The sequences of D3ED3 and D4ED3 were retrieved from UniProt (37, 38). The genes encoding ED3s were artificially synthesized and cloned into a pET15b vector (Novagen) at the endonuclease NdeI and BamH1 sites, as previously described (23, 39). The nucleotide sequences encoding the SCP-tags were added at the C-terminus of ED3 by QuikChange (Stratagene, USA) site-directed mutagenesis, and all sequences were confirmed by DNA sequencing. SCP-tagged ED3 variants were named according to the number and type of amino acids added to the C-terminus of D3ED3 and D4ED3 (20) ( Figure 1 ). For example, D3C5D stands for a D3ED3 with five aspartic acids added after two glycines acting as a spacer between the tag and C-terminus of D3ED3.

The ED3 variants, namely, D3ED3, D4ED3, D3C5D, D3C5K, D4C5D, and D4C5K, were overexpressed in E. coli JM109 (DE3) pLysS as inclusion bodies. Protein expression was induced by the addition of 1.0 mM IPTG at OD590=0.5-0.6. The proteins were purified using Ni-NTA (Wako, Japan) in 6M guanidine-HCl (20, 43), refolded through dialysis, and the proteins in the soluble fraction were collected. The His-tag was cleaved by thrombin proteolysis, and the proteins were purified by a second passage through the Ni-NTA column, followed by reverse-phase HPLC (43). Protein identities were confirmed by analytical HPLC and MALDI-TOF mass spectrometry (Bruker Daltonics, autoflex III), and the lyophilized powders were stored at -30°C.

Samples for all biophysical and immunization experiments were prepared as follows: The lyophilized protein powders were first dissolved in MilliQ water (Millipore A10 ultra-pure water purifier, EMD Millipore, Germany) at 1mg/mL, and the samples were centrifuged at 20,000xg for 20 minutes at 4°C to remove any aggregates that might have accumulated during sample preparation. The supernatants containing soluble proteins were aliquoted into 1mg/mL stock solutions.

We prepared two homo-serotypic combinations, (D3C5D+D3C5K and D4C5D+D4C5K), and two hetero-serotypic combinations (D3C5D+D4C5K and D3C5K+D4C5D). In order to generate the aggregates, C5D and C5K tagged proteins were mixed at a 1:1 ratio to yield a final protein concentration of 0.3 mg/mL in phosphate-buffered saline (PBS, pH 7.4), phosphate buffer (PB, pH 7.0) and Tris-HCl (pH 8.8). For example- D3C5D and D4C5K proteins were mixed at 0.15 mg/mL concentration yielding a total ED3 concentration of 0.3 mg/mL (0.15 + 0.15 mg/mL). The sample was named D3C5D+D4C5K. As a control, the untagged D3ED3 and D4ED3 proteins were mixed in the same way, and the formulation was termed D3+D4. The samples were incubated for 20 minutes at 25°C to generate aggregates. Monomeric ED3s were prepared in the same way. In brief, lyophilized protein powders were reconstituted at 0.3 mg/mL concentrations in PBS or PB and incubated for 20 minutes at 25°C. The final concentrations of ED3 molecules were identical (0.3 mg/mL) regardless of the fraction of homo-, hetero-, and monomeric proteins in the sample. In all cases, the final protein concentration was confirmed by measuring the extinction coefficient (A₂₈₀) using Nanodrop-2000 (Thermo Fisher Scientific, USA).

The hydrodynamic radii (R _h) of the individual SCP-tagged ED3 variants and D3C5D+D4C5K aggregates were measured at 25° and 37°C by dynamic light scattering (DLS) (44) on a Zetasizer Nano-S (Malvern, UK). D3C5K+D4C5D formed visible aggregates (precipitates) and no biophysical measurements could be performed. For the other variants, three independent readings were recorded for calculating the average R _h from the number distributions using the Stokes-Einstein equation (45). The presence of sub-visible aggregates was also monitored by static light scattering (SLS) (46) at a wavelength of 600 nm using an FP-8500 spectrofluorometer (JASCO, Japan) and using a quartz cuvette with a 3 mm optical path length. Each SLS measurement was repeated three times, and the values were averaged.

The secondary structure contents were characterized by far-UV circular dichroism (CD) spectroscopy using a JASCO J820 CD spectropolarimeter (JASCO, Japan), and spectra were measured using a 2 mm optical path length quartz cuvette (20, 47). Trp-fluorescence spectra with λ_ex 295 nm were measured on a JASCO FP-8500 spectrofluorometer using a quartz cuvette with a 3 mm optical path length in order to examine the conformational stability of the ED3s (20, 23). Both measurements were conducted at a final protein concentration of 0.3 mg/mL at 25° and 37°C.

Four-week-old female mice (Jcl : ICR, CLEA, Japan) were used for the immunization experiments ( Supplementary Table S2 ). Control mice were injected with PBS only. The D3ED3, D4ED3, and combinations of ED3s were formulated (see sample preparation section) and administrated at a final concentration of 30 µg/dose (100 µL/mice) as previously described (20). No adjuvant was used because the IgG antibody level was sufficiently increased by the sub-visible aggregates without a need for external adjuvant (20, 27). Besides, the aim of our study was to examine the effects of sub-visible aggregates on immunogenicity, per see, without external factors.

The samples were injected subcutaneously five times at weekly intervals. Note that, in addition to the aforementioned DLS measurements, we monitored the R _hs just before injection by taking an aliquot of the injecting sample. Dose-dependent (doses 1-5) anti-D3ED3 and anti-D4ED3 antibody titers (IgGs) were measured three days after each inoculation using tail-bleed sera by enzyme-linked immunosorbent assay (ELISA).

After the fifth dose, the mice were kept untreated for seven weeks to assess their long-term IgG titers. During this period, the serum IgG levels were monitored weekly using the tail-bleed sera. On the seventh week, a sixth dose was injected, all mice were sacrificed after a week, and blood samples were collected from the heart, centrifuged, and the sera were preserved at -30°C. All of the animal experiments were performed in compliance with the Tokyo University of Agriculture and Technology (TUAT) review panel for animal experimentation and Japanese governmental regulations.

ELISA was carried out as previously described (20, 33). In short, anti-ED3 IgG levels were evaluated using the untagged D3ED3 and D4ED3 (2.5 µg/mL in PBS) as coating antigens on a 96-well microtiter plate (TPP microtiter plates, Japan). Anti-ED3 sera were applied to the PBS-washed wells at an initial dilution of 1:50 for the tail-bleed samples. Plates were then incubated at 37°C for 2 hours. In all ELISA plates, previously developed anti-dengue-ED3 (20) and anti-BPTI-19A sera (33) were used as positive and negative controls, respectively.

After washing the plates thrice with PBS-0.05% Tween-20, each well received a 100 μL of anti-mouse IgG HRP conjugate (Thermo Fisher Scientific, USA) at 1:3000 dilution in 0.1% BSA-PBS-Tween-20 and incubated at 37°C for 1.5 hours. As a substrate, O-phenyl Di-amine (OPD) was added. The color intensities were measured at 492 nm using a microplate reader (SH-9000 Lab, Hitachi High-Tech Science, Japan) immediately after stopping the reaction with 1 N sulfuric acid (50 µL/well). Antibody titers were calculated from the power fitting of OD_492nm versus the reciprocal of the antisera dilution using a cutoff of OD_492nm = 0.1 above the background values. The values were averaged over the number of mice (n) in the respective groups.

Experimental samples were prepared according to our reported protocol (20). In short, single-cell suspension of mice splenocyte was prepared in FACS (Fluorescence Activated cell Sorter) buffer [PBS supplemented with 2% FBS (Fetal Bovine Serum), 1 mM EDTA, and 0.1% sodium azide]. Afterward, 1 X RBC lysis buffer (0.15 M ammonium chloride, 10 mM potassium bicarbonate, 0.1 mM EDTA) was used to lysis the red blood cells (RBCs). Furthermore, 1 million splenocyte cells in 100 µL pre-cooled FACS buffer were surface stained with different fluorescent labeled antibodies according to the manufacture’s guidelines. To study the CD4 T-lymphocytes, cells were stained with anti-CD3-Pcy5, CD4-Pcy7, CD44-FITC, and CD62L-PE-conjugated antibodies in one tube, and for CD8 T- lymphocytes, cells were stained with anti-CD3-Pcy5, CD8-Pcy7, CD44-FITC, and CD62L-PE-conjugated antibodies in another tube (0.2 μg of antibodies/100 μL) for 30 minutes in the dark. Unbound excess conjugated antibodies were removed by centrifugation, and the cells were resuspended in 500 μL of FACS buffer. The data were collected using a CytoFlex flow-cytometer (Beckman Coulter, USA).

The influence of C5D and C5K tags on the formation of sub-visible aggregates were examined by dynamic light scattering (DLS) and static light scattering (SLS) at 25° and 37°C. The hydrodynamic radii of the untagged D3ED3, D4ED3, and tagged ED3s ranged from 1.6-2.4 nm ( Figures 2A, C , and Table 1 ), strongly suggesting that the untagged ED3s, as well as the un-mixed tagged ED3s, did not aggregate ( Supplementary Figures S1A, B and Table 1 ).

On the other hand, mixing ED3 tagged with residues of opposite charges produced, as we expected, sub-visible aggregates of about 500 nm, which were stable over time ( Supplementary Figures S1C, D ). In particular, D3C5D and D4C5K mixed at an equimolar concentration (D3C5D+D4C5K) generated submicron aggregates with R _h of ~522 nm at 37°C ( Table 1 and Figures 2A, C ) and showed an increased SLS signal ( Figure 2B and Supplementary Figure S1B ). To date, no aggregates were detected in sodium-acetate buffer (pH 4.7) ( Figure 2E ), where the aspartic acid side chains lose their negative charges, and the electrostatic interactions disappear. In addition, the aggregate’s size decreased at high salt concentrations, further corroborating our hypothesis that the aggregates originated from the electrostatic attraction between the C5D and C5K tags ( Figure 2D ).

We assessed the impacts of the SCP-tags on the conformational stability of ED3 tagged variants and their aggregates using Trp-fluorescence and circular dichroism (CD) at 25° and 37°C. The SCP-tagged ED3 variants maintained their native-like conformations as assessed by Trp-fluorescence ( Figures 3A, B ) and CD ( Figures 3C, D ) at 25° and 37°C. Furthermore, ED3s attached to oppositely charged SCP-tags oligomerized in a native-like conformation ( Figure 4 ) when mixed. Namely, Trp-fluorescence (48) ( Figures 4A, B ) and CD spectra ( Figures 4C, D ) of the mixed sample (D3C5D+D4C5K) showed spectra that were the averages of D3ED3 and D4ED3. The monovalent samples (D3C5D+D3C5K and D4C5D+D4C5K) exhibited spectra identical to the respective untagged D3ED3 and D4ED3 spectra ( Figure 3 ). Altogether, C5D and C5K tags could form both homo and hetero-oligomeric D3ED3 and D4ED3 in essentially a native-like conformation, unlike the previously reported C4I that oligomerized ED3 in a partially unfolded state due to the reverse hydrophobic effect (20, 49).

The effects of C5D and C5K favored sub-visible aggregates of ED3s on the immunogenicity of templates D3ED3 and D4ED3 were examined in mice models. We injected five doses at one week interval as described in our previous study (20). Two doses did not induce an immune response. The IgG levels started to increase after dose-3 ( Supplementary Figures S2A–D ). After dose-5, the titer differences among mice were clearly observable. The untagged-D3ED3 was moderately immunogenic, while the D4ED3 was barely immunogenic ( Figure 5 and Supplementary Table S1 ) as assessed by anti-ED3 IgG antibody detected using ELISA. The immunogenicity (namely, anti-D3ED3 IgG antibody titers) of the monovalent combination of D3C5D+D3C5K was ~5 times higher than that of the untagged-D3ED3 ( Figure 5A , Supplementary Figure S2A , and Table S1 ); and the immunogenicity of D4C5D+D4C5K was ~15 times higher than that of the untagged D4ED3 ( Figure 5B , Supplementary Figure S2B and Table S1 ). On the other hand, the immune responses against the individual and thus monomeric D3C5D, D3C5K, D4C5D and D4C5K were close to that against the untagged ED3s ( Figures 5A, B and Supplementary Table S1 ).

Noteworthy, the anti-ED3 IgG antibodies generated by the heterotypic D3C5D+D4C5K produced anti-sera with almost identical anti-D3ED3 and anti-D4ED3 titers in 3 of 4 responsive mice ( Figure 5 ; Supplementary Figure S2 and Tables S1, S2 ), and the titer was >100 fold higher than those generated by the untagged D3+D4 formulation ( Figures 5C, D and Supplementary Table S1 ). In addition, the antibody responses of D3ED3 and D4ED3 are sero-specific with minimal or no sero-cross-recognition (20, 43). Namely, anti-D3ED3 recognizes only D3ED3 and anti-D4ED3 recognizes only D4ED3. Therefore, we concluded that the bivalent anti-D3ED3 and anti-D4ED3 raised by D3C5D+D4C5K injection originated from the bundling of the two proteins.

The anti-IgG antibody titers remained high for several weeks ( Figures 5E, F ), and a booster dose of D3C5D+D4C5K generated a robust immune response against both D3ED3 ( Figure 5E ) and D4ED3 ( Figure 5F and Supplementary Table S1 ). We then assessed the generation of immunological memory in mice immunized against D3C5D+D4C5K by analyzing the cell surface CD markers [CD4 (T-helper) and CD8 (T-cytolytic); n=3] 15 weeks after the first immunization). We used a randomly selected mouse for the flow cytometry experiments. The mouse injected with untagged ED3s (D3+D4) produced a high percentage of CD44-CD62L+ co-expressed CD4 (85%) and CD8 (93%) cells, indicating their naïve immunological status. On the other hand, immunization with heterotypic SCP-tagged ED3s (D3C5D+D4C5K) showed a low number of CD44-CD62L+ (on both CD4+ and CD8+ cell) when compared with those observed following D3+D4 co-immunization ( Figure 6 ). These observations indicate that a central and effector T-cell memory was established through D3C5D+D4C5K immunization.