The safety of four different adjuvants was assessed in lupus-prone New Zealand black/New Zealand white (BW)F1 mice. Four groups of mice were injected intraperitoneally with incomplete Freund's adjuvant (IFA), complete Freund's adjuvant (CFA), squalene (SQU) or aluminium hydroxide (ALU). An additional group received plain phosphate-buffered saline (PBS) (UNT group). Mice were primed at week 9 and boosted every other week up to week 15. Proteinuria became detectable at weeks 17 (IFA group), 24 (CFA group), 28 (SQU and ALU groups) and 32 (UNT group). Different mean values were obtained among the groups from weeks 17 to 21 [week 17: one-way analysis of variance (anova) P = 0·016; weeks 18 and 19: P = 0·048; weeks 20 and 21: P = 0·013] being higher in the IFA group than the others [Tukey's honestly significant difference (HSD) post-test P < 0·05]. No differences in anti-DNA antibody levels were observed among groups. Anti-RNP/Sm antibody developed at week 19 in only one CFA-treated mouse. Mean mouse weight at week 18 was lower in the ALU group than the IFA (Tukey's HSD post-test P = 0·04), CFA (P = 0·01) and SQU (P < 0·0001) groups, while the mean weight in the SQU group was higher than in the IFA (P = 0·009), CFA (P = 0·013) and UNT (P = 0·005) groups. The ALU group weight decreased by almost half between weeks 29 and 31, indicating some toxic effect of ALU in the late post-immunization period. Thus, SQU was the least toxic adjuvant as it did not (i) accelerate proteinuria onset compared to IFA; (ii) induce toxicity compared to ALU or (iii) elicit anti-RNP/Sm autoantibody, as occurred in the CFA group.
Systemic lupus erythematosus (SLE) is the prototype autoimmune disease, and can severely affect various different organs (i.e. the kidney, lung and nervous central system). The disease is still associated with high morbidity and mortality rates, despite the use of a wide range of drugs that can control symptoms, delay progression and/or improve the quality of life, but never bring about a complete cure. Among the most innovative therapeutic approaches, inducing B cell depletion by the passive administration of anti-CD20 monoclonal antibodies (mAb) seems to show promise in counteracting SLE activity, as demonstrated in mice  and humans (albeit in open non-controlled clinical trials) [2-7].
We have previously characterized 11 7-mer rituximab-specific peptides [8, 9], some of which can induce an anti-CD20 response in mice, with biological effects similar to those obtained with the passive administration of anti-CD20 mAb [8, 9]. The properties of these peptides led us to consider them as a potential vaccine in SLE. The mouse strain (New Zealand black × New Zealand white)F1 (BWF1) can be considered a suitable model for testing the efficacy of the CD20 mimotope peptide, as it spontaneously develops a human-like SLE characterized by the presence of an autoimmune glomerulonephritis (indicated by the appearance of proteinuria) and the presence of lupus-specific autoantibody . However, to obtain an efficient response, a peptide (which behaves as a hapten) has to be coupled to a macromolecule (carrier) to achieve an adequate presentation , and incorporated with an appropriate adjuvant.
Adjuvants have greatly enhanced the efficacy of active immunotherapy thanks to their ability to stimulate inflammatory and adaptive immune responses in a non-specific fashion [12-14]. Aluminium hydroxide (ALU) and mineral (or organic) oils are the adjuvants employed most commonly in animals and humans. ALU is used currently in human vaccines against diphtheria, tetanus, pertussis, Haemophilus influenziae B, hepatitis A/B, human papilloma virus and anthrax [15-17]. Oil adjuvants include pristane (2,6,10,14 tetramethylpentadecane), incomplete Freund's adjuvant (IFA) (an emulsion of mineral oil Bayol F and water), complete Freund's adjuvant (CFA) (composed of inactivated and dried mycobacteria added to Bayol F) and MF59, consisting of an emulsion of the organic oil squalene (SQU) and water (∼2% w/v). With the exception of pristane and CFA, used only in animals, IFA has been employed in humans against influenza , is currently being tested in experimental HIV  and as a tumour vaccine , while MF59 is included in seasonal influenza and influenza A (H1N1) vaccine preparations [21, 22]. However, even the last generation of adjuvants (SQU adjuvants), like the early one , specifically elicited anti-nuclear antibody in non-autoimmune prone Balb/c mice  and pathogenic anti-phospholipid antibodies in C57/B6 mice . Indeed, Balb/c mice receiving adjuvants developed either an autoimmune glomerulonephritis  or autoantibody to ribonuclear protein/Smith (RNP/Sm) and to single-stranded DNA (ssDNA) , the specificity profile/pattern of these autoantibodies being dependent upon the type of adjuvant used .
Recently, a clinical syndrome referred to as ‘autoimmune/autoinflammatory syndrome induced by adjuvant’ (ASIA), encompassing four different clinical entities, namely siliconosis [28, 29], the ‘gulf war syndrome’ , macrophagic myofasciitis syndrome  and post-vaccination phenomena [29, 32-34], has been described in humans. The symptoms are overlapping and have been attributed to the probable effect of adjuvants (reviewed in [35-38]). All the above findings raised concerns as to whether adjuvants alone could be harmful in our animal model of SLE, by acting as a sort of autoimmune disease ‘activator’, and so reversing the potential beneficial effect of a CD20-mimotope peptide vaccine.
To identify the least harmful adjuvant, as no definite data are yet available on the effects of adjuvants on human-like SLE-prone mice BWF1, we evaluated the effects of IFA, CFA, ALU and SQU on the disease course in this mouse strain.
Materials and methods
Reagents and antibodies
Keyhole limpet haemocyanin (KLH) was purchased from Calbiochem (San Diego, CA, USA). IFA and CFA were purchased from Thermo Fisher Scientific Inc. (Rockford, IL, USA). SQU adjuvant (AddaVax) and ALU adjuvant (Alhydrogel) were purchased from Invivogen (San Diego, CA, USA). Double-stranded calf thymus DNA was purchased from Sigma (St Louis, MO, USA). RNP/Sm antigen was purchased from Arotec Diagnostics Limited (Wellington, New Zealand). Phycoerythrin-cyanin 5·1 (PE-Cy5)-conjugated anti-mouse CD3 antibody, phycoerythrin (PE)-conjugated anti-mouse CD8 antibody, and fluorescein isothiocyanate (FITC)-conjugated anti-mouse CD4 and CD19 antibodies were purchased from BD Pharmingen (Franklin Lakes, NJ, USA).
Mice and immunization strategy
Nine-week-old BWF1 female mice were purchased from Harlan Laboratories (San Pietro al Natisone, Italy). Mice were kept at the University of Bari with a 12-h light/dark cycle and experiments were performed in accordance with institutional and internationally approved guidelines on animal research. Animal study protocols were approved by the University of Bari Medical School ethical review Committee.
BWF1 mice were treated with adjuvants that are commercially available for human use (Table 1), namely IFA, SQU and ALU, which were mixed with KLH before injection. Immunogen (200 μl of adjuvant emulsion/injection) was prepared by mixing 100 μl of phosphate-buffered saline (PBS) containing 100 μg KLH with an equal volume of adjuvant. Four groups of mice (four mice per group) were primed at the 9th week (when no proteinuria was detected in any of the mice) with IFA, CFA, SQU or ALU adjuvant. Boosters were given fortnightly (up to the 15th week) until proteinuria ≥100 mg/dl was detected in at least one mouse in each group. Two additional groups of mice treated with PBS-only (adjuvant untreated group; UNT) and with CFA, served as the negative and positive  control groups. Groups with only one mouse left were excluded from further evaluations.
Table 1. Adjuvants most commonly employed in vaccine for human use
Blood was collected every 2 weeks by an orbital technique from the ophthalmic venous plexus. An aliquot of blood was mixed rapidly with 900 μl of PBS containing 1·8 mg K3-ethylenediamine tetraacetic acid (EDTA) and another aliquot was used for serum recovery. Sera were stored at −80°C until used, while EDTA blood was used immediately for white blood cell (WBC) counts and cytofluorimetric assays. Mice were monitored for proteinuria, serum levels of anti-ssDNA and anti-RNP/Sm antibody and weight.
Urine samples were collected weekly and proteinuria was assessed with the urine strip test (Dyaset s.r.l, Portomaggiore, Italy). Urine was graded colorimetrically on a scale from 0 to 4+, corresponding to the following protein concentrations: 0, negative or trace (15–20 mg/dl), 1+ (30 mg/dl), 2+ (100 mg/dl), 3+ (300 mg/dl) or 4+ (>2000 mg/dl). Mice were considered to show proteinuria if two consecutive urine samples were scored ≥1 (30 mg/dl).
Analysis of serum anti-DNA antibodies and anti-RNP/Sm antibodies
Anti-ssDNA antibodies were detected by enzyme-linked immunosorbent assay (ELISA), as described previously , with minor modifications. Briefly, polyvinylchloride 96-well plates were precoated with 50 μl of poly-l-lysine (Sigma) by a 1-h incubation at 37°C. Then, wells were washed three times with PBS and incubated with ssDNA (5 μg/ml), denaturated previously by boiling calf thymus DNA for 10 min and cooling on ice for 15 min. Following 12 h at 4°C, wells were washed once with PBS containing 0·05% Tween 20 (PBS–T20) and blocked with PBS containing 0·5% bovine serum albumin (BSA) (PBS–BSA). Serum samples (diluted 200 times in PBS–BSA) were added to the wells and incubated for 4 h at 25°C. Wells were washed three times with PBS–T20. Bound immunoglobulin (Ig)G was detected by sequential incubation with horseradish peroxidase (HRP)-conjugated xeno-antibody to the Fc portion of mouse IgG (1 h incubation at 25°C) and o-phenylenediamine (0·5 mg/ml; 100 μl/well); colour development was stopped by adding 100 μl 2 N H2SO4 and the absorbance at 490 nm was read with the Benchmark microplate reader (Bio-Rad Laboratories, Hercules, CA, USA). Background binding was determined from the absorbance generated in wells with blocking solution alone. Specific binding was determined by subtracting the background absorbance from the absorbance in the experimental wells. Results were expressed as units/ml extrapolated from a calibration curve constructed using different dilutions of a positive control serum.
Anti-double-stranded DNA (dsDNA) antibodies were detected by ELISA as described previously [41, 42], with minor modifications. Briefly, calf-thymus DNA was passed through a 0·45 μm nitrocellulose Millex syringe filter (Millipore Corporation, Bedford, MA, USA) to remove any ssDNA fragments. Then, 50 μl of dsDNA solution (50 μg/ml per well) was added to plates precoated with poly-l-lysine, and incubated overnight at 37°C. Afterwards, plates were washed and the assay was continued as described above for anti-ssDNA antibody quantification.
Anti-RNP/Sm antibodies were detected in ELISA, using microtitre plates coated with RNP/Sm antigen after overnight incubation at 4°C of 50 μl/well of PBS solution containing 5 μg/ml of antigen. Wells were then washed once with PBS–T20 and blocked with PBS–BSA. Serum samples (diluted 200 times in PBS–BSA) were added to the wells and incubated for 3 h at 25°C. Wells were washed three times with PBS–T20. Detection of bound IgG and colour reactions was performed as described above for the anti-DNA antibody assay.
Flow cytometry analysis
The white blood cell (WBC) count per μl of EDTA/whole blood was determined with the Sysmex KX-21N™ Automated Hematology Analyzer (Lincolnshire, IL, USA). Peripheral blood mononuclear cells (PBMC) were isolated by Ficoll-Hypaque density gradient centrifugation. One hundred μl EDTA–blood were layered on 300 μl of Ficoll in an Eppendorf tube. Following a 30-min centrifugation (500 g) at 20°C, the lymphocyte layer was recovered and washed twice in PBS. Cells (2 × 104) were incubated for 30 min at 4°C with anti-mouse CD3 (PE-Cy5), CD8 (PE), CD4 (FITC) and CD19 (FITC) antibodies, blocked with rabbit IgG (50 μg/ml) and washed with PBS–BSA, fixed in 2% formaldehyde and analysed by flow cytometry (Cyan; Beckman Coulter, Brea, CA, USA). Lymphocyte subset counts were calculated by multiplying the total lymphocyte count by the percentage of each cell type. Results are expressed as the mean number of cells (1 × 103/μl).
All statistical analyses were performed using spss software for windows version 20. Differences between groups in terms of the arithmetic mean of proteinuria, mouse weight and anti-DNA titre were analysed by one-way analysis of variance (anova). In cases of statistically significant differences, set at P < 0·05, the Tukey's honestly significant difference (HSD) post-test was performed for comparisons.
BWF1 mice spontaneously develop an autoimmune disease resembling human SLE. In these mice, anti-DNA antibodies appear in their sera between the 16th and 24th weeks, and proteinuria between the 28th and 32nd weeks, affecting 80% of mice at the 35th week .
Two accidental (not disease-related) deaths were recorded: one in the IFA group at week 17 and one in the ALU group at week 19.
Mice were primed at the 9th week and immunization was discontinued from the 17th week, when a proteinuria score ≥2 was detected in one mouse (IFA group) (Fig. 1). In the subsequent week, proteinuria appeared significantly earlier in the IFA group than in the others (Fig. 1). Indeed, in the IFA group, a proteinuria mean score of 1 was observed as early as at week 17 and increased to 2 and then 2·5 at weeks 27 and 28. In the CFA group, proteinuria became detectable at week 24 (mean score 0·5) and reached 1, and 1·5 at weeks 27 and 28. Proteinuria showed a similar trend in the SQU and ALU groups, becoming detectable at week 28 (SQU: 0·5; ALU: 1) and increasing from week 29 (SQU: 1·7; ALU: 2·3) to week 31 (SQU: 2; ALU: 2·5) (representative results are shown in Fig. 1). In the UNT group, proteinuria became detectable at week 32 (mean score: 0·5), then increased to 1·5 and 3 by weeks 34 and 35. Overall, the mean proteinuria value was statistically different among the groups from weeks 17 to 21 (week 17: one-way anovaP = 0·016; weeks 18 and 19: P = 0·05; weeks 20 and 21: P = 0·013). The Tukey's HSD post-test indicated that proteinuria was statistically higher in the IFA group than in the other groups at weeks 17 (P = 0·026), 18 and 19 (P = 0·05) and at weeks 20 and 21 (P = 0·021), while no differences were observed among the other groups (P > 0·05). After the 21st week, no significant differences in proteinuria were observed among groups up to week 35. Overall, the results indicate that IFA not only accelerated the onset, but also the progression of proteinuria, in the 5 weeks following the last injection of adjuvant compared to all the other groups, while CFA, SQU and ALU affected the disease course to a lesser extent (or not at all) compared to the UNT group.
To define whether the type of adjuvant had any influence on the development of anti-ssDNA, anti-dsDNA and anti-RNP/Sm antibody, these antibodies were evaluated from weeks 9 to 32. Figure 2 shows that no significant differences were detectable in the titres of anti-ssDNA (Fig. 2a) or anti-dsDNA antibodies (Fig. 2b) in the adjuvant-treated groups compared to the UNT mice (one-way anovaP > 0·05), while anti-RNP/Sm antibodies were not detected in any of the mice in the SQU and ALU groups (Fig. 2c). One mouse only, in the CFA group, developed these antibodies early, at the 19th week, and the titre reached a plateau at the 29th week. Anti-RNP/Sm antibody levels became barely detectable in one mouse in the UNT group at week 32.
Based on previous studies indicating that adjuvant immunization may have an influence on mouse weight changes , this parameter was monitored before immunization and weekly, after week 18, when a visible difference in mice weight became evident (Fig. 3). The increased weight observed up to week 18 was most probably attributable to physiological growth, as similar changes were observed in the UNT mice. However, a statistically significant difference was recorded among groups at weeks 18 (one-way anovaP < 0·001), 19 (P = 0·01), 20 (P = 0·02) and 22 (P = 0·01) and at weeks 29 (P = 0·037), 30 (P = 0·008) and 31 (P = 0·035). Specifically, at the 18th week, the mean weight in the ALU group was lower than in the IFA (Tukey's HSD post-test P = 0·04), CFA (P = 0·01) and SQU (P < 0·0001) groups, while the mean weight in the SQU group was higher than in the IFA (P = 0·009), CFA (P = 0·013) and UNT (P = 0·005) groups. No significant differences in the mean weight were found between mice treated with CFA and IFA (P = 0·98) and the control group (P = 0·68), or between the latter and ALU-treated mice (P = 0·28). The ALU group mouse weight remained statistically lower than that of the CFA and SQU groups up to week 22. In subsequent weeks up to week 28, no statistically significant differences were observed in mouse weight among the groups. After the 28th week, the mean weight of the ALU-treated mice decreased from 39 to 28 g (at week 31) (Fig. 3). This effect was probably due specifically to ALU, as no similar marked changes were observed in UNT mice or mice treated with an adjuvant other than ALU.
Cytofluorimetric analysis of peripheral blood lymphocytes from adjuvant-treated mice (Fig. 4, closed symbols) showed no significant changes/differences in the number of CD4-, CD8- and CD19-positive cells compared to UNT mice (Fig. 4, open symbols) for all the monitoring period, indicating no specific effect of adjuvant in the changing lymphocyte subpopulation distribution.
This explorative study evaluated the effect of different adjuvants for human use in BWF1, and showed that IFA, more than CFA, SQU and ALU, markedly accelerated the onset of proteinuria (and hence of glomerulonephritis). It is of interest that the differences in proteinuria between the groups were not associated with the anti-ssDNA and anti-dsDNA antibody titres, which were not statistically different. These findings are in agreement with previous observations by Gilkeson et al. , who showed that immunization of BWF1 mice with Escherichia coli DNA (in association to methylated BSA and adjuvant) attenuated murine SLE, although these mice developed pathogenic anti-DNA antibodies earlier, and at higher titres, than control mice (immunized with BSA and adjuvant).
Regarding anti-RNP/Sm antibody levels, a previous study was unable to demonstrate the presence of these lupus-specific antibodies in BWF sera . Even so, the lack of detection of anti-RNP/Sm antibodies in almost all adjuvant-treated mice was still an unexpected finding, considering (i) the documented ability of mineral and organic oil adjuvants to elicit anti-RNP/Sm antibodies in non-autoimmune Balb/c mice  and (ii) the previous detection of Sm-derived peptide-specific CD4+ T cells in BWF mice . The lack of an anti-RNP/Sm response indicates that none of the adjuvants (but CFA in one mouse) were able to overcome the genetic resistance to the development of these lupus-specific autoantibodies even in this lupus-prone mouse model. Overall, these data suggest that the accelerated proteinuria observed in the IFA mice was unrelated to the titre of either autoantibodies to DNA or RNP/Sm.
A recent study by Bassi et al.  reported a fairly similar time lapse between the appearance of proteinuria in CFA-treated mice and in the UNT mice group (about 8 weeks) to the time observed in our study. Even so, they found that BWF1 mice treated with CFA developed anti-dsDNA antibodies together with proteinuria earlier, and at higher titres, than BWF1 mice treated with PBS . One possible explanation for these different results may be the different immunization routes employed, as we used an intraperitoneal route, while Bassi et al. injected adjuvant subcutaneously in the posterior feet . Whether or not a subcutaneous posterior foot injection may have caused tissue stress and the release of higher amounts of DNA from dying host cells, as observed with aluminium , triggering hypersensitivity reactions, toxicity and/or higher stimulation than occurs with intraperitoneal route immunization, remains to be evaluated.
The notably earlier onset and higher amount of proteinuria observed in the IFA compared to the CFA group may be explained by considering the two different cytokine profiles probably induced by the two adjuvants: CFA has been reported to induce a significantly higher production of interleukin (IL)-17, IL-6 and transforming growth factor (TGF)-β, while IFA appears to enhance IL-10 production [48, 49]; this cytokine is known to be involved in autoantibody production and to contribute to the pathogenesis of SLE [50, 51].
A weight increase post-immunization has been reported previously by Oscherwitz et al. in a non-autoimmune C57BL/6 mouse strain . The observed increase was attributed to adjuvant-induced oedema and/or inflammation, which may promote water retention . Indeed, these investigators demonstrated that IFA and CFA were equally effective, and more so than ALU, in generating a mouse weight increase. In fact, while no differences were detected between the IFA- and CFA-treated groups, these mice showed a significantly higher weight than the mice in the ALU group [44, 52]. Thus, in agreement with these previous findings, IFA- and CFA-treated mice displayed no weight differences, and weighed more than the mice in the ALU-treated group. At variance with the observations by Oscherwitz et al., we found no statistically significant differences between the weight of mice in the IFA or CFA groups compared to the negative control group. Whether the discrepancy between Oscherwitz et al.'s and our results reflects the different strains of mice involved in the study, and hence a different susceptibility to immunization and/or bleeding stress, remains to be determined. A salient aspect of this analysis was the divergence between the weight differences observed in the early post-immunization period (from the 18th to the 22nd weeks), and the lack of differences in the middle post-immunization period (from the 23rd to the 29th weeks). By contrast, in the late post-immunization period (from the 29th to the 31st weeks), an impressive and progressive loss of weight by almost one-half was observed in two of three ALU-treated mice, along with the appearance of a cachexia-like syndrome. An ALU-induced neurotoxic effect  was excluded as a possible cause of this syndrome, as the mice continued to be mobile. It was also unlikely that ALU-induced peritoneal adhesions might have affected the gastrointestinal tract, inducing a malabsorption syndrome, as post-mortem observation of the peritoneal cavity showed fewer adhesions in the ALU compared to the IFA- and CFA-treated mice (data not shown). Finally, nephropathy was ruled out as a possible cause of this syndrome because it was not supported by proteinuria, which showed a similar progress to that observed in UNT mice.
Caution must be exercised in extrapolating the observed effect of adjuvants to the clinical setting in humans, considering: (i) that the ‘adjuvant quantity/host's weight’ ratio in this study was markedly higher than that in human vaccination; and (ii) the different route of immunization employed, in that adjuvant was injected intraperitoneally in this study while the intramuscular route is used commonly for vaccination in humans.
In conclusion, even if the power of the study is limited by the relatively low number of mice in each group, our data indicate that SQU seems to be a less toxic adjuvant compared to ALU, did not accelerate proteinuria, as observed in the IFA-treated group, and did not elicit anti-RNP/Sm autoantibody, as observed in the CFA group.
F. P., Y. S. and E. F. conceived and designed the study; E. F. and F. P. drafted the manuscript. Y. S. and F. P. participated in critical revision of the manuscript for important intellectual content. E. F., E. I. F. and L. D. carried out the immunoassays and acquired the data. E. I. F., L. D. and V. R. participated in the analysis and interpretation of data. E. I. F. and V. R. performed statistical analyses. All authors read, revised and approved the final version of the manuscript.
This work was supported by a grant in 2012 from the ‘SLE Italian Group’. The authors are grateful to Mr Vito Iacovizzi for his excellent secretarial assistance.
Y. S. declares that he has appeared in court as an expert witness in cases of subjects who developed autoimmune phenomena following vaccinations or silicone implants. All the remaining authors declare no conflicts of interest.