularization
S1. Analysis efficiencyefficiency
of circularization of Group IIofintron
Groupwith different
II intron split
with sites.
different
split sites.diagram for the design of CirCode system with different split sites. The
Top, schematic
Top, schematic diagram for the design of CirCode system with different split sites.
autocatalytic self-splicing group II intron was split into two fragments at the D1, D3, or D4
The autocatalytic self-splicing group II intron was split into two fragments at the
domain,
D1,and
D3,a customized
or D4 domain, exonsand
containing IRES andexons
a customized codingcontaining
region of aIRES
gene and
of interest
coding
region of a gene of interest (GOI) were inserted between the split introns. Bottom,
(GOI) were inserted between the split introns. Bottom, the circRNA is in vitro synthesized and
the with
analyzed circRNA is ingel.
agarose vitro synthesized and analyzed with agarose gel.

30
 40
Luminescence signal (X104)

30

20

10

0
SP1 SP2 SP3 SP4 SP5

Fig. S2. Analysis of protein expression from circRNA with different spacer
Fig. S2. Analysis of protein expression from circRNA with different spacer regions.
egions.
The protein
The protein expression from expression from circRNA
circRNA containing containing
CVB3 IRES CVB3 IRESORF
and Gluc and Gluc
with ORF with different
ifferent spacer regions (from Fig.
spacer regions (from2B)
Fig.at 48athours
2B) after
48 hours transfection.
after transfection.

31
 Analysis: 40 μg circGluc 100✖ scale up Preparation: 4 mg circGluc
in each injection in each injection

Peak 2 Peak 2
Peak 1: precursor Peak 1: precursor
Peak 1
Peak 2: circular Peak 1 Peak 2: circular
Peak 3: intron Peak 3: intron
Peak 3
Peak 3

1234 3

Fraction 1 2 3 4 Fraction 3

Fig. S3 Analytical and scale-up SEC of circRNA from splicing reactions by HPLC.
Fig. S3 Analytical
Left and scale-up
panel indicates SECHPLC
the analytical of circRNA from splicing
chromatogram reactions
of 40μg circGluc withby HPLC.
three
injections (left top) and agarose gel of collected fractions (left bottom). Right panel
Left panel indicates the analytical HPLC chromatogram of 40μg circGluc with three injections
demonstrates HPLC scale-up chromatogram of 4mg circGluc with two injections (right
(left top)
top)and
andagarose
agarose gel ofof collected
collectedfraction
fractions (leftbottom)
(right bottom). Right panel demonstrates HPLC
scale-up chromatogram of 4mg circGluc with two injections (right top) and agarose gel of
collected fraction (right bottom)

32
Reference:
1. Sahin, U., Kariko, K. & Tureci, O. mRNA-based therapeutics--developing a new class of
drugs. Nat Rev Drug Discov 13, 759-80 (2014).
2. Li, X., Yang, L. & Chen, L.L. The Biogenesis, Functions, and Challenges of Circular RNAs.
Mol Cell 71, 428-442 (2018).
3. Wang, Y. & Wang, Z. Efficient backsplicing produces translatable circular mRNAs. RNA 21,
172-9 (2015).
4. Wesselhoeft, R.A. et al. RNA Circularization Diminishes Immunogenicity and Can Extend
Translation Duration In Vivo. Mol Cell 74, 508-520 e4 (2019).
5. Breuer, J. et al. What goes around comes around: artificial circular RNAs bypass cellular
antiviral responses. Mol Ther Nucleic Acids (2022).
6. Petkovic, S. & Muller, S. RNA circularization strategies in vivo and in vitro. Nucleic Acids
Res 43, 2454-65 (2015).
7. Beaudry, D. & Perreault, J.P. An efficient strategy for the synthesis of circular RNA
molecules. Nucleic Acids Res 23, 3064-6 (1995).
8. Puttaraju, M. & Been, M.D. Group I permuted intron-exon (PIE) sequences self-splice to
produce circular exons. Nucleic Acids Res 20, 5357-64 (1992).
9. Wesselhoeft, R.A., Kowalski, P.S. & Anderson, D.G. Engineering circular RNA for potent
and stable translation in eukaryotic cells. Nat Commun 9, 2629 (2018).
10. Ford, E. & Ares, M., Jr. Synthesis of circular RNA in bacteria and yeast using RNA cyclase
ribozymes derived from a group I intron of phage T4. Proc Natl Acad Sci U S A 91, 3117-
21 (1994).
11. Liu, C.X. et al. RNA circles with minimized immunogenicity as potent PKR inhibitors. Mol
Cell 82, 420-434 e6 (2022).
12. Lambowitz, A.M. & Zimmerly, S. Group II introns: mobile ribozymes that invade DNA. Cold
Spring Harb Perspect Biol 3, a003616 (2011).
13. Rybak-Wolf, A. et al. Circular RNAs in the Mammalian Brain Are Highly Abundant,
Conserved, and Dynamically Expressed. Mol Cell 58, 870-85 (2015).
14. Peebles, C.L. et al. A self-splicing RNA excises an intron lariat. Cell 44, 213-23 (1986).
15. Pyle, A.M. Group II Intron Self-Splicing. Annu Rev Biophys 45, 183-205 (2016).
16. McNeil, B.A., Simon, D.M. & Zimmerly, S. Alternative splicing of a group II intron in a
surface layer protein gene in Clostridium tetani. Nucleic Acids Res 42, 1959-69 (2014).
17. Fan, X., Yang, Y., Chen, C.Y. & Wang, Z. Pervasive translation of circular RNAs driven by
short IRES-like elements. BioRxiv doi: https://doi.org/10.1101/473207(2020).
18. Kariko, K., Muramatsu, H., Ludwig, J. & Weissman, D. Generating the optimal mRNA for
therapy: HPLC purification eliminates immune activation and improves translation of
nucleoside-modified, protein-encoding mRNA. Nucleic Acids Res 39, e142 (2011).
19. Chen, Y.G. et al. Sensing Self and Foreign Circular RNAs by Intron Identity. Mol Cell 67,
228-238 e5 (2017).
20. Dai, L. & Gao, G.F. Viral targets for vaccines against COVID-19. Nat Rev Immunol 21, 73-
82 (2021).

33
21. Meier, S., Guthe, S., Kiefhaber, T. & Grzesiek, S. Foldon, the natural trimerization domain
of T4 fibritin, dissociates into a monomeric A-state form containing a stable beta-hairpin:
atomic details of trimer dissociation and local beta-hairpin stability from residual dipolar
couplings. J Mol Biol 344, 1051-69 (2004).
22. Vogel, A.B. et al. BNT162b vaccines protect rhesus macaques from SARS-CoV-2. Nature
592, 283-289 (2021).
23. Dai, L. et al. A Universal Design of Betacoronavirus Vaccines against COVID-19, MERS, and
SARS. Cell 182, 722-733 e11 (2020).
24. Edara, V.V. et al. mRNA-1273 and BNT162b2 mRNA vaccines have reduced neutralizing
activity against the SARS-CoV-2 omicron variant. Cell Rep Med 3, 100529 (2022).
25. Corbett, K.S. et al. SARS-CoV-2 mRNA vaccine design enabled by prototype pathogen
preparedness. Nature 586, 567-571 (2020).
26. Qu, L. et al. Circular RNA vaccines against SARS-CoV-2 and emerging variants. Cell (2022).
27. Yang, Y. & Wang, Z. IRES-mediated cap-independent translation, a path leading to hidden
proteome. J Mol Cell Biol (2019).
28. Jarrell, K.A. Inverse splicing of a group II intron. Proc Natl Acad Sci U S A 90, 8624-7 (1993).
29. Pegu, A. et al. Durability of mRNA-1273 vaccine-induced antibodies against SARS-CoV-2
variants. Science 373, 1372-1377 (2021).
30. Chen, C.K. et al. Structured elements drive extensive circular RNA translation. Mol Cell 81,
4300-4318 e13 (2021).
31. Polack, F.P. et al. Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine. N Engl J
Med 383, 2603-2615 (2020).

34