Immunologic Research

, Volume 39, Issue 1, pp 4–14

Structural requirements and applications of inhibitory oligodeoxyribonucleotides

Article

DOI: 10.1007/s12026-007-0065-4

Cite this article as:
Ashman, R.F. & Lenert, P. Immunol Res (2007) 39: 4. doi:10.1007/s12026-007-0065-4

Abstract

Synthetic oligodeoxyribonucleotides (ODN) bearing certain sequence characteristics mimic bacterial DNA by activating B cells and dendritic cells through Toll-like receptor (TLR) 9, an event that potentiates both humoral and cell-mediated immunity. ODN sharing some of the sequence characteristics of strong stimulatory (ST-) ODN, but substituting GGG for CGTT, competitively inhibit ST-ODN-driven events. An ODN with the same length and base composition as a strong ST-ODN, but lacking both ST- and IN-sequence requirements, has neither ST- nor IN-activity. Whereas, certain sequence changes strongly influence ST-ODN activity in human cells relative to mouse cells and B cells relative to non B cells, the strongest IN-ODN appear to work well in both species and multiple cell types. Converting from the natural phosphodiester backbone to a nuclease-resistant phosphorothioate backbone increases the sensitivity to ST-ODN about 2 logs and to IN-ODN 3 logs, while increasing the impact of critical base changes in ST-ODN and diminishing it in IN-ODN. Examples where IN-ODN have been used in vivo to interrupt autoimmune and other TLR-9-induced inflammatory states are described.

Keywords

Inhibitory oligonucleotides TLR9 Stimulatory oligonucleotides Autoimmunity Bacterial DNA 

Introduction

One of the features of the mammalian immune system most conserved in evolution is a set of receptors called Toll-Like Receptors (TLRs), because of their resemblance to the Toll molecule of fruit flies. The striking conservation of their structure between insects and humans testifies to their usefulness in innate immune defense. Each of our TLRs recognizes a different molecular pattern typical of pathogens but absent from mammals. For example, TLR9 recognizes a DNA sequence motif centering on an unmethylated CpG pair which is much rarer in mammalian DNA than in microbial DNA [1].

In mice and humans, TLR9 expression is largely restricted to antigen-presenting cells. B cells respond directly by proliferation, cytokine and IgM production, and protection from apoptosis. Type 1 interferons from CpG-stimulated dendritic cells trigger interferon γ from NK cells, strongly biasing T cell cytokine production toward TH1.

Clinical applications

These properties have created much interest in the development of synthetic stimulatory oligonucleotides (ST-ODN) as therapeutic agents, including (a) vaccine adjuvants that simultaneously stimulate cell-mediated and humoral immunity, (b) anticancer agents, acting by potentiating T-cell responses to tumor antigens that have been weakened by tumor products, (c) anti-allergy agents, by inhibiting TH2 cytokine production. These properties of CpG ODN have been well-reviewed [2, 3]. On the other hand, there is evidence that innate immune responses through TLRs may contribute to autoimmunity [4], so that inhibitors of TLRs may also prove to have therapeutic applications [5].

CpG signaling pathways

CpG signaling has been explored both in primary cells and in cells transfected with TLR9. Resting cells express TLR9 diffusely in their endoplasmic reticulum. Synthetic single-stranded ODNs enter the cell rapidly by a facilitated diffusion process, concentrating in endosomal vesicles, to which TLR9 is rapidly recruited [6, 7]. The MyD88 signaling intermediate is also then recruited, as the vesicles mature and acidify [6]. MyD88 −/− and TLR9 −/− cells do not respond to CpG DNA [8, 9]. Direct binding of a TLR9-IgFc fusion protein to CpG-ODN takes place [10], but there is as yet no evidence that the avidity of binding to TLR9 quantitatively determines biologic activity.

Downstream signaling from TLR9/MyD88 is only partially understood. PI3 kinase is activated, acting on rab 5 [11]. Disappearance of IkBa is associated with the transport of NFkB subunits to the nucleus, including p50, p65, and cRel in B cells treated with ST-ODN [12]. Activation of MAP kinases, Jun kinase, the AP-1 complex and NF-IL6 occurs [13]. Other transcription factors may also be involved in the complex biologic responses to ODN, which are dramatically different in different cell types. For example, only ODN-driven dendritic cells use STATs 1 and 4 to generate INFα/β.

Structural requirements for ST-ODN

Though equally dependent on TLR9, B cells and non B cells differ in their preference for single stranded ODNs with nuclease-resistant phosphorothioate backbones (B cells) versus ODN with secondary structure, featuring a central core with a phosphodiester backbone, capable of internal hydrogen bonding, plus mutually adhesive poly G ends [14]. In B cells, the unmethylated CpG is strongly required for activity, but activity is enhanced by flanking the CpG with two purines on the 5′ side and two pyrimidines on the 3′ side [15]. We further refined this scheme to show the requirement for TCC at the 5′ end and for the absence of Cs one or two bases 5′ to CG [16]. Thus, in mouse B cells the strongest ST-ODN with a single motif was #2084 (TCCTGACGTTAAGT) with 50% maximal activity at 14 nM for apoptosis protection, 34 nM for G1 cell cycle entry at 16 h, and 119 nM for IL6 production at 72 h. ST-ODN with phosphodiester backbones perform the same functions, but these require about a 100-fold higher concentrations [16, 17]. Human B cells require much higher ST-ODN concentrations than mouse. While titration curves have not been published, single motif ODN are generally used in the 600–1,000 nM range with thioate backbones and 100-fold higher with diester backbones [18]. Human cells also differ from mouse cells in preferring longer ST-ODN with three motifs instead of one, a T instead of A adjacent to the CG on the 5′ side, and a 5′ end beginning with TCG instead of TCC [18, 19]. Upregulation of CD86 on B cells, IFNα/β secretion by plasmacytoid dendritic cells (almost exclusively), and CD69 upregulation on NK cells (responding to IFN) have been the most sensitive indicators of ST-ODN activity in human cells. But with natural DNA (diester backbone) the advantages of multiple motifs is much less.

The single stranded ODNs preferred by B cells have been dubbed class B [19] or K [20], and the ODN with secondary structure preferred by non B cells have been called class A [14, 21] or D [20]. Fortunately, the desirable properties of activating B cells and generating interferons from plasmacytoid dendritic cells can be combined in Type C which have a combination of class A and B structural features [22], a potential advantage in human therapeutics.

Inhibitory ODN

While studying the impact of sequence changes on the stimulatory activity of synthetic ODN in mouse B cells we discovered that as few as two or three base changes could cause a ST-ODN to become inactive. When mixed with a ST-ODN, such an ODN proved inhibitory; for example, at a 1:1 molar ratio, the strongest inhibitory (IN-) ODN produced 80–90% reduction in activity for B cells. The concentrations required for half-maximal inhibitory activity in B-cells were highest for apoptosis protection, intermediate for GI entry and lowest for IL6 production, the opposite trend to that observed with ST-ODN [23]. The inhibition was fully reversible by adding more ST-ODN. Inhibition of cell cycle entry, apoptosis protection, IL-6 secretion, NFkB nuclear translocation, and generation of AP1, indeed all readouts studied to date, occurred only when they were driven by ST-ODN and not by LPS, anti CD40, or IL4 + anti Ig [12, 13, 23]. IN-ODN titrations produced symmetrical sigmoid curves, indicating control of all biologic readouts by a population of receptors with homogeneous sensitivity to ODN [23, 24]. This behavior strongly suggested competition between IN-ODN and ST-ODN for a uniform populations of binding sites. Most likely, these sites are on TLR9, since TLR9 expression is both necessary for ODN activity in gene knockout studies [25] and sufficient in transfection studies [9], and since direct binding of ODN to TLR9 has been shown [10].

Of the 15 bases in our prototype strongest IN-ODN, at only six positions did changing a base have a major (>2x) effect on activity [24]. The most active sequence was xCC n notC notC nn GGG nnn where n is any base, x is optional, and notC is any base except C. The underlined portions we call the “critical bases”. CC and GGG must be separated by four or five intervening bases, and must be covalently linked [16]. The central CG required by ST-ODN is replaced by nn. This formula pertains not only for the sequences 2088 [12, 23] and 2114 [24], used by ourselves and by others, [4, 26, 27, 28], but also to those used by the Coffman lab [29]. Other inhibitory sequences developed independently by Klinman share only some of the features described here [30, 31]. While adding features of our formula to the Klinman inhibitors increases activity, and subtracting features decreases activity [24], nevertheless it is apparent that our formula does not describe the whole universe of inhibitory sequences. This situation is reminiscent of studies of the fine structure of epitopes reacting to the combining sites of monoclonal antibodies, where sometimes unrelated epitope sequences could compete for binding to the same site through a different selection of non-covalent interactions. For example, the sequence (TTAGGG)3, has about 60% of the inhibitory activity of 2114 [24, 31], and is the most frequent six-base sequence in mammalian DNA, because it is the sequence of the telomere repeats at the end of chromosomes [32]. TTAGGG thus contributes to the excess of potentially inhibitory over stimulatory sequences in mammalian DNA, which confers its ability to inhibit stimulation by CpG-DNA [33].

The structural formula for optimal IN-ODN activity in B cells prompted us to reexamine the requirements for stimulatory activity [16], already partially worked out long ago [1, 2]. We found ST-ODN and IN-ODN shared the requirement that the 5′-half should contain CC and notC notC in the appropriate positions. However, they differed in that (unmethylated) CGTT was required for stimulation in place of n GGG required for inhibition, that the 5′ T was essential instead of optional, and that truncating the 3′ end had more severe effects, mainly related to the shorter length [16]. These findings suggested that the 5′ half contained the portions of IN-ODN and ST-ODN that compete for binding to TLR9, whereas, the central portion determined whether the first signal intermediate (MyD88) was engaged. Explanations of ODN-induced TLR9 signaling include a conformational change in TLR9 [24], or an associative model where the orientation of two TLR9s into a dimer with a site for MyD88 is effected by CGTT and prevented by GGG [16]. These mechanisms might both be correct.

Since type A/K (preferred by dendritic cells) and B/D (preferred by B cells) ST-ODN have such different structural properties, we expected the structural requirements for inhibition of these types of ST-ODN would be different. This expectation was convincingly refuted by the data [24]. Even the requirement for the 5′ CC is preserved in IN-ODN when they oppose Type A ST-ODN, though the prototype Type A ST-ODN lacks the CC [14, 21]. The mechanism of inhibition is clearly more complicated than our simple models suggest, and it may require crystallizing TLR9 with and without various ODN ligands to sort it out. IN-ODN, as well as ST-ODN does bind directly to the TLR9 IgFc fusion protein (E. Latz and D. Golenbock, personal communication).

At present, our concept of TLR9 structure derives from the published crystallographic model of the TLR3 ectodomain, which binds double-stranded RNA [34]. The series of hydrophobic amino acids recurring at a regular three base interval more than 20 times (leucine-rich repeats) is the most striking feature of the primary structure of TLR3, TLR9 and other TLRs. In their model of TLR3, these hydrophobic amino acids bind firmly to one another at the center of a spiral coil stabilized also by hydrogen bonds [34]. The authors identify positively charged surfaces that could provide an opportunity for negatively charged RNA to alter the conformation of the TLR3 dimer [34]. TLR9 differs by having a stretch in the middle free of hydrophobic groups [35] which could form a hinge between two smaller rigid solenoids. Golenbock and Latz have used FRET analysis to demonstrate that ST-ODNs not only cause a closer approach of TLR9 molecules to one another inside the endosomal vesicle, but also bring the distal coil closer to the membrane proximal coil, probably representing motion at the hinge (D. Golenbock and E. Latz, manuscript submitted). Studies introducing IN-ODN into this system are in progress.

Relevance of TLR9 species and cell type

Despite the high degree of homology between the primary structures of TLR9 in mouse and man, the ineffectiveness of the strongest ST-ODN for mouse (example: 2084) in human cells was disappointing. Extensive comparisons led to the routine use of the phosphorothioate S-backbone ODN 2006, possessing a 5′ TC and 3 GTCGTT motifs separated by TT spacers [18]. The GTCGTT motif was superior in human cells to GACGTT, although in mouse cells, the substitution of T for A causes a six-fold reduction in activity [16]. Omitting one of the TT spacers made little difference to its activity in human cells, but greatly decreased its activity in chimp and rhesus monkey cells [19]. These early findings highlighted the importance of maintaining the CGTT motif across species, but also the importance of species differences.

Our first attempt to construct IN-ODN for human cells was to convert one or more of the CGTTs in 2006 to CGGG, an inhibitory motif for mouse. However, in Namalwa cells (a human B cell line) none of these ODN had more than 5% of the inhibitory activity of the mouse prototype IN-ODN 2114 (TCCTGGAGGGGAAGT). Table 1 compares the inhibitory activity of a panel of IN-ODN when opposing 1 μM 2006 in Namalwa cells or 100 nM 2084 in mouse spleen B cells. In general, changes in the critical bases that impair activity in mouse also does so in human. ODN 2310 has no inhibitory (or stimulatory) activity in either mouse or human. It has C at the “forbidden” location and neither complete inhibitory nor stimulatory motifs; it also has the same base composition and length as the ST-ODN 2084 [23]. Examples where inhibitory activity differs between mouse and human are mostly ODN with truncated ends. Table 2 shows that one of these, the truncation variant 4171 (10 mer), shows the most variation in activity between three cell types expressing human TLR9, whereas IN-ODN with critical residue substitutions had less activity, not only in the three cell types expressing human TLR9, but also in the two cell types expressing mouse TLR9. Specifically, the activity of 4171 is much greater in TLR9-transfected HEK cells than in primary cells. Since these differences between responses of HEK and primary cells to IN-ODN cannot be ascribed to TLR9, they may be due to differences in accessory molecules that alter the signaling responses of TLR9 to different ODN. IN-ODN 4171 might prove useful in investigating such differences (Table 3).
Table 1

Comparison of IN-ODN activity in mouse and human cells

IN-ODN#

Sequence

% Activity in human

% Activity in mouse

2114

T

C

C

T

G

G

A

G

G

G

G

A

A

G

T

100

100

2088

C

100

100

2310

C

T

T

0.2

<1

4028

X

X

C

6

1

4265

X

C

100

52

4266

X

A

10

119

4141

C

C

5

6

4263

T

T

2

13

4262

A

A

 

1.6

3

4115

A

5

23

4110

A

2

1

4171

X

X

X

X

X

100

23

4004

C

1

1

4005

T

17

26

4008

T

3

40

% activity in human was based on concentration inhibiting CD86 induction by 50% in Namalwa cells by FACS, stimulated by 100 nM ST-ODN 2006. % activity in mouse was determined by 50% inhibition of apoptosis protection, GI entry, and IL-6 production stimulated by 100 nM ST-ODN 2084. The potency of 2114 was set at 100% activity. Mean of three experiments

Table 2

Inhibitory ODN: truncation variants tested in mouse B cells, backbone effect

# bases

 

O backbone

S backbone

15

TCCTGGAGGGGAAGT

4228

100% (prototype)

2114

100%

14

_CCTGGAGGGGAAGT

4329

101%

4033

393%

14

_CCAGGAGGGCAAGT

4335

152%

4266

119%

13

TCCTGGAGGGGAA__

4315

85%

4030

44%

12

_CCTGGAGGGGAA__

4330

128%

4104

120%

10

_CCTGGAGGGG_____

4255

100%

4171

31%

9

_CCTGGAGGG_______

4336

92%

4320

12%

9

__CTGGAGGGG______

4334

0.5%

4172

<1%

Mouse spleen B cells were cultured either: (a) with 10 μM ODN 1916, with and without a series of concentrations of truncated IN-ODN with O backbones, or (b) with 100 nM ODN 2084, with and without a series of concentrations of truncated IN-ODN with S backbones, with the same sequences as the IN-ODN in (a). ODNs 2084 and 1916 also have the same sequence. Apoptosis protection, G1 entry and IL6 secretion were determined. The concentrations providing 50% of maximal inhibition were determined, and the geometric mean calculated. The percent activity of a test IN-ODN relates to the activity of the prototype IN-ODN 2114/4428, which is set at 100%

Table 3

Inhibitory ODN activity in human and mouse cells

 

HEK/hu (%)

Namalwa (%)

Blood B (%)

HEK/mo (%)

Mo B (%)

Prototype

2114

100

100

100

100

100

X-7, A-4

4266

12.5

10

19

1100

119

X-7 +5 to 8

4171

75

100

45

280

23

A-6

4115

1.5

5

9

18

23

T+3

4008

1.5

3

3

6

50

T+3

4005

5

17

12

1.4

26

Neutral

2310

0.3

0.2

0.5

0.3

0.2

The percentage of the activity of the prototype IN-ODN 2114 was measured and calculated as in Table 2. HEK (Human embryonic kidney cells) transfected with human (hu) or mouse (mo) TLR9 and an NFκB-binding promoter attached to the luciferase gene were used to provide a luminescence response to CpG stimulation. The measured intensity of CD86 expression accessed by surface immunofluorescence was the activity readout for human peripheral blood B cells and the Namalwa B cell line. Human cells were stimulated with ST-ODN 2006 (100 nM) and mouse cells with 2084 (100 nM) for 24 h, with or without a series of concentrations of IN-ODN with S backbones representing truncation (4266 and 4171) and critical base (4115, 4008, and 4005) changes from the prototype (2114). The sequence of neutral ODN 2310 is TCCTG CAGGTTAAGT [23]

Effect of phosphorothioate (S) versus phosphodiester (O) backbones

The phosphodiester (O), backbone of natural DNA is flexible and highly susceptible to degradation by DNAases. One sulfur substitution in each phosphate group (phosphoothioate), renders it stiffer and much less susceptible to degradation. Therefore cell biology studies mainly use ODN with the O backbone, but drug development studies use ODN with the phosphorothioate (S) backbone. Both require TLR9 for activity. In mouse cells, equivalent stimulation requires about 100× as much when the backbone is O than when the same base sequence is S [23], a difference usually attributed to the degradation rate (without direct measurement). The IN-ODN 2114 concentration needed to inhibit 100 nM 2084 (S backbone) and 10 μM 1916 (same sequence with O backbone) is the same (Fig. 1), and the curves are superimposed, supporting the hypothesis that 2084 and 1916 bind to the same site.
Fig. 1

Mouse B cells were cultured with a series of concentrations of the prototype IN-ODN 2114 in the presence of either ST-ODN 2084 (100 nM, S backbone) or 1916 (10 μM, O backbone with same sequence). Percent inhibition relative to 2114 is calculated from the concentration required for 50% inhibition of IL6 secretion, with 0% being the activity of ST-ODN alone, and 100% being maximal inhibition by the prototype IN-ODN 2114

A limited set of comparisons has so far shown only minor differences in sequence preferences between O and S backbone ODN in mouse cells [36]. Changes in the critical bases diminished activity whether the backbone was O or S (Fig. 2). However, consistently, changes in critical bases impaired activity of ST-ODN much less and IN-ODN much more if the backbone is O than if it is S (Fig. 2). A possible explanation for this behavior derives from the old observation that at concentrations of about 10 μM or higher the backbone can exert effects independent of sequence which are either stimulatory (when used alone) or inhibitory (when present with another ODN), involving a complex mix of causes including interference with uptake [37, 38]. Thus the ratio of sequence-specific to nonspecific backbone effects would be less with O- than with S-stimulators (Fig. 2A), whereas with weak O-backbone inhibitors, sequence specific effects might be below the detection threshold (Fig. 2B).
Fig. 2

Influence of phosphodiester (o) and phosphorothioate backbones on ST-ODN and IN-ODN activity in mouse B cells. The sequences listed were tested with both O and S backbones. (A) Stimulatory activity was measured by G1 entry and apoptosis protection, using fluorescence with acridine orange [49]. (B) Inhibition was measured as in Table 1, using 10 μM ODN 1916 or 100 nM 2084 to provide an equal stimulus (Fig. 1) for comparing S and O backbone inhibitors

We would confidently have predicted that more rapid degradation would cause truncated IN-ODN with O backbones to show much greater loss of activity than S-backbone ODN with the same sequence and length. For example, the loss of the 5′ T from S-backbone IN-ODN may even enhance activity (Fig. 2), but the loss of only one more 5′ base causes a 2-log loss of activity [24]. Thus, we were surprised to find that IN-ODNs with O-backbones sustained activity after truncation at least as well, and in some cases better, than their S-backbone counterparts (Table 2). These results suggest a potent protective mechanism preserving the ends of O-backbone ODN. Nevertheless, the same IN-ODN sequence has 3 logs less activity with an O-backbone than with an S-backbone, suggesting that it is the internal cleavage (perhaps between the critical bases) that determines the life of an ODN in the cell.

In conclusion, despite the differences in ST-ODN sequence preference between humans and mice, the prototype strong IN-ODN 2114 (S)/4228 (O) and some of its truncation mutants (Table 2) are excellent choices for work with human cells as well as mouse cells.

In vivo effects of Inhibitory ODN

The knowledge that purified mammalian DNA is not stimulatory (actually somewhat inhibitory—33), led to surprise at the results of the Marshak-Rothstein lab showing that immune complexes from lupus patients containing nucleosomes would activate B cells when they could simultaneously engage the B cell receptor and TLR9 [4, 26]. TLR9 was implicated because IN-ODN 2088 [4, 12, 23] blocked the activation of B cells whose BCRs recognized components of these immune complexes [4, 26]. The positive role of TLR9 as a critical sensor of bacterial DNA to warn us of infection was greatly complicated by the concept of endogenous TLR ligands, in this case our own nucleosomes, using TLR9 to initiate an autoimmune response. Certain B cells expressing BCRs directed to chromatin continue to proliferate ex vivo; this proliferation may be driven by chromatin released from apoptotic cells in culture, as it is inhibited by IN-ODN 2088 but not by neutral ODN 2310 [39]. Parenteral ST-ODN accelerated the production of anti-dsDNA and progression of lupus in lupus prone NZB/W mice [40], whereas IN-ODN 2114 [28] and TTAGGG3 [41] have the opposite effect. The prospects of using IN-ODN to treat lupus have been reviewed [5, 42].

Too much exogenous TLR9 ligand can cause harmful inflammation, raising expectations that IN-ODN could act in vivo to prevent these harmful effects. In a model of CpG-induced lung inflammation, IN-ODN blocked the development of pneumonitis [43]. Like LPS, two doses of ST-ODN at the right interval can induce a lethal cytokine storm in mice resembling septic shock, and inhibitory ODN (in this case TTAGGG3) can prevent it [44]. Injection of ST-ODN into joints causes inflammation [45], preventable by simultaneous intraarticular [46] or systemic [47] injection of IN-ODN.

Summary

The implication of nucleosome-Ig complexes as activators of TLR9 in lupus, and the potential toxicity associated with use of high doses of stimulatory ODN as therapy for cancer or allergy, or as vaccine adjuvants, create a need for antidotes for CpG-DNA activation. Synthetic inhibitory ODN with the phosphorothioate backbone represent a potential solution for this problem.

Acknowledgments

Expert secretarial assistance by Deanna Ollendick, technical assistance by Adam Goeken, and collaboration with Eicke Latz and Douglas Golenbock are gratefully acknowledged. Supported by RO1 AI 47374–04 from the NIH.

Copyright information

© Humana Press Inc. 2007

Authors and Affiliations

  1. 1.Graduate Program in Immunology, Carver College of MedicineUniversity of IowaIowa CityUSA
  2. 2.Department of Internal Medicine, Carver College of MedicineUniversity of IowaIowa CityUSA

Personalised recommendations