Foresight | doi: 10.1111/j.1365-2796.2010.02224.
Inverse vaccination, the opposite of Jenner’s concept, for
therapy of autoimmunity
L. Steinman
From the Department of Neurology and Neurological Science, Interdepartmental Program in Immunology, Stanford University, Stanford, CA, USA
Abstract. Steinman L (Stanford University, Stanford,
CA, USA). Inverse vaccination, the opposite of
Jenner’s concept, for therapy of autoimmunity (fore-
sight). J Intern Med 2010: 267: 441–451.
DNA-based vaccines to induce antigen-specific
inhibition of immune responses in human autoim-
mune diseases represent the inverse of what Jenner
intended when he invented vaccination. Jenner’s
vaccine induced antigen-specific immunity to small
pox. DNA vaccines for autoimmunity have been
developed in preclinical settings, and now tested in
human trials. The first two clinical trials, one in
relapsing remitting multiple sclerosis, and the other
Introduction
What is meant by the inverse of Jennerian vaccination?
In 1796, Edward Jenner wrote,
I selected a healthy boy, about eight years old, for the
purpose of inoculation for the cow-pox. The matter
was taken from a sore on the hand of a dairymaid,
who was infected by her master’s cows, and it was in-
serted, on the 14th of May, 1796, into the arm of the
boy by means of two superficial incisions, barely pe-
netrating the cutis, each about half an inch long [1].
He continued,
In order to ascertain whether the boy, after feeling so
slight an affection of the system from the cow-pox
virus, was secure from the contagion of the smallpox,
he was inoculated the 1st of July following with vario-
lous matter, immediately taken from a pustule. Sev-
eral slight punctures and incisions were made on
both his arms, and the matter was carefully inserted,
but no disease followed. The same appearances
were observable on the arms as we commonly see
when a patient has had variolous matter applied,
in type 1 diabetes indicate that specific inhibition of
antigen-specific antibody and T-cell responses is
attainable in humans. Further development of this
approach is ongoing. This new version of immuni-
zation termed ‘inverse vaccination’ when applied to
autoimmune diseases, may allow targeted reduc-
tion of unwanted antibody and T-cell responses to
autoantigens, while leaving the remainder of the
immune system intact. The method of specifically
reducing a pathological adaptive autoimmune re-
sponse is termed inverse vaccination.
Keywords: autoimmunity, diabetes mellitus, Jenner,
multiple sclerosis, vaccination.
after having either the cow-pox or smallpox. Several
months afterwards he was again inoculated with
variolous matter, but no sensible effect was produced
on the constitution.
Jenner’s powerful assertion was that ‘the cow-pox
protects the human constitution from the infection of
the smallpox’ [1]. These classic experiments per-
formed in the eighteenth century established the
remarkable idea that the immune system could be
turned on, in a highly specific manner that afforded
protection from a highly virulent and often fatal viral
infection. In the past three centuries, quite surpris-
ingly there has been very little progress, at least in
man, in establishing what would be the very opposite
of Jenner’s vaccination. In experimental animals
there are scores of papers demonstrating that it is
possible to induce control or eliminate specific im-
mune responses, although reducing ongoing autoim-
mune responses has been more challenging [2]. The
opposite of what Jenner intended would involve
reducing or eliminating specifically a particular
undesired immune response. Such an approach
would with high likelihood provide enormous benefit
to an individual suffering from an autoimmune
ª 2010 Blackwell Publishing Ltd 441
 L. Steinman
| Inverse vaccination for autoimmunity therapy

disease stemming from an immune response to that
component of what we call ‘self’, in the immunological
sense.
Inverse Jennerian vaccination in the context of current approaches to au-
toimmunity
Current approaches to treatment of autoimmunity
were for the most part taken from drugs that have al-
ready been approved to treat lymphoid malignancy
like the anti-CD20 monoclonal Rituxan [3–5] or the
anti-CD52 monoclonal antibody [6, 7], Campath, or
to reduce transplant rejection, with various monoclo-
nal anti-CD3 antibodies [8, 9]. Thus, some of the most
popular approaches to treating autoimmunity in-
clude extinguishing nearly all of one’s CD20 B cells
[3–5], paralysing all of one’s CD3 T cells [8–10], killing
all of one’s CD52 white blood cells [6, 7], impairing in-
gress of all T and B cells to affected organs via integrin
blockade [11–14] or impeding egress of lymphoctyes
out of lymph nodes with modulators of spingosine
phosphate receptors [15]. So it is instructive to
emphasize Jenner’s observation:
slides [16–19]. We customized the content of these ar-
rays. The first array studied was comprised of many
of the known target autoantigens identified for sys-
temic lupus erythematosus [16]. To study MS, all the
known myelin proteins in their recombinant forms,
their peptide epitopes including post-translationally
modified variants were printed on slides [17]. Using a
two-stage system after ‘printing’ the slides, fluid ta-
ken either from serum or from cerebrospinal fluid
(CSF) was first applied to the surface of these slides. A
second stage with fluorescent antihuman immuno-
globulin was then applied and the slides were
scanned for their fluorescence. See Fig. 1 showing a
‘heatmap’ indicating the major antibody responses in
the spinal fluid of individuals with relapsing remitting
(RR) MS [19].
Quantitative measures of antibodies to various mye-
lin components were obtained and validated with
alternate methods of measurement [16, 17]. We were
able to reveal the major targets of the immune re-
sponse to myelin in various models of MS, collectively
RRMS Controls

what renders the cow-pox virus so extremely singular 1000

is that the person who has been thus affected is for- 500
250
ever after secure from the infection of the small pox; 100
50
neither exposure to the variolous effluvia, nor the Brain mass 25
Numbness
Headache

Headache

Headache
insertion of the matter into the skin, producing this Leukemia 10
Seizure

Seizure
Cancer

Cancer
5
RRMS
RRMS
RRMS
RRMS
RRMS
RRMS
RRMS
RRMS
RRMS
RRMS
RRMS

RRMS

Lyme

AIDS
distemper [1]. <1

The singularity of his approach to modulation of
adaptive immunity is the essence of what my col-
leagues and I are attempting in human autoimmune
disease: We are trying to shut down specific un-
wanted T- and B-cell responses in individuals already
afflicted with diseases thought to be autoimmune
such as multiple sclerosis (MS) and type 1 diabetes.
Instead of undertaking approaches that have been
translated from the fields of transplant rejection and
from cancer therapy, where wide swathes of the im-
mune response are extinguished, we are attempting a
much focused approach for controlling unwanted
adaptive immune responses.
Identification of targets in particular autoimmune diseases for inverse
vaccination
In 1999 we initiated a two-pronged attack to first
identify the targets of adaptive immunity in various
autoimmune diseases, and then to develop a way to
reduce these unwanted adaptive responses. Large-
scale arrays were developed where hundreds of puta-
tive autoantigens were printed on microscopic glass
CRYAB 21–40
J37
PLP 50–69
MBP 40–54
Rat MBP 63–88
CRYAB PROTEIN
CRYAB 116–135
Abeta 1–12
MBP 19–37 CIT X2
HSP 70
PLP 180–199
Fig. 1 Autoantibody responses in the spinal fluid of
patients with relapsing remitting multiple sclerosis. Analy-
sis was performed on cerebrospinal fluid from relapsing
remitting multiple sclerosis (RRMS) and other neurologic dis-
ease (OND) patients. Statistical Analysis of Microarrays
identified significant differences in antibodies in RRMS com-
pared with OND. Samples are arranged with hierarchical
clustering, and displayed as a heat map. RRMS pat-
ients demonstrated significantly increased autoantibodies
against various myelin epitopes including CRYAB protein
and peptides (J37, Golli-myelin basic protein isoform J37);
PLP, proteolipid protein; MBP, myelin basic protein; HSP,
heat shock protein; Abeta, amyloid beta). A false discovery
rate threshold of 1.9% and numerator threshold of 2.0 were
used. Prediction Analysis of Microarrays yielded a classifi-
cation model with 21 markers, with cross-validated sensi-
tivity of 10 ⁄ 12 = 83% and specificity of 11 ⁄ 12 = 92% [19].
With permission from Nature Publishing.

442 ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451
 L. Steinman
| Inverse vaccination for autoimmunity therapy
called experimental autoimmune encephalomyelitis
[17], as well as the major targets of the adaptive im-
mune response in the spinal fluid of patients with
RRMS [18, 19].
Development of inverse vaccines to attenuate specific ongoing adap-
tive autoimmunity
We chose to develop a technology based on immuni-
zation with DNA encoding a specific antigen that had
first been discovered in the early 1990s [2]. In 1990,
Wolff et al. [20] showed that naked DNA could trans-
fect cells in vivo. This breakthrough led to the first
successful vaccine in animal models when Ulmer
et al. [21] injected a naked DNA encoding influenza,
and conferred protection against infection. Since
then there have been numerous attempts to use DNA
vaccines to induce what might be considered conven-
tional Jennerian immunization against a variety of
microbial targets. For the past 20 years, investigators
have been developing DNA vaccines to induce immu-
nity to microbial infections including HIV [22], rabies
[23], ebola [24] and influenza [21]. There are no ap-
proved DNA-based vaccines as yet.
In 1996, we published experiments where a DNA vac-
cine was first used to treat autoimmune disease [25].
The vaccine targeted a key portion of the T-cell recep-
tor that is found on pathogenic T cells that cause
paralysis in experimental autoimmune encephalo-
myelitis (EAE), a collection of autoimmune diseases
that bears some important resemblances with MS. A
variable region gene of the T-cell receptor, V beta 8.2,
is rearranged, and its product is expressed on patho-
genic T cells that induce EAE in H-2u mice after
immunization with myelin basic protein (MBP). Vac-
cination of these mice with naked DNA encoding V
beta 8.2 protected mice from EAE. Analysis of T cells
reacting to the pathogenic portion of the MBP mole-
cule indicated that in the vaccinated mice there was a
reduction in the Th1 cytokines such as interferon-c
(IFN-c). In parallel, there was an elevation in the pro-
duction of interleukin (IL)-4, a Th2 cytokine associ-
ated with suppression of disease. A novel feature of
DNA immunization for autoimmune disease, modifi-
cation of the autoimmune response from Th1 to Th2,
indicated that this approach might be relevant for
treatment of autoimmune diseases like MS, type 1
diabetes and rheumatoid arthritis, if the particular
critical V region genes could be identified [25]. How-
ever, given the diversity of T-cell receptor rearrange-
ments it became clear that the approach to targeting
autoimmunity via ‘critical’ T-cell receptors was going
to be a daunting task. Hence, efforts quickly evolved
towards trying to reduce autoimmune responses to
pathological autoantigens.
Several groups attempted DNA vaccination with plas-
mids encoding various myelin antigens in various
models of EAE. The Karolinska group showed that
EAE was suppressed with a DNA vaccine encoding an
immunodominant portion of MBP for the Lewis rat,
fused to a dimerized synthetic analogue of the immu-
noglobulin G (IgG)-binding B domain of staphylococ-
cal protein A [26].
We demonstrated that a DNA vaccine encoding an
unmodified myelin epitope, PLP139–151 was shown
to suppress EAE, with a reduction in the degree and
frequency of paralysis in mice. Proliferative re-
sponses and production of the quintessential Th1
cytokine, IFN-c, were reduced in T cells responsive to
PLP139-151. In the brains of mice that were success-
fully vaccinated, mRNA for IL-2, IL-15 and IFN-c were
reduced. The mechanism underlying this reduction
in both severity and incidence of paralysis and the
reduction in Th1 cytokines involved diminished
co-stimulation of T cells. DNA immunization with a
plasmid encoding proteolipid protein (PLP) peptide
139–151 altered expression of the co-stimulatory
molecules, CD80, and CD86 on antigen presenting
cells in the spleen [27].
We also explored whether co-administration of plas-
mids encoding Th2 cytokines might enhance DNA
vaccination with plasmids encoding myelin proteins
[28]. Using a combination of local gene delivery and
tolerizing DNA vaccination, we demonstrated that co-
delivery of the gene encoding IL-4 gene and a DNA
plasmid encoding the self-peptide PLP 139–151 pro-
tected against development of EAE. Mechanistic
studies revealed that IL-4 expressed from the naked
DNA is secreted and acts locally on autoreactive T
cells via activation of STAT6 to shift their cytokine
profile from T helper 1 (Th1) to Th2. This approach
with a DNA vaccine plus co-delivery of the gene for IL-
4 reversed established EAE and diminished paralysis
in a second model of EAE induced with myelin oligo-
dendrocyte glycoprotein (MOG) [28].
Some experiments with DNA vaccination in EAE lead
to worsening of disease. CpG motifs in bacterial DNA
were shown to activate the immune system via toll-
like receptors and to directly activate T and B cells
[29]. Raz et al. [30] showed that CpG sequences were
critical for induction of DNA vaccination via induction
of Th1 responses. Fujinami et al. [31] recognized that
oligonucleotide motifs, termed CpG, could induce
ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451 443
 L. Steinman
| Inverse vaccination for autoimmunity therapy
Th1 cytokines. They showed that bacterially derived
DNA plasmids, termed pCMV, could induce IL-6, IL-
12 and IFN-c. They also showed that pCMV injection
also enhanced R-EAE with increased IFN-c and IL-6
responses [31]. In contrast, Lobell et al. [32] at the
Karolinska showed that plasmids without any of
these so-called CpG motifs were incapable of protect-
ing mice from EAE. Thus, treatment with a DNA vac-
cine encoding MBP peptide 68–85 and containing
three CpG motifs with their characteristic 5¢-
AACGTT-3¢ sequence, suppressed paralysis in EAE.
A DNA vaccine without any of these CpG motifs failed
to suppress EAE in two different models of EAE, one
induced with a peptide to MBP and the other with a
peptide to MOG. The issue was therefore raised
whether CpG motifs were beneficial or not in inverse
vaccination for autoimmune disease.
An oligonucleotide motif termed GpG to suppress Th1 T-cell responses
The flagship cytokine of Th1 immunity is IFN-c [33].
Although administration of IFN-c is associated with
protection from paralysis in EAE [33, 34], adminis-
tration of this quintessential Th1 cytokine leads to
worsening of MS [35]. Segal et al. [36] showed that
EAE could be induced with lymphocytes that were
first sensitized to PLP 139–151, and were then acti-
vated with CpG. Our own work in agreement with that
of Lobell et al. [32] showed that at least for EAE, CpG
oligonucleotides in bacterial plasmids could by them-
selves suppress EAE. We demonstrated that injection
of plasmid DNA could suppress the prototypic T-cell-
mediated autoimmune disease, EAE, by inducing
IFN-c [37]. Given the potential danger of inducing a
Th1 response, with high IFN-c production when we
translated this approach to MS and to other diseases
where Th1 responses are undesirable, we decided to
design an oligonucleotide motif that might actually
compete with the CpG motif. We predicted that
replacing CpG motifs with a competitor might be the
best course of action, given that we wanted to avoid
amplifying Th1 immunity in those with autoimmune
disease.
We discovered that a GpG oligonucleotide, with a sin-
gle base switch from CpG to GpG, which effectively
inhibited the activation of Th1 T cells was associated
with autoimmune disease [38]. A CpG oligonucleo-
tide with the sequence 5¢-TGACTGTGAACGTTAGA-
GATGA-3¢ was changed to a GpG oligonucleotide, 5¢-
TGACTGTGAAGGTTAGAGATGA-3¢. The immuno-
modulatory GpG-ODN suppresses the severity of
EAE in mice, a prototypic Th1-mediated animal dis-
ease model for MS. The GpG oligonucleotide stoichio-
444 ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451
metrically competed with CpG for binding to its recep-
tor the toll-like receptor 9, TLR9. The combination of
stimulatory CpG oligonucleotides with inhibitory
GpG oligonucleotides resulted in a marked reduction
in phosphorylation of IkB- at Ser32. Overall, GpG oli-
gonucleotides effectively competed with CpG oligonu-
cleotides to inhibit Th1 responses [38].
Matching the inverse vaccine to targets detected on autoantibody ar-
rays
We used the two-pronged approach to treat autoim-
mune disease using large-scale analysis of the speci-
ficity of autoantibody responses in EAE to guide
selection of the DNA vaccine used to suppress EAE.
To survey autoantibody responses in EAE, we used
our antigen microarrays containing a spectrum of
proteins and peptides derived from the myelin
sheath, the target of the autoimmune response in
EAE and the 2304-feature myelin proteome arrays
contain 232 distinct antigens, including proteins and
sets of overlapping peptides representing MBP, PLP,
MOG, myelin-associated oligodendrocytic basic pro-
tein (MBOP), oligodendrocyte-specific protein (OSP),
aB-crystallin, cyclic nucleotide phosphodiesterase
(CNPase) and Golli-MBP. We used these customized
myelin arrays to profile autoantibody responses in
serum derived from mice with EAE. Images of repre-
sentative arrays are presented in Fig. 2.
We applied our myelin proteome arrays to study the
evolution of the autoreactive B-cell responses in EAE.
Based on results from these arrays showing large-
scale spreading of the immune response to multiple
myelin antigens in EAE, we devised a DNA vaccine
with plasmids encoding four myelin proteins, MBP,
PLP, myelin oligodendroglial glycoprotein and mye-
lin-associated glycoprotein. This vaccine referred to
as a ‘cocktail’ vaccine was given to mice after the first
attack of paralysis. Beginning on day 17 after recov-
ery from the acute paralytic attack in EAE (7–8 days
after disease onset), mice were injected intramuscu-
larly on a weekly basis with control therapies or DNA
encoding this cocktail of array-determined myelin
targets The vaccine with plasmids encoding the four
myelin proteins reduced relapses when given after
the initial attack of paralysis; see Fig. 3. Autoantibody
responses to myelin proteins were reduced on a large
scale; see Fig. 3.
We next tested to see what would happen when we
continued to optimize the DNA vaccination strategy,
employing in ongoing EAE an optimized set of
antigens, the cocktail of four plasmids encoding four
 L. Steinman
| Inverse vaccination for autoimmunity therapy

Normal SCH PLP 139–151 MBP 85–99

> 20 000
10 000
5000
2500
1000

PLP-10 (2)
Normal-10
Normal-11

MBP-8 (3)
MBP-7 (3)
MBP-6 (4)
MBP-5 (2)
MBP-4 (1)
MBP-1 (2)
MBP-3 (1)
MBP-2 (2)
SCH-4 (3)
SCH-2 (5)
SCH-3 (3)
SCH-1 (2)
PLP-9 (2)
PLP-7 (1)
PLP-6 (2)
PLP-8 (1)
PLP-2 (2)
PLP-3 (4)
PLP-4 (2)
PLP-1 (3)
PLP-5 (2)
Normal-3
Normal-1

Normal-2
Normal-8
Normal-4
Normal-9
Normal-7
Normal-6
Normal-5 500
250
100
< 25

PLP 50–69
PLP 139–151
PLP 139–151
hMBP 31–49
hMBP 61–79
hMBP 151–170
hMBP 141–159
MBP 85–99
hMBP 71–89
MOG plate P31
hMBP 1–20
PLP210–229
MOG protein
hMBP 131–153
P2 26–48
hMBP 6–14
hMBP protein
gpMBP protein
PLP 190–209
rMBP 63–88
3.3 1.5 2.5 3.3 1.6 Relapse rate

P = 0.088 P = 0.025

Fig. 2 The diversity of autoantibody responses in acute experimental autoimmune encephalomyelitis (EAE) predicts subsequent
disease activity. Hierarchical clustering of antigen features with statistically significant differences in myelin proteome array
reactivity between sera derived from groups of normal control mice and from groups of mice upon recovery from acute EAE induced
with proteolipid protein 139)151 (day 17), myelin basic protein 85)99 (day 22) or spinal cord homogenate (day 25). Mice were la-
ter scored daily for 10 weeks to determine the number of relapses for each mouse (indicated in parentheses). The average relapse
rates for mice included in the primary subnodes of the dendrogram, and P-values by Mann–Whitney test for the differences in re-
lapse rate between these nodes, are indicated [17]. With permission from Nature Publishing.

different myelin proteins, a gene for IL-4 and the GpG Inverse vaccines with increased GpG, high antigen expression and
oligonucleotide [39]. The results of this experiment intracellular localization
were illuminating and quite positive. In chronic
In 2001, we reported that nonobese diabetic (NOD)
relapsing EAE, combining a myelin cocktail plus IL-
mice injected intramuscularly with a bacterial plas-
4-tolerizing DNA vaccine with a suppressive GpG-
mid encoding the insulin B chain peptide showed sig-
ODN induced a shift of the autoreactive T-cell re-
nificantly lower disease incidence and delayed onset
sponse towards a protective Th2 cytokine pattern
of disease when compared with controls [40]. Protec-
[39]. Myelin microarrays demonstrated that toleriz-
tion was mediated by insulin B (9–23)-specific down-
ing DNA vaccination plus GpG-ODN further de-
regulation of IFN-c. We have pursued this observation
creased antimyelin autoantibody epitope spreading
performing experiments with a succession of plasmid
and shifted the autoreactive B-cell response to a pro-
to optimize therapeutic efficacy. We gave particular
tective IgG1 isotype, indicating a shift from a Th1 re-
attention to installing GpG constructs in the plasmid
sponse to a Th2 response. Moreover, the addition of
backbone, to assessing whether higher levels of anti-
GpG-ODN to the inverse vaccine effectively reduced
gen expression were beneficial and to assessing
overall mean disease severity in both the chronic
whether localizing the site of expression to intracellu-
relapsing EAE and chronic progressive EAE mouse
lar locations was of benefit.
models [39]. Thus, a method was in hand that re-
duced both the frequency of relapses when given after
The backbone of the plasmid for the inverse vaccine
the first attack and overall reduced disease progres-
had its CpG motifs eliminated and reconstituted with
sion caused by the recurrent relapses.

ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451 445
 L. Steinman
| Inverse vaccination for autoimmunity therapy

Cocktail We identified pDNA-encoding proinsulin II as the

+ IL4 Cocktail Controls most potent at reducing type 1 diabetes in both pre-
ventive situations before hyperglycaemia was present
and in treatment models, where blood glucose levels
were between 190 and 250 mg mL–1. Furthermore,
we were able to elucidate an optimal frequency of dos-
> 20 000
ing, optimal levels of protein expression and whether
Cocktail + IL4-5
Cocktail + IL4-3
Cocktail + IL4-4
Cocktail + IL4-1
Cocktail + IL4-2

10 000
5000
2500 intracellular localization of the antigen contributed to
Cocktail-2
Cocktail-4
Cocktail-1
Cocktail-5
Cocktail-3
pTarget-4
pTarget-1
pTarget-3
pTarget-2

pTarget-5
Vehicle-2
Vehicle-3
Vehicle-1
Vehicle-5
Vehicle-4

1000 efficacy or not [41]. Targeting of the proinsulin II pro-

500
250 tein product to the cytoplasm by removing the signal
100
< 25 sequence from preproinsulin II proved to be the best
hMOG 71–83 strategy for delaying diabetes onset. Furthermore,
hMOG 70–82
hMBP protein the higher expression vector, created by adding a chi-
hMOG 68–80 meric intron upstream of the proinsulin II gene se-
hMOG 72–84
PLP 89–106 quence provided more robust protection from type 1
MOBP 21–39
hMBP 84–101 diabetes. This chimeric intron is composed of the 5¢
rMOG 13–30
hMBP 13–21 donor splice site from the first intron of the human b-
hMBP 1–20 globin gene and the branch and 3¢ acceptor splice site
hMBP 71–89
hMOG 1–26 from the intron of an Ig gene H chain V region. Weekly,
hMBP 6–14
gpMBP 72–85 biweekly and monthly dosing frequencies were all
rMBP 71–83 efficacious [41].
rMBP 72–84
hMBP 15–23
hMBP 14–22
hMBP 21–39 We undertook further studies to analyse the mecha-
rMBP 14–27 nism of action of this inverse vaccine. As reported,
hMOG 63–87
hMBP 148–156 administration of a single injection of proinsulin II
PLP 50–69
MOG 91–106 DNA to hyperglycaemic mice significantly decreased
MOG 35–55 the number of B9-23-specific, IFN-c producing cells
0.8 1.4 3.0 Relapse rate compared with the hyperglycaemic phosphate buf-
fered saline control mice (P = 0.02). Furthermore is-
Fig. 3 Tolerizing DNA vaccines reduce autoantibody epitope let-associated antibody (IAA) levels, reflecting humor-
spreading. At day 7 after onset of and after partial recovery al immune responses to insulin, were reduced in
from acute paralytic experimental autoimmune encephalo- NOD mice treated with the 10 lg dose of the inverse
myelitis (day 17) induced with proteolipid protein (PLP) vaccine [41].
139)151, SJL ⁄ J mice were treated weekly with phosphate
buffered saline vehicle, empty pTARGET vector, pTARGET
The conclusions from these studies on NOD mice,
expressing myelin basic protein, PLP, myelin oligodendrocyte
glycoprotein and MAG (cocktail) or pTARGET expressing the
formed the basis for the design of a multicentre phase
cocktail and interleukin-4. After the 10-week treatment, ser- I clinical trial testing pDNA encoding human proinsu-
um was obtained, array analysis carried out and Statistical lin II, termed BHT-3021, in individuals with a diagno-
Analysis of Microarrays was used to identify and analyse the sis of T1D, which began in October 2006.
hierarchical cluster to order antigen features [17]. With per-
mission from Nature Publishing
Human trials with an inverse vaccine in MS
An initial clinical trial with an inverse vaccine was at-
GpG motifs. Twelve CpG sequences were eliminated tempted in 30 individuals with RRMS and secondary
from the original plasmid by converting the cytosine progressive MS at four clinical sites in North America.
to guanosine. This CpG-reduced plasmid, where Although preclinical testing indicated that an inverse
CpGs were converted to GpGs was named pBHT1 vaccine using multiple plasmids encoding multiple
[41]. Using this plasmid, we studied NOD mice at var- myelin proteins would be optimal, after discussions
ious stages of the disease to compare plasmids with regulatory authorities in the United States and
encoding distinct islet antigens that have been impli- Europe, it was decided to test a vaccine, termed BHT-
cated in the pathogenesis of T1D. The pBHT1 back- 3009, encoding just one protein – MBP. BHT-3009
bone by itself, without an autoantigen insert could was administered intramuscularly at weeks 1, 3, 5
not modulate experimental type 1 diabetes mellitus and 9 after randomization into one of three groups:
in the NOD mice. placebo, BHT-3009 alone or BHT-3009 with the oral

446 ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451
 L. Steinman
| Inverse vaccination for autoimmunity therapy
statin drug, atorvastatin. Atorvastatin was added in
one group because there were indications that it may
act as an adjuvant to enhance the efficacy of inverse
immunization [42].
The inverse vaccine, BHT-3009, was safe and well tol-
erated. There were favourable trends on brain mag-
netic resonance imaging (MRI), with a trend towards
a reduction in gadolinium lesion activity and volume
after treatment relative to the placebo group. There
was no additional benefit from atorvastatin. We ob-
served beneficial antigen-specific immune changes
including a marked decrease in proliferation of IFN-c
producing, myelin-reactive CD4+ T cells from periph-
eral blood [Fig. 4] and a reduction in titres of myelin-
specific autoantibodies from CSF as assessed by pro-
tein microarrays [42].
A phase 2 trial was then performed in patients
with RRMS with the inverse vaccine, encoding
MBP, termed BHT-3009. As reported in 2008,
compared with placebo, in the 267 patient analy-
sis population the median 4-week rate of new
enhancing lesions during weeks 28–48 was 50%
lower with 0.5 mg of BHT-3009 (P = 0.07) and
during weeks 8–48 was 61% lower with 0.5 mg
of BHT-3009 (P = 0.05). The mean volume of
enhancing lesions at week 48 was 51% lower on
0.5 mg of BHT-3009 compared with the placebo
(P = 0.02). No significant improvement in MRI le-
sion parameters was observed with 1.5 mg of
BHT-3009. Dramatic reductions in 23 myelin-spe-
cific autoantibodies in the 0.5-mg BHT-3009 arm
were observed in CSF, but not with placebo or
1.5 mg of BHT-3009 [43]; see Fig. 5.
We had the opportunity to study a predefined sub-
set of 80 patients who contributed CSF for protein
array analysis at baseline, and CSF at week 44 for
repeat protein array analysis. This allowed us to
determine whether treatment with BHT-3009 had
an effect on the levels of antimyelin autoantibodies.
Treatment with the 0.5-mg BHT-3009 dose was
associated with a significant decrease in the auto-
antibody titres to 23 myelin autoantigens, whereas
treatment with placebo did not result in a statisti-
cally significant net change in any of the antimyelin
autoantibodies measured in the CSF. Antibody lev-
els to MBP peptides decreased with 0.5 mg of BHT-
3009. We also saw a fall in autoantibodies binding
to other components of the myelin sheath in RRMS
including aB-crystallin, PLP, MOG, MBOP and OSP.
The antibody levels of various components of the
myelin sheath decreased in a statistically significant
manner as determined by the Statistical Analysis of
Microarrays algorithm [43].
We studied whether baseline antibody profile in the
CSF predicted response to the study drug. In a pro-
spectively defined protocol we described that ‘in the
upper half of the anti-MBP reactive patients (n = 13
on placebo, n = 11 on 0.5-mg BHT-3009 and n = 14
on 1.5-mg BHT-3009), there was a significantly lower
number of new Gd-enhancing lesions per patient in
MRIs of weeks 28–48 with 0.5 mg of BHT-3009 com-
pared with placebo (mean ± SD, 2.5 ± 4.03 on 0.5-
mg BHT-3009 vs. 3.3 ± 4.59 on placebo; P = 0.02). In
contrast, in this subgroup, there was no significant
difference between 1.5-mg BHT-3009 and placebo
[43]. Further analysis of baseline antibody profiles
has allowed us to identify a serum antibody profile
involving MBP where high responders not only had
significant reductions in parameters of MRI, but also
had reductions in relapse rate in the 0.50-mg group
(in preparation). Further trials are planned in both
patients with RRMS as well as those patients preiden-
tified by their high responder profiles in a serum as-
say identifying the detailed immune response to dif-
ferent components of MBP.
Human trials with inverse vaccines in type 1 diabetes
In June 2009, Gottlieb and colleagues reported
results of a trial with an inverse vaccine encoding pro-
insulin in patients, aged 18–40 with type 1 diabetes.
Patients were randomized into placebo or treatment
groups and were given 12 weekly injections of various
doses of BHT-3021. In all dosage groups at 6 months
we reported stability of C-peptide as compared with
levels seen in the placebo group. Levels of haemoglo-
bin A1c, were stabilized at all doses compared with
the placebo. Full results of the trial at 1 year will be re-
ported in 2010. Based on the results reported in June
2009, a partnership to continue development of this
inverse vaccine for type 1 diabetes mellitus was con-
summated between Bayhill Therapeutics and Genen-
tech, a division of Roche [44].
Further human trials with inverse vaccines
In addition to further trials in RRMS and type 1 diabe-
tes mellitus, a plasmid to the a-chain of the acetylcho-
line receptor was effective in models of experimental
autoimmune myasthenia gravis models in mice and
rats. This approach will now be taken forward into
the clinic in patients with myasthenia gravis. Another
neurological condition, neuromyelitis optica, where
there is a strong autoantibody response targeting the
ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451 447
 L. Steinman
| Inverse vaccination for autoimmunity therapy

TT MBP83-99 PLP peptide mix

104 104 104
20.0 5.36 25.9 7.33 14.5 5.76

103 103 103

Baseline
102 102 102

101 101 101

8.31 65.5 7.52 59.2 11.7 68.1

100 0 100 100 0
1 2 3 4
10 10 10 10 10 100 101
102
103
10 4
10 10 1
102
10 3
104

104 104 104

20.5 3.24 1.2 4.32 1.75 4.56

103 103 103

Week-9
IFN-γ

102 102 102

101 101 101

12.5 57.8 4.98 89.5 5.2 68.5

100 100 100
100 10 1
10 2
10 3
104
100 101
102
103
10 4
100 10 1
102
10 3
104

104 104 104

26.4 1.6 0.15 2.56 0.36 2.83

103 103 103

Week-50
102 102 102

101 101 101

13.7 58.3 0.66 96.6 0.29 96.5

100 0 100 100 0
10 101 102 103 104 100 101 102 103 104 10 101 102 103 104

CFSE

Fig. 4 Example of decreased T-cell response with BHT-3009. An example of one patient whose myelin basic protein
(MBP)- and proteolipid protein (PLP)-specific T-cell proliferative response decreased in response to BHT-3009 is shown.
Proliferation was measured using a dye dilution method with the vital dye 5,6-carboxyfluorescein diacetate succinimidyl
ester. Peripheral blood mononuclear cells were incubated with a variety of antigens and controls, but for simplicity, only
the responses to tetanus toxoid (TT), MBP83-99 peptide and a PLP peptide mix are shown. The upper three panels corre-
spond to the baseline response; the middle three, to the week 9 response; and the bottom three, to the week 50
response. Proliferating interferon (IFN)-positive CD4+ T cells are shown in the upper left quadrant of each fluorescent-
activated cell sorter plot. Numbers in red indicate the percentage of cells in each quadrant. A dramatic decrease in IFN-
positive cells specific for MBP and PLP is demonstrated by week 9 and persists till week 50. Importantly, the response
to TT is unchanged with dosing, confirming the antigen-specific nature of BHT-3009 [42]. Permission obtained from
Archives of Neurology.

448 ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451
 L. Steinman
| Inverse vaccination for autoimmunity therapy

Fig. 5 Treatment with 0.5 mg of BHT-3009 is associated with a reduction in antimyelin antibody titres. Myelin array analysis
was performed to quantify antimyelin peptide antibodies in baseline and post-treatment samples in patients treated with 0.5 mg
of BHT-3009. Antibodies with pre- to post-treatment changes within each treatment group are represented either as fold change in
intensity with increases false coloured red, no change false coloured black, and decreases false coloured green. Grey represents
samples with no data available. Significant Analysis of Microarrays was applied to identify statistical differences in antibody
reactivity between the pre- and post-treatment samples, and identified 23 myelin peptides (listed to right of heat map) that exhib-
ited overall significant changes (all were decreases) in the post-treated as compared with pretreated cerebrospinal fluid samples
in patients treated with 0.5 mg of BHT-3009 (false discovery rate, q < 0.1). There were no statistically significant differences be-
tween pre- and post-reactivity in placebo-treated patients. Myelin basic protein peptide sequences are derived from and numbered
based on the 18.5-kDa isoform (Genbank accession no.: AAA59562), with the exception of peptides denoted by *s which are num-
bered based on the 21.5-kDa isoform (Genbank accession no.: AAA59564). MOG, myelin oligodendrocyte glycoprotein;
MOBP, myelin-associated oligodendrocytic basic protein; OSP, oligodendrocyte-specific protein. With permission from Annals of
Neurology.

water channel, aquaporin-4 is being planned. As au-
toantigens are identified in other diseases inverse
vaccines will be developed to treat these conditions in
time. Initially diseases where there is clear evidence
of a dominant immune response against one protein
will have the highest priorities. Over time, as we get
more experience with inverse vaccines in man, we
shall attempt therapy with multiple plasmids to
inhibit responses to several autoantigens on a broad
front. One hope is that we will be able to use this
approach for reducing specific immune responses
instead of current therapies for autoimmune
diseases (Fig. 6).
Many of the strongest drugs now applied to treat au-
toimmunity were first approved for cancer therapy or
have shown promise for reducing transplant rejec-
tion. Such therapies are associated with severe
opportunistic infections such as progressive multifo-
cal leukoencephalopathy, which has been reported
with Rituxan and Natalizumab [4, 12–14]. There are
other complications with these drugs that have been

Fig. 6 CpG motifs in red are re-

duced and replaced by GpG mo-
tifs shown in blue. DNA encoding
autoantigen is inserted into the
GpG-enriched plasmid. Trials
are planned to perform inverse
vaccination to acetylcholine rec-
eptor in myasthenia gravis, and
to aquaporin-4 in neuromyelitis
optica.

ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451 449
 L. Steinman
| Inverse vaccination for autoimmunity therapy
taken from the realms of cancer therapy or therapy to
reduce transplant rejection. These complications in-
clude increased risk of cancer and other opportunis-
tic infections besides progressive multifocal leukoen-
cephalopathy. One drug, the anti-CD52 approved for
chronic lymphocytic leukaemia has shown stellar re-
sults in phase 2 trials for RRMS. Yet, 40% of patients
taking the drug for RRMS develop thyroiditis, either
Hashimoto’s, Graves or idiopathic thrombocytopae-
nic purpura [6, 7]. Another drug anti-CD3, approved
for transplant therapy, induced skin rashes and acti-
vation of Epstein–Barr virus in patients undergoing
therapy for type 1 diabetes [8–10].
In contrast, inverse vaccination offers the promise of
specifically turning the ‘off’ switch’ for unwanted im-
mune responses, while leaving the rest of the immune
system intact. These unwanted complications from
the current therapies are perhaps inextricably tied to
the fact that they were originally developed to inten-
tionally modulate important components of the im-
mune response, some of them being necessary for
effective immune surveillance.
It is wise to remember at this early juncture in the
development of inverse vaccination, what Jenner said
over three centuries ago,
Thus far have I proceeded in an inquiry founded, as it
must appear, on the basis of experiment; in which,
however, conjecture has been occasionally admitted
in order to present to persons well situated for such
discussions, objects for a more minute investigation.
In the mean time I shall myself continue to prosecute
this inquiry, encouraged by the hope of its becoming
essentially beneficial to mankind’ [1].
Only further experimentation in man with these in-
verse vaccines will allow us to see whether this new
variation of Jenner’s idea of vaccination will achieve
its intended goal to specifically reduce unwanted
autoimmune responses, leaving the rest of the im-
mune system intact.
Acknowledgements
The author’s colleagues William Robinson, P.J. Utz
and Hideki Garren have been working on this every
step of the way since their first meeting at Stanford in
the late 1990s. Peggy Ho and Paulo Fontoura made
seminal discoveries on the GpG motif. Colleagues at
Bayhill Therapeutics particularly Nanette Solvason,
Michael Leviten, Robert King and JoAnne Quan have
played critical roles in the development of DNA
450 ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451
vaccines. None of this would have happened without
support from the National Institutes of Health, the
Juvenile Diabetes Research Foundation, the Na-
tional Multiple Sclerosis Society and the loyal inves-
tors in Bayhill Therapeutics.
Conflict of interest statement
The author is on the Board of Directors and serves as
a consultant to Bayhill Therapeutics, the developer of
DNA vaccines in clinical trials for MS and type 1 dia-
betes mellitus.
References
1 Jenner E. The Three Original Publications on Vaccination against
Smallpox, The Harvard Classics. New York: P.F. Collier & Son,
Bartleby Classics, 2001.
2 Garren H. DNA vaccines for autoimmune vaccines. Expert Rev
Vaccines 2009; 8: 1195–203.
3 Hauser SL, Waubant E, Arnold DL et al. B-cell depletion with Ri-
utximab in relapsing-remitting multiple sclerosis. N Engl J Med
2009; 358: 676–88.
4 Food and Drug Administration. FDA Public Health Advisory: Life-
Threatening Brain Infection in Patients with Systemic Lupus
Erythematosus after Rituxan (Rituximab) Treatment. Rockville,
MD: Food and Drug Administration, 2006. Accessed 18
January 2008, at: http://www.fda.gov/cder/drug/advisory/
rituximab.htm.
5 Cohen SB, Emery P, Greenwald MW et al. Rituximab for rheuma-
toid arthritis refractory to anti-tumor necrosis factory therapy:
results of a multicenter, randomized, double-blind, placebo-con-
trolled, Phase III trial evaluating primary efficacy and safety at
twenty-four weeks. Arthritis Rheum 2006; 54: 2793–806.
6 The CAMMS223 Trial Investigators. Alemtuzumab vs. interferon
beta-1a in early multiple sclerosis. N Engl J Med 2008;
359:1786–801.
7 Coles AJ, Wing M, Smith S et al. Pulsed monoclonal antibody
treatment and autoimmune thyroid disease in multiple sclerosis.
Lancet 1999; 354: 1691–5.
8 Keymeulen B, Candon S, Fafi-Kremer S et al. Transient Epstein
Barr virus reactivation in CD3 monoclonal antibody-treated pa-
tients. Blood 2010; 115: 1145–55.
9 Herold KC, Hagopian W, Auger JA et al. Anti-CD3 monoclonal
antibody in new-onset type 1 diabetes mellitus. N Engl J Med
2002; 346: 1692–8.
10 Keymeulen B, Vandemeulebroucke E, Ziegler AG et al. Insulin
needs after CD3-antibody therapy in new-onset type 1 diabetes.
N Engl J Med 2005; 352: 2598–608.
11 Polman CH, O’Connor PW, Havrdova E et al. A randomized, pla-
cebo-controlled trial of natalizumab for relapsing multiple sclero-
sis. N Engl J Med 2006; 354: 899–910.
12 Langer-Gould A, Atlas SW, Green AJ, Bollen AW, Pelletier D. Pro-
gressive multifocal leukoencephalopathy in a patient treated
with natalizumab. N Engl J Med 2005; 353: 375–81.
13 Steinman L. Blocking adhesion molecules as therapy for multiple
sclerosis: natalizumab. Nat Rev Drug Discov 2005; 4: 510–19.
14 Linda H, von Heijne A, Major ED et al. Progressive multifocal leu-
koencephalopathy after natalizumab monotherapy. N Engl J
Med 2009; 361: 1081–7.
 L. Steinman
| Inverse vaccination for autoimmunity therapy
15 Kappos L, Antel J, Comi G et al. Oral fingolimod [FTY720] for
relapsing multiple sclerosis. N Engl J Med 2006; 355: 1124–
40.
16 Robinson WH, DiGennaro C, Hueber W et al. Antigen arrays for
multiplex characterization of autoantibody responses. Nat Med
2002; 8: 295–301.
17 Robinson WH, Fontoura P, Lee BJ et al. Reverse genomics: pro-
tein microarrays guide tolerizing DNA vaccine treatment of auto-
immune encephalomyelitis. Nat Biotechnol 2003; 21: 1033–9.
18 Kanter J, Narayana S, Ho P et al. Lipid microarrays identify key
mediators of autoimmune brain inflammation. Nat Med 2006;
12: 138–43.
19 Ousman SS, Tomooka BH, Van Noort JM et al. Protective and
therapeutic role for aB-crystallin in autoimmune demyelination.
Nature 2007; 448: 474–9.
20 Wolff JA, Malone RW, Williams P et al. Direct gene transfer into
mouse muscle in vivo. Science 1990; 247: 1465–8.
21 Ulmer JB, Donnelly JJ, Parker SE et al. Heterologous protection
against influenza by injection of DNA encoding a viral protein. Sci-
ence 1993; 259: 1745–9.
22 Smith JM, Amara RR, McClure HM et al. Multiprotein HIV type 1
clade B DNA ⁄ MVA vaccine: construction, safety, and immunoge-
nicity in Macaques. AIDS Res Hum Retroviruses 2004; 20: 654–
65.
23 Osinubi MO, Wu X, Franka R et al. Enhancing comparative ra-
bies DNA vaccine effectiveness through glycoprotein gene modifi-
cations. Vaccine 2009; 27: 7214–18.
24 Martin JE, Sullivan NJ, Enama ME et al. A DNA vaccine for Ebola
virus is safe and immunogenic in a Phase I clinical trial. Clin Vac-
cine Immunol 2006; 13: 1267–77.
25 Waisman A, Ruiz PJ, Hirschberg DL et al. Suppressive vaccina-
tion with DNA encoding a variable region gene of the T cell recep-
tor prevents autoimmune encephalomyelitis and activates Th2
immunity. Nat Med 1996; 2: 899–906.
26 Lobell A, Weissert R, Storch MK et al. Vaccination with DNA
encoding an immunodominant myelin basic protein peptide tar-
geted to Fc of immunoglobulin G suppresses experimental auto-
immune encephalomyelitis. J Exp Med 1998; 187: 1543–8.
27 Ruiz P, Garren H, Ruiz I et al. Suppressive immunization with
DNA encoding a self-peptide prevents autoimmune disease:
modulation of T cell co-stimulation. J Immunol 1999; 162: 3336–
41.
28 Garren H, Ruiz P, Watkins TA et al. Combination of gene delivery
and DNA vaccination to protect from and reverse Th1 autoim-
mune disease via deviation to the Th2 pathway. Immunity 2001;
15: 15–22.
29 Krieg AM, Yi AK, Matson S et al. CpG motifs in bacterial DNA
trigger direct B-cell activation. Nature 1995; 374: 546–9.
30 Sato Y, Roman M, Tighe H et al. Immunostimulatory DNA se-
quences necessary for effective intradermal gene immunization.
Science 1996; 273: 352–4.
31 Tsunoda I, Tolley ND, Theil DJ, Whitton JL, Kobayashi H, Fuji-
nami RS. Exacerbation of viral and autoimmune animal models
for multiple sclerosis by bacterial DNA. Brain Pathol 1999; 9:
481–93.
32 Lobell A, Weissert R, Eltayeb S, de Graaf KL et al. Suppressive
DNA vaccination in myelin oligodendrocyte glycoprotein peptide-
induced experimental autoimmune encephalomyelitis involves a
T1-biased immune response. J Immunol 2003; 170: 1806–13.
33 Steinman L. A brief history of TH17, the first major revision in the
TH1 ⁄ TH2 hypothesis of T cell-mediated tissue damage. Nat Med
2007; 13: 139–45.
34 Voorthuis JA, Uitdehaag BM, De Groot CJ, Goede PH, van der
Meide PH, Dijkstra CD. Suppression of experimental allergic
encephalomyelitis by intraventricular administration of inter-
feron-gamma in Lewis rats. Clin Exp Immunol 1990; 81: 183–8.
35 Panitch HS, Hirsch RL, Schindler J, Johnson KP. Treatment of
multiple sclerosis with gamma interferon: exacerbations associ-
ated with activation of the immune system. Neurology 1987; 37:
1097–102.
36 Segal BM, Chang JT, Shevach EM. CpG oligonucleotides are po-
tent adjuvants for the activation of autoreactive encephalitogenic
T cells in vivo. J Immunol 2000; 164: 5683–8.
37 Boccaccio G, Mor F, Steinman L. Non-coding plasmid DNA in-
duced gamma interferon in vivo and suppresses autoimmune
encephalomyelitis. Int Immunol 1999; 11: 289–96.
38 Ho P, Fontoura P, Ruiz P, Steinman L, Garren H. An immuno-
modulatory GpG oligonucleotide for the treatment of autoimmu-
nity via the innate and adaptive immune systems. J Immunol
2003; 171: 4920–6.
39 Ho P, Fontoura P, Platten M et al. A suppressive oligodeoxynucle-
otide enhances the efficacy of myelin cocktail ⁄ IL-4 tolerizing DNA
vaccination and treats autoimmune disease. J Immunol 2005;
175: 6226–34.
40 Urbanek-Ruiz I, Ruiz P, Garren H et al. Immunization with DNA
encoding an immunodominant peptide of insulin prevents diabe-
tes in NOD mice. Clin Immunol 2001; 100: 164–71.
41 Solvason N, Lou YP, Peters W et al. Improved efficacy of a toleriz-
ing DNA vaccine for reversal of hyperglycemia through enhance-
ment of gene expression and localization to intracellular sites. J
Immunol 2008; 181: 8298–307.
42 Bar-Or A, Vollmer T, Antel J et al. Induction of antigen-specific
tolerance in multiple sclerosis after immunization with DNA
encoding myelin basic protein in a randomized, placebo-con-
trolled Phase 1-2 trial. Arch Neurol 2007; 64: 1407–15.
43 Garren H, Robinson W, Krasulová E et al. Phase 2b trial of a DNA
vaccine encoding myelin basic protein in relapsing multiple scle-
rosis. Ann Neurol 2008; 63: 611–20.
44 Young D. Bayhill Therapeutics snags $350 million deal with
Genentech. Bioworld Today 2009; 20: 1.
Correspondence: Lawrence Steinman, Department of Neurology and
Neurological Science, Interdepartmental Program in Immunology,
Stanford University, Stanford, CA 94305, USA.
(fax: 650-725-0627; e-mail: Steinman@stanford.edu).
ª 2010 Blackwell Publishing Ltd Journal of Internal Medicine 267; 441–451 451