Peptides identify multiple hotspots within the ligand binding domain of the TNF receptor 2

Background Hotspots are defined as the minimal functional domains involved in protein:protein interactions and sufficient to induce a biological response. Results Here we describe the use of complex and high diversity phage display libraries to isolate peptides (called Hotspot Ligands or HSPLs) which sub-divide the ligand binding domain of the tumor necrosis factor receptor 2 (TNFR2; p75) into multiple hotspots. We have shown that these libraries could generate HSPLs which not only subdivide hotspots on protein and non-protein targets but act as agonists or antagonists. Using this approach, we generated peptides which were specific for human TNFR2, could be competed by the natural ligands, TNFα and TNFβ and induced an unexpected biological response in a TNFR2-specific manner. Conclusions To our knowledge, this is the first report describing the dissection of the TNFR2 into biologically active hotspots with the concomitant identification of a novel and unexpected biological activity.


Background
Hotspots have been defined as the minimal functional protein:protein interaction domains through which biological activity can be modulated [1][2][3]. A novel strategy called Phenogenix ® has been developed as a means for studying these interactions. To enable Phenogenix ® , we use large and diverse phage display libraries consisting of randomized 20 mer and 40 mer amino acid peptides with > 10 11 independent clones. Because of their ability to modulate protein: protein interactions, the resulting peptides are called Hotspot Ligands (HSPLs). Using this approach, we have successfully identified peptide agonists and antagonists for a number of biologically important molecules including growth hormone receptor, insulin receptor and the insulin-like growth factor receptor [2,3].
In this report, we describe the use of Phenogenix ® to identify the critical protein:protein interactions underlying the TNF/TNFR axis. TNFα and TNFβ have been extensively studied and are involved in immune and pro-inflammatory responses, playing an important role in host defenses against infection and other disease states [4][5][6][7]. The biological effects of TNFα and TNFβ are mediated through the two membrane associated receptors, TNFR1 (p55) and TNFR2 (p75), that are expressed on the target cells [8,9]. The postulated pathogenic roles for TNF include sepsis and other bacterial and viral pathologies [10][11][12], certain cancers [13], metastasis [14] and chronic autoimmune disorders such as rheumatoid arthritis [15,16], multiple sclerosis [17] and Crohn's disease [18]. It is worth noting that TNFα mutants have been identified that selectively bind TNFR1. One such mutant induced less systemic toxicity yet showed no compromise in its anti-tumor activity in nude mice [13]. These observations indicated that finding modulators specific for each of the TNF receptors is indeed possible.
While TNFα binds equally well to both receptors, the majority of biologic responses attributed to TNF are mediated via TNFR1 although there is evidence that activation of TNFR2 is responsible for some adverse effects of TNFα [19][20][21][22][23][24]. Thus, the differential regulation of these two receptors may require drugs that specifically target either TNFR1 or TNFR2 allowing therapeutic efficacy without the expected toxicity profile. One obvious possibility then is to search for Surrogates that selectively bind and regulate the individual receptors. To date several antagonistic protein reagents have been isolated for treating diseases with an underlying TNF etiology. These include neutralizing monoclonal antibodies and soluble forms of TNFR1 and TNFR2 [25][26][27]. However, both antibodies and extracellular receptors are large proteins and not amenable to oral administration. They also carry the unwanted risk of stimulating an auto-immune response during chronic use.
By panning our peptide libraries against the TNFR2, we isolated four HSPL peptides binding specifically to human TNFR2 that allow the subdivision of the receptor into a minimum of at least three independent hotspots involved in its interaction with TNFα. In a biological model for TNFα activation of TNFR2, three of the four peptides induced cytotoxic response in the target cell line suggesting that at least one of these domains induced a unique TNFR2-specific biological response. Overall, these data show the utility of using the HSPL peptides for subdividing protein:protein interacting domains and for defining unique novel biological activities not observed through the natural ligand/receptor interaction.

Results
Human TNFR2 was panned with random 20 mer and 40 mer phage display peptide libraries of high diversity [2,3,28]. After four rounds of panning, four phage clones were isolated with specificity for TNFR2. The peptides were initially divided into two groups: a 40 mer designated KcF12 and three 20 mers designated KcC7, KcD11 and KcF6 (Table 1). From the amino acid sequences, KcC7, KcD11 and KcF12 were found to have a pair of cysteine residues suggesting the possibility of intra-chain disulfide bonding. In addition, a putative motif, WxEYxxRGG, was found for peptides KcF6 and KcD11 (Table 1). It was interesting that this motif was related to a sequence Wx-ExxxxxGG found at the C-terminal (amino acids 190-198) of TNF-α. No appreciable sequence homology was seen between either of the natural ligands, TNFα or TNFβ, and KcF12 although the spacing of cysteine loop was close to that seen in the cysteine rich domains of TNFR2 [29]. At present, it is not known whether these structural elements play a role in binding of the peptides to TNFR2 and it may be that the peptides mimic rather than recapitulate the contact domains of the natural ligands. ELISA analysis showed that all four phage clones bound specifically to human TNFR2 and did not react with either human or murine TNFR1 or murine TNFR2 (data not shown).
Competition ELISAs were performed using TNFα and TNFβ vs. the four phage clones at their ED 50 values ( Figure  1a,1b,1c,1d). More TNFα and TNFβ were required to compete KcF12 than the other phage clones suggesting that certain intrinsic properties of KcF12 (such as its length and/or amino acid content) may contribute to a higher relative affinity ( Table 2). In addition, the synthetic forms of KcC7, KcD11, KcF6 and the recombinant KcF12 (rKcF12) were able to compete the binding of TNFα and TNFβ, as well as KcF6 and KcF12 phage clones in the micromolar range (unpublished data). Interestingly, the three 20 mer HSPLs but not KcF12 were competed by an anti-TNFR2 monoclonal antibody, mAb226, ( Figure 2) suggesting that KCF12 binds to a hotspot on TNFR2 distinct from the other 3 peptides. A summary of the phage/peptide competition data is shown in Table 3.

KcC7
GLQRWCRWAESACGDSGPLR The sequences shown are the deduced amino acid sequences of the human TNFR2-specific phage clones that were selective for TNFR2, but not TNFR1. The cysteine residues are underlined and highlighted in grey. A potential motif, WxEYxxRGG, found in peptides KcF6 and KcD11 are indicated as bolded one letter amino acid codes.
Binding of KcF12 to native TNFR2 was measured in Jurkat and Jiyoye cells using a radioreceptor assay. Jurkat cells have been reported to have approximately equal numbers of TNFR1 and TNFR2 per cell [27] whereas Jiyoye cells express mainly TNFR2 (unpublished data). KcF12 inhibits TNFα binding to Jiyoye cells with a Ki of about 2 uM whereas the Ki for the peptide in Jurkat cells was found to be >30 uM ( Figure 3). Since TNFα binding to Jurkat prob-ably involves both TNFR1 and TNFR2, the data offers additional confirmation of KcF12's selectivity for TNFR2.
It has been previously reported that TNFα induced a TNFR2-specific proliferation of the neuroblastoma cell line SK-N-BE [30]. KcF12, KcC7 and KcF6 inhibited the proliferative response to TNF-α in this cell line indicating that they could act as antagonists (data not shown). In addition, the same three peptides were found to be cytotoxic to SK-N-BE cells in the absence of TNFα (Figure 4a). KcF12 possessed the strongest killing activity with an IC 50 of ~700 nM whereas the IC 50 s for KcC7 and KcF6 were 2 uM and 7 uM, respectively. None of the peptides induced a cytotoxic response in cell lines not expressing TNFR2 (data not shown). Interestingly, KcD11, which shares a motif with KcF6 (Table 1), had no effect on SK-N-BE cells. The absence of activity by KcD11 did not appear to be due to breakdown of the peptide during culturing since mass spectral analyses on aliquots retrieved from culture media showed relatively no degradation of peptides between the initial and the final time points (data not shown). Therefore, the proliferation data suggests evidence of a hotspot on TNFR2 recognized by KcF12, KcC7 and KcF6 that is linked to this unique biological effect. The inhibition of the KcC7 and KcF6 cytotoxic effect with mAb226 ( Figure  4b) confirms the relationship between the peptide-induced cytotoxic event and TNFR2 because of the antibody's specificity for TNFR2. Overall, these data indicate HotSpot Peptide Ligand phage clones were added to 96 well plates with immobilized TNFR2 at a concentration which allowed 50% occupancy of receptor binding sites. Anti-TNFR2 antibody, mAb226, was added and serially diluted for competition. To visualize receptor bound HSPL-phage the reaction was probed with HRP labeled anti-M13 antibody. To demonstrate mAb226 interacted with the TNF binding pocket, TNF-alpha was added in a manner as described for the phage clones and TNF binding was determined using an anti-TNF antibody followed by HRP-anti-mouse IgG antibody. Data is presented as percent inhibition and were fit to a four-parameter logistic equation using Xlfit™, an add-in for Microsoft Excel.
that the surface topology of the TNFR2 binding site is very complex and may be composed of multiple hotspots capable of different regulatory functions.

Discussion
Peptide libraries have been successfully used for identifying novel peptides mimicing natural molecules in terms of the critical protein:protein interactions involved in biological activity. Yanofsky et al. [31] described a peptide antagonist to IL-1 that was active in the nM range and Wrighton et al. and Livnah et al. [32,33] identified peptides binding to the erythropoietin (EPO) receptor with full agonist activity in vivo. Neither the EPO agonist nor the IL-1 antagonist peptides show any significant sequence homology to the natural ligand suggesting that they act as mimics of the natural ligand. In the case of the EPO agonists, X-ray crystallographic analysis revealed that one of these peptides spontaneously formed non-covalent homodimers that allowed the dimerization of two EPO receptors. Two families of small peptides binding to the human thrombopoietin receptor (TPOR) were isolated that competed with the natural ligand TPO [34]. The peptide with the highest affinity was subsequently dimerized and shown to function as a potent TPO agonist in vivo. Recently, a small, non-peptide organic molecule binding to one chain of the EPO receptor was multimerized and shown to mimic the biological activities of the EPO [35].
In this report we show that HSPLs can be isolated that bind to human TNFR2 at its ligand-binding site and subdivide it into multiple hotspots. All of the peptides were competed by both TNFα and TNFβ implying they recognized a hotspot that mapped either within or at least overlapping the ligand-binding domain of TNFR2. The HSPLs KcC7, KcD11, KcF6 but not KcF12 were also inhibited by the neutralizing antibody mAb226 suggesting the 20 mers but not KcF12 interact with a hotspot overlapping with the one recognized by antibody. The fact that recombinant KcF12 peptide competed all of the phage expressed peptides (Table 3) indicated that the four peptides also recognize a common hotspot which may be the one responsible for TNF ligand binding to the receptor. In addition, the biological data suggests that KcD11 binds to a hotspot that can be further separated from KcF12, KcF6 and KcC7. While we cannot rule out possible allosteric interactions between the peptides, the data suggests that the HSPLs were able to dissect the ligand-binding region of TNFR2 into several structural and/or functional hotspots. In addition, the HSPLs were specific for TNFR2 unlike the native ligands TNF-α and TNF-β which bind to both receptors. These results demonstrate that the approach has the ability to dissect critical protein:protein interaction domains in terms of both structural and functional elements. Using a similar approach, we recently reported on    the subdivision of the ligand binding domain on the insulin receptor into two separate and non-overlapping hotspots with different biological activities [3].
It has been shown that TNFα induced proliferation of the undifferentiated neuroblastoma cell line, SK-N-BE, resulted from activation of TNFR2 [30]. Proliferation of SK-N-BE cells could be suppressed by the pre-incubation of cells with anti-TNFR2 blocking antibodies without concomitant cytotoxicity (data not shown). While all of the peptides blocked the agonist effects of TNFα indicating that they can act as antagonists, KcC7, KcF6 and KcF12 by themselves induced a cytotoxic response by killing SK-N-BE cells in the absence of TNFα. This is surprising because these peptides share a hotspot with the natural ligand TNFα (which is proliferative rather than-cytotoxic) and KcC7 and KcF6 bind to a hotspot common to the neutralizing but non-cytotoxic, antibody mAb226. Overall, these data suggest that activation (agonism or cytotoxicity) of TNFR2 is mediated through unique hotspots on the receptor that can be defined by the peptides. Under the present experimental conditions, it is unclear whether the natural ligand, TNFα, touches the hotspot responsible for cytotoxicity although one can speculate that, if it does, the cytotoxic effect might be nullified by the sum of the natural ligand:receptor interactions. Competition experiments showing that mAb 226 reversed the effects of KcC7 and KcF6 confirm that these two peptides induce their effects though their interaction with TNFR2. These findings suggest that TNFR2 has multiple biological functions. In addition to transferring TNFα to TNFR1 as proposed by the ligand passing model [36], TNFR2 may have a role in modulating proliferative responses in both positive and negative manner. Ongoing studies will attempt to elucidate the mechanism of action behind the phenomenon such as the effects on downstream signaling pathways including NFkB, and apoptosis. However, it is possible that peptide-induced cytotoxicity works through a mechanism similar to TNF mediated apoptosis in T-cells [37].

Conclusions
In summary, HSPLs were isolated with which we were able to dissect the ligand-binding domain of a natural receptor and generate information about novel and unexpected bioactivities. This technology has also been successfully used for known protein targets, such as the insulin receptor [3] and TNF-β [38] as well as non-protein targets such as disease-specific RNAs [38]. We have also shown that the HSPL libraries identify hotspots on novel genes of unknown function and identify functional interaction pathways (manuscript in preparation). Overall, the use of HSPLs have the potential to become a general approach for scanning and understanding complex protein:protein interactions.

Design of the Primary Peptide Libraries
A description of library design and production has been previously published [3,28].

Subtractive Panning and ELISA
Panning was done as described [3,28], except that only 6 wells of a 96-well microtiter plate were coated with target at concentrations ranging from 10-100 ng/well. The cross reactive phages from the eluted phage were then subtracted by panning sequentially against human TNFR1, TNFα and TNFβ as a way to obtain peptides specific to the original target. ELISA was done as previous described [3,28].

Phage Competition ELISA Studies
Target was coated on MaxiSorp plates and blocked with NFM-PBS as described [3,28]. HSPLs (phage clones or synthetic peptides) are added in the presence of varying concentrations of competitor (e.g., TNF ligand or soluble receptor). IC 50 s are determined by probing for phage or peptide binding to target in the presence of competitor.