Modulating CD40 and integrin signaling in the proinflammatory nexus using a 15-amino-acid peptide, KGYY₁₅

Abstract

CD40 signaling has long been a target in autoimmunity. Attempts to block signaling between CD40 and CD154 during clinical trials using monoclonal antibodies suffered severe adverse events. Previously, we developed a peptide, KGYY₁₅, that targets CD40 and, in preclinical trials, prevents type 1 diabetes in >90% of cases and reverses new-onset hyperglycemia in 56% of cases. It did so by establishing normal effector T-cell levels rather than ablating the cells and causing immunosuppression. However, the relationship between KGYY₁₅ and other elements of the complex signaling network of CD40 is not clear. Studying interactions between proteins from autoimmune and nonautoimmune mice, we demonstrate interactions between CD40 and integrin CD11a/CD18, which complicates the understanding of the inflammatory nexus and how to prevent autoinflammation. In addition to interacting with CD40, KGYY₁₅ interacts with the integrins CD11a/CD18 and CD11b/CD18. We argue that modulation of CD40-CD154 signaling may be more advantageous than complete inhibition because it may preserve normal immunity to pathogens.

Targeting CD40 in inflammatory diseases, including autoimmunity, has long been a quest. However, the development of new treatments has been fraught with complications. Almost all CD40 antibodies are stimulatory, which is opposite the desired effect for autoimmune disease. CD40 antibodies are pursued in cancer treatments where an inflammatory response targeting the cancer cells is warranted. For autoimmune diseases, focus shifted to CD154 antibodies, which target this natural CD40-ligand to restrict interaction with the CD40 receptor. Success was demonstrated in preclinical autoimmune models where CD154 antibodies prevented disease onset or ameliorated symptoms (1, 2, 3, 4). However, human clinical trials using CD154 antibodies were halted despite evidence of clinical efficacy because in some subjects unexpected severe adverse events, thrombotic emboli, developed (5, 6). Other means of targeting CD40 signaling, including random peptides (7, 8) and targeted organic small molecules (9), were tested. However, those approaches were not efficacious in vivo. Small organic compounds could inhibit CD40–CD154 interaction in vitro but lost activity in protein-rich medium (10). In addition, those compounds are similar to suramin (11), which has toxic side effects (12).

Considering these difficulties, we designed a peptide based on one of the regions of CD154 that was determined from X-ray crystallography and site-directed mutational analysis to interact with CD40 (13). The peptide spans amino acids 136 to 150 of the mouse CD154 protein (UniProtKB—P27548; VLQWAKKGYYTMKSN). Preclinical testing demonstrated that the peptide, KGYY₁₅, prevents disease onset in a mouse model of type 1 diabetes mellitus and reverses overt hyperglycemia in 56% of new-onset diabetic mice (14). KGYY₁₅ was shown, by coimmunoprecipitation followed by Western blot analysis, to interact with CD40 (14). These findings were challenged by studies using surface plasmon resonance (15), ELISA-type binding assays, and in vitro stimulation of cells in the absence/presence of the peptide (16), which indicated weak binding of the peptide to CD40.

Based on reports that CD154 whole protein interacts with several integrins in addition to CD40, we considered that other molecules, integrins, might be at play, creating a more complicated interaction.

Integrins, which regulate cell adherence and trafficking as well as activation, are receptors found on many cell types. They can be activated, from a bent or folded state to an erect or unfolded state, by manganese divalent ions (17). Integrin CD11a is strongly expressed by T cells (18) and CD11b is expressed by neutrophils, macrophages, and B cells, among others (17). Reports show that CD154 binds to the integrins αIIbβ3, α5β1 (CD49e/CD29), and αMβ2 (CD11b/CD18) (19). Integrin α5β1 and CD40 can simultaneously bind to CD154, but it is unclear if α5β1 and CD40 work cooperatively in CD40/CD154 signaling or how α5β1 binds to CD154 (20). It is predicted that the integrin-binding site of human CD154 is located in the trimeric interface of CD154 that forms a cryptic binding site (20).

CD40 is prominent in many autoimmune conditions. While CD40 certainly is present in nonautoimmune conditions, it is strongly expressed in autoimmune conditions and is known to be at the core of disease onset and progression (2, 21, 22, 23, 24, 25, 26). Therefore, we compared autoimmune nonobese diabetic (NOD) to nonautoimmune BALB/c mice and here we demonstrate that CD40 strongly interacts with integrin CD11a, in autoimmune conditions, in addition to CD154. We also show that, under assay conditions with phosphate-buffered saline (PBS) and bovine serum albumin (BSA) in kinetic exclusion assays (KinExAs) in vitro, the CD40-targeting peptide KGYY₁₅ interacts with CD40, contrary to what was previously reported for other in vitro assays (15). The peptide does not interact with CD11a/CD18 or CD11b/CD18 alone in this in vitro assay. However, when CD11b/CD18 is combined with CD40 and then allowed to interact with KGYY₁₅, an interaction stronger than that with CD40 alone is observed. Conversely, when CD11a/CD18 is combined with CD40, the interaction is weakened compared with CD40 alone. KGYY₁₅ peptide coimmunoprecipitates CD11a and CD11b from cells of both nonautoimmune and autoimmune background but with different patterns. Our findings bring to light that KGYY₁₅ may preferentially bind activated cells and may modulate inflammatory cell signaling rather than completely inhibiting or ablating all CD40-positive cells. The findings suggest that such modulation leads to control of autoimmune inflammation while not targeting cells that are not currently partaking in an immune response.

Results

CD40 coimmunoprecipitates integrin CD11a

Reports demonstrate that CD154 binds to not only CD40 but also integrins (19, 20). In fact, CD154 can bind CD40 simultaneously with the integrin α5β1 (20). Therefore, we considered that CD40 rather than being a standalone receptor may itself interact directly with integrins. We performed Western blots on proteins immunoprecipitated by the CD40 antibody 4F11 (27). 4F11 readily coimmunoprecipitated CD11a in Th40 cells from autoimmune NOD mice (Fig. 1A; upper panel). The interaction was also evident in Th40 cells from nonautoimmune BALB/c mice (Fig. 1A; upper panel). Although much lower, the interaction was present in peripheral blood mononuclear cells (PBMCs) as well as conventional CD4 T cells from both NOD and BALB/c. In all cases, a smaller band was present (Fig. 1A; upper panel; a longer exposure reveals bands better [not shown]). Integrins are known to be heavily glycosylated, thus a different glycoform may account for the smaller band (addressed in Fig. 3). Because manganese (Mn²⁺) activates integrins (17), we hypothesized that addition of the ion may increase the amount of immunoprecipitated CD11a. However, Mn²⁺ did not affect the amount of CD11a that was coimmunoprecipitated (Fig. 1, A and B). When assaying for CD11b and CD18 in the 4F11 immunoprecipitated proteins, no bands were visible (data not shown).

Figure 1.

CD40 interacts with CD11a. Sorted cells were Mn²⁺ activated, or not, then immunoprecipitation was performed with anti-CD40 antibody 4F11. A, Western blot for CD11a was performed. B, fold change of band intensities in response to Mn²⁺ was analyzed, and significant differences were calculated by one-tailed t test (for CD11a) or two-tailed t test (for CD40). Significant p-values are shown in the figure. Data are representative of three different experiments.

Figure 3.

Glycosylation accounts for the different size bands of CD11a, CD11b, and CD18.A, NOD cell lysates were either assayed as is (NOD whole cell lysates) or subjected to immunoprecipitation by an OVA-peptide (OVA-pep IP; a control for KGYY₁₅ peptide) or by CD3 antibody (CD3-ab IP; a control for CD40 antibody 4F11) then assayed for CD11a, CD11b, CD18, or CD40 in Western blots. CD3-ab immunoprecipitation eluates were also assayed for CD3. B, NOD PBMC lysates were subjected to deglycosylation by PNGase F (removal of N-linked deglycosylation) or by PNGase F and total O-linked deglycosylation (Total deglyc.), or not (Lysates), then Western blots for CD11a, CD11b, and CD18 were performed. Completely deglycosylated bands are indicated (compl. deglyc.). C, lysates were subjected to deglycosylation by PNGase F alone, PNGase F and total O-linked deglycosylation (Total deglyc.), deglycosylation with four different enzymes that remove O-linked glycosylations (O-linked deglyc.), or the four different enzymes that remove O-linked glycosylations separately (O-glycosidase, neuraminidase, β-galactosidase, or glucosaminidase). Western blots were then done for CD11a. Data are representative of three experiments.

We confirmed that CD40 was present in the same samples. Unlike CD11a, the amount of CD40 from NOD and BALB/c cells increased but was only significant in NOD cells, if the cells were first activated by Mn²⁺ (Fig. 1, A and B; bottom panel). This suggests that CD11a may shield CD40 exposure.

KGYY₁₅ immunoprecipitates CD11a, CD11b, and CD18 in addition to CD40

CD154 is known to interact with CD11b/CD18 (19) and a portion of the amino acids of CD154 that are critical in binding to CD40 also are involved in that interaction (19). KGYY₁₅ contains the amino acids from CD154 known to interact with integrins as well as with CD40; therefore, we determined whether KGYY₁₅ could interact with CD11a and CD11b in addition to CD40. CD11a was immunoprecipitated by KGYY₁₅ from both NOD and BALB/c immune cells and displayed two major bands (Fig. 2A; top panels). CD11b was also readily immunoprecipitated but more prominently from NOD than BALB/c cells (Fig. 2A; second row panels). The patterns differed between the two strains; in autoimmune NOD mice there was prominent CD11a immunoprecipitation from MHCII⁺/CD8⁺ cells (done as depletion in order to get CD4 cells but assayed here as a mix) while it was virtually undetected in those cells from nonautoimmune BALB/c (Fig. 2A; top panels). Conversely, CD11a immunoprecipitation was absent in Th40 cells from NOD mice but present in Th40 cells from BALB/c mice (Fig. 2A; top panels). CD11a was prominently immunoprecipitated from splenic conventional CD4 T cells from both NOD and BALB/c mice (Fig. 2A; top panels). CD11b also displayed very different patterns in the two strains; in BALB/c little to no CD11b was immunoprecipitated from splenic cells while in NOD splenic MHCII/CD8 and conventional CD4 it was readily detected (Fig. 2A; second row panels). In NOD splenic Th40 cells, as was the case for CD11a, CD11b was not detected (Fig. 2A; second row panel). Since Mn²⁺ is known to activate integrins, we speculated that adding this ion would increase KGYY₁₅ interaction with the integrins. Exposure to Mn²⁺ significantly increased KGYY₁₅ interaction with CD11a in splenic MHCII/CD8 and conventional CD4 from NOD (Fig. 2, A and B; p-values displayed in B). In NOD PBMC, only the smaller CD11a band was significantly increased (Fig. 2B). In BALB/c, KGYY₁₅ interaction with the small CD11a band increased significantly in splenic conventional CD4 T cells and in PBMC while the larger band did not change or, in the case of PBMC, decreased significantly (Fig. 2B). Mn²⁺ addition significantly increased KGYY₁₅ interaction with CD11b in NOD splenic MHCII/CD8 cells (Fig. 2B). There was no significant effect of Mn²⁺ on that interaction in the other NOD cells or in BALB/c cells (Fig. 2B).

Figure 2.

KGYY₁₅interacts with CD11a, CD11b, and CD18. Sorted NOD and BALB/c cells were activated with Mn²⁺, or not, then immunoprecipitation was performed with KGYY₁₅-conjugated microbeads. A, Western blots for CD11a (top panels), CD11b (second from top panels), CD18 (third from top panels), and CD40 (bottom panels) were performed. B and C, fold change of band intensities in response to Mn²⁺ was analyzed, and significant differences were calculated by one-tailed t test (for CD11a, CD11b, and CD18; except for BALB/c PBMC large CD11a band where two-tailed test was used) or two-tailed t test (for CD40). Significant p-values are shown in the figure. Data are representative of least three different experiments.

Because cells from autoimmune NOD mice are more activated than the same cells from nonautoimmune BALB/c mice, we purified BALB/c cells and activated them with LPS overnight prior to Mn²⁺ stimulation and KGYY₁₅ immunoprecipitation. Interestingly, LPS activation ablated the KGYY₁₅ interaction with CD11a in splenic Th40 cells, mimicking NOD Th40 cells (Fig. 2A; top panels). The interaction remained absent in LPS-activated BALB/c splenic MHCII/CD8 cells, which is opposite of the case in NOD splenic MHCII/CD8 cells (Fig. 2A; top panels). After LPS activation, both the large and small CD11a band interactions with KGYY₁₅ increased significantly in response to Mn²⁺ in splenic conventional CD4 T cells (Fig. 2B). That was only the case for the small band in non-LPS-exposed BALB/c conventional CD4 T cells (Fig. 2B). The significant decrease in CD11a large band interaction with KGYY₁₅ in response to Mn²⁺ noted in non-LPS-exposed PBMC was lost when the cells were first activated with LPS, which is similar to PBMC from NOD mice (Fig. 2B). KGYY₁₅ interaction with CD11b was increased in the LPS-activated splenic Th40 and PBMC cells and became detectable in splenic conventional CD4 T cells (Fig. 2A). Mn²⁺ significantly increased that interaction in splenic conventional CD4 T cells (Fig. 2B).

KGYY₁₅ interaction with CD18 was detectable mainly in splenic conventional CD4 T cells from both mouse strains (Fig. 2A; third row panels). That interaction was significantly increased in response to Mn²⁺ (Fig. 2C; p-values displayed in figure). When BALB/c cells were exposed to LPS, some CD18 immunoprecipitation became detectable in Mn²⁺-activated splenic Th40 cells (Fig. 2, A and C). CD18 was still detectable in splenic conventional CD4 T cells, but the significant increase that was detected in non-LPS-exposed cells in response to Mn²⁺ was lost (Fig. 2C).

We tested whether there were changes in KGYY₁₅ interaction with CD40 under these conditions. CD40 was most prominently immunoprecipitated in splenic MHCII/CD8 cells from BALB/c and was detectable in splenic conventional CD4 T cells from this strain (Fig. 2A; bottom panels). The splenic conventional CD4 T cells from BALB/c decreased CD40 interaction with KGYY₁₅ significantly in response to Mn²⁺ (Fig. 2C; p-values displayed in figure). In NOD splenic conventional CD4 T cells and Th40 cells, CD40 was detected and decreased significantly in response to Mn²⁺ in the former cell subset (Fig. 2, A and C; p-values displayed in C). When BALB/c cells were activated with LPS, the KGYY₁₅ immunoprecipitated CD40 from splenic MHCII/CD8 cells was decreased and was ablated from conventional CD4 T cells while it became detectable in Th40 cells (Fig. 2A).

CD11a, CD11b, and CD18 each have glycoforms

As controls for the experiments in Figures 1 and 2, we performed Western blots on the lysates prior to immunoprecipitation. CD11a, CD11b, and CD18 were readily detected in whole cell lysates from PBMC and splenic MHCII/CD8, conventional CD4, and Th40 cells (Fig. 3A). Interestingly, larger bands than the 100-kD band detected for CD18 in Figure 2 were detected in the lysates (Fig. 3A; third panel). As a control for the CD40 antibody 4F11 immunoprecipitation in Figure 1, we performed immunoprecipitations using an antibody to CD3 to assess any nonspecific binding. No CD11a, CD11b, or CD18 was immunoprecipitated by the CD3 antibody (Fig. 3A; CD3ab IP). As a confirmation, CD3 content in the CD3 antibody immunoprecipitates was analyzed and CD3 was readily detected (Fig. 3A; bottom panel). To control for the KGYY₁₅ peptide immunoprecipitation, we performed immunoprecipitations using an unrelated peptide, the 17-amino-acid OVA (323–339) peptide, to assess any nonspecific binding. No CD11a, CD11b, or CD18 was immunoprecipitated by the OVA-peptide (Fig. 3A; OVA-pep IP).

We performed deglycosylation experiments to assess whether the different bands in the Western blots were due to glycosylation. CD11a was not deglycosylated by PNGase F, indicating that the bands were not due to N-linked glycosylation (Fig. 3B; top panel). However, when including enzymes to remove O-linked glycosylation the larger band completely disappeared and a faint band at 130 kD was the only band detected; totally deglycosylated CD11a is predicted to be 128 kD (Fig. 3B; Total deglyc.). The fact that the CD11a, 130-kD band was faint indicates that the polyclonal Western blot antibody recognizes polysaccharides on the CD11a protein in addition to protein. For CD11b we detected a band of 160 to 170 kD in Figure 2, and the predicted size of completely deglycosylated CD11b is 127 kD. Using PNGase F alone decreased the size of CD11b to 130 kD (Fig. 3B; second panel). Adding enzymes to remove O-linked glycosylation did not appreciably decrease the size, although the intensity of the band was lessened (Fig. 3B; second panel). In Figure 2, the CD18 band was 100 kD. The predicted size of a completely deglycosylated CD18 is 85 kD. PNGase F treatment to remove N-linked glycosylation led to a smaller band, and treatment with enzymes to also remove O-linked glycosylation led to an even smaller band, about 85 kD (Fig. 3B; third panel). Interestingly, several larger CD18 bands detected in Figure 3A, but not in the immunoprecipitations in Figure 2A, were minimally affected by deglycosylation (Fig. 3B; third panel). Therefore, it is not clear what those larger CD18 bands represent.

KGYY₁₅ binding to cells is partially blocked by recombinant integrin proteins or integrin antibodies as well as by CD40 protein

As mentioned above, CD154 is known to interact with CD11b/CD18 through amino acids that are represented in KGYY₁₅ (19). Therefore, we performed cell staining using a fluorescenated version of the KGYY₁₅ peptide in the absence/presence of competition by recombinant CD11a/CD18 or CD11b/CD18 proteins, or by antibodies to those proteins. When exposed to fluorescenated KGYY₁₅, NAMALWA cells stained positively compared with unstained cells (Fig. 4A). If the peptide was incubated with recombinant CD11a/CD18 or CD11b/CD18 prior to being used for staining, a significant portion of the staining was competed for, a 29.4% reduction with CD11a/CD18 and 22.7% reduction with CD11b/CD18 (Fig. 4A; comp. CD11a/CD18 and comp. CD11b/CD18). If the cells were incubated with antibodies to CD11a or CD11b prior to staining with KGYY₁₅, a significant portion of the staining was blocked, 14.0% with anti-CD11a and 13.2% with anti-CD11b (Fig. 4A; anti-CD11a and anti-CD11b). Preincubation of the peptide with recombinant CD40 prior to using it for staining prevented a significant portion of the staining as well, 11.2% (Fig. 4B; comp. CD40). As a control, preincubation of the peptide with recombinant CD28 protein was done, which did not affect the peptide staining (Fig. 4C; comp. CD28).

Figure 4.

KGYY₁₅binding is partially blocked by antibodies to CD11a and CD11b as well as by recombinant CD11a/CD18, CD11b/CD18, and CD40 proteins, but not recombinant CD28 protein. NAMALWA cells were stained with KGYY₁₅-FITC in the absence/presence of competition. A, competition with anti-CD11a, anti-CD11b, or recombinant CD11a/CD18 or CD11b/CD18 proteins. B, competition with recombinant CD40 protein. C, competition with recombinant CD28 protein. Data are represented as mean ± SEM. p-Values were calculated by two-tailed t test and are shown in the figure.

KGYY₁₅ binds to recombinant CD40 in KinExA

Binding of KGYY₁₅ to CD40 in solid phase methods was not successful (15, 16, 28), albeit buffers different from here and a recombinant CD40 fused with human IgG Fc protein were used. In order to assay the interactions between KGYY₁₅ and CD40 and to determine the K_d, we performed KinExAs (29). This is a solution phase assay and was done in Tris-buffered saline containing bovine serum albumin (Tris/BSA; pH 7.5), to somewhat mimic physiologic conditions. KinExA detected an interaction between KGYY₁₅ and CD40 with a Kd of 109.69 nM (Fig. 5A).

Figure 5.

KGYY₁₅interacts with recombinant CD40 and CD11a/CD18 or CD11b/CD18 proteins in KinExAs. KGYY₁₅ interactions were assayed by KinExA solution phase assay with recombinant CD40, CD11a/CD18, CD11b/CD18, or combinations of CD40 and one or the other integrin. A, KGYY₁₅ with CD40; K_d = 109.69 nM. B, KGYY₁₅ with CD11a/CD18+CD40; K_d = 166.78 nM. C, KGYY₁₅ with CD11b/CD18+CD40; K_d = 7.09 nM.

Recombinant CD11a/CD18 and CD11b/CD18 modulate CD40 interaction with KGYY₁₅ in KinExA

When assayed in KinExA with Tris/BSA and 1 mM MnCl₂ to activate the integrins and using detection antibodies to the CD11 portion of the proteins, KGYY₁₅ interaction with CD11a/CD18 or CD11b/CD18 alone was not evident. It is possible that the recombinant CD11a/CD18 or CD11b/CD18, which are noncovalently associated, became dissociated in the process of binding the KGYY₁₅ peptide and that the peptide binds primarily to CD18. Therefore, we used a detection antibody to CD18 and again did not detect interaction between the integrins and the peptide. We determined whether combining CD11a/CD18 or CD11b/CD18 with CD40 would impact interaction between KGYY₁₅ and CD40. The combination of CD11a/CD18 and CD40 had a K_d of 166.78 nM when interacting with KGYY₁₅ (Fig. 5B). The K_d was 1.5 times higher than CD40 alone (K_d = 109.69 nM), demonstrating that the combination caused the interaction to become less strong. When combining CD11b/CD18 and CD40, the combination had a K_d of 7.09 nM when interacting with KGYY₁₅ (Fig. 5C). That K_d was 15 times lower than the K_d of CD40 alone interaction (K_d = 109.69 nM), demonstrating gain of interaction strength.

Mn²⁺ activation of T cells changes the surface expression phenotype and increases the level of CD40 while decreasing CD18

Integrins are activated by Mn²⁺ (30). Since there is an interaction between integrins and CD40, we speculated that the phenotypic surface availability of CD40 and CD11a/CD11b/CD18 would change when the integrins are activated by Mn²⁺ into their erect state, essentially uncovering CD40. We isolated PBMC from NOD and BALB/c mice and stained for CD4, CD40, CD18, CD11a, and CD11b in the absence/presence of Mn²⁺. Gates were set on CD4hi, CD4lo, and CD4-negative cells to represent conventional CD4 T cells, Th40 cells, and CD4-negative cells (B cells, CD8 T cells, macrophages), respectively (Fig. 6, A and B). Mn²⁺ activation did not increase the percentage of cells expressing CD40 on their surface; however, the intensity of CD40 surface expression was significantly elevated on BALB/c Th40 cells as well as on NOD Th40 and CD4-negative cells (Fig. 6A). The Mn²⁺-activated cells appeared to form new CD40 peaks, but the locations of those peaks were not consistent across several experiments; rather, the overall mean fluorescence intensity increased. When analyzing CD18, there was a decrease in the CD18^hi population and a corresponding increase in the CD18^lo population on BALB/c Th40 cells as well as on NOD Th40 and CD4-negative cells in response to Mn²⁺ activation (Fig. 6B). While there were some intermediate peaks of CD18, those peaks were not consistent across several experiments; rather, the high/low designation was consistent. Conventional CD4 T cells from both strains did not change surface expression of CD40 or CD18 molecules in response to Mn²⁺ activation (Fig. 6, A and B; Conv. CD4). There was no change in CD11a or CD11b surface expression in response to Mn²⁺ on any cell type from either strain (Fig. S1).

Figure 6.

CD4 negative and Th40 cells alter CD40 and CD18 surface expression in response to Mn²⁺. PBMCs were stained for CD4, CD40, CD18, CD11a, and CD11b and analyzed by flow cytometry. In the lymphocyte gate, cells were gated on CD4-negative (CD4^neg), CD4^lo, or CD4^hi, then the expression of CD40 (A) and CD18 (B) was analyzed. Tinted histogram, isotype; solid line, no activation; dotted line, Mn²⁺ activation. p-Values were calculated, in A, by two-tailed t test and, in B, by one-way ANOVA with Tukey's multiple comparisons test.

In addition, examining the same cell populations among spleen cells from the two mouse strains did not reveal any change in surface expression of CD40, CD18, CD11a, or CD11b in response to Mn²⁺ (Fig. S2).

Since Mn²⁺ can be considered a form of activation, we analyzed the surface expression of the classic activation molecules CD69, CD154, and CD25 (while CD25 is associated with regulatory T cells it is also activation induced on effector cells). None of the three molecules was affected by Mn²⁺ activation for 30 min (Fig. S3, A–D).

Discussion

Despite decades of work, the intricacies of CD40 at the nexus of inflammation remain incompletely understood. It is becoming clearer that there are many interactions that can be ascribed to this molecule that depend on the inflammatory state and on the cell type being studied. The quest has long been to inhibit CD40 signals since it is well understood that CD40 is intimately involved in driving autoimmune disease (2, 3, 4, 31, 32, 33, 34, 35).

In this work, we demonstrate that, rather than being a standalone receptor, CD40 interacts with different integrins in a cell, as well as activation, specific manner, thus complicating the understanding of the interaction between CD40 and CD154. Targeting CD40 therapeutically therefore becomes more complicated. There have been clues that this complication could be the case from the observations that CD154 interacts with integrins in addition to CD40 and that it can in some cases do so simultaneously (20). That fact and the present data highlight the point that, when considering CD40–CD154 interaction, a more complicated scenario must be contemplated rather than a simplistic molecule 1 interacts with molecule 2. This new understanding can account for the many, diverse outcomes of signaling that have been reported for this molecular duo.

Integrins are cell surface adhesion molecules that are intimately involved in immune cell trafficking as well as cell–cell interactions; they can be involved in both pro- and anti-inflammatory events (17, 18). CD11a/CD18 and CD11b/CD18 are considered leukocyte-specific integrins; T cells strongly express CD11a/CD18 (18). Integrins exist in a low-affinity, or nonactivated, form that has a folded, or bent, structure (Fig. S4). There is cross talk between integrins and other surface molecules, e.g., the T-cell receptor (TCR) (18), on the same cell and signals generated from those other surface molecules can activate the integrins to their high-affinity, erect state (Fig. S4). Different ions can have an activating or inactivating effect on the integrins. Ca²⁺ is considered to have an inactivating effect while Mn²⁺ is considered an activator, causing the integrin to take on its high-affinity, erect state (Fig. S4). Manganese is a nutritional trace element that is required for various physiological processes, is a cofactor for many enzymes, and is critical for immune responses (36), e.g., host defense against DNA viruses (37). It is known that manganese is present in blood, bone, liver, pancreas, and kidneys, and it has been demonstrated that it is present in the cytoplasm, endoplasmic reticulum, Golgi, and mitochondria of cells (38). It is also understood that manganese shares transport and trafficking mechanisms of other metals (38); however, little is known about the flux of manganese in the microenvironment. This is due to the limitations of the existing methods for measuring the ion (38). The present experiments demonstrate that Mn²⁺ increased the interaction of the KGYY₁₅ peptide with both CD11a Western blot bands in NOD PBMC while this was only the case for the smaller band in BALB/c cells. In fact, the interaction with the larger band in BALB/c PBMC was decreased by Mn²⁺. It is possible that the autoimmune/inflammatory conditions in NOD cells creates a microenvironment, such as a difference in availability of other ions, where activation by Mn²⁺ more readily takes place. In the nonautoimmune/noninflammatory environment, perhaps other ions compete with Mn²⁺. It is also conceivable that the glycosylation status of the integrin, minute saccharide differences not measurable in crude deglycosylation experiments, are responsible for access of the Mn²⁺ ion to the protein, accounting for the difference between NOD and BALB/c.

Our experiments show a direct interaction between CD40 and CD11a that is particularly strong in autoimmune conditions. CD11a typically associates with CD18 (39), but curiously, even though CD11a was immunoprecipitated by the CD40 antibody 4F11 in NOD and BALB/c Th40 cells, those cells did not demonstrate coimmunoprecipitation of CD18. Integrins are composed of an α (e.g., CD11a) and β (e.g., CD18) subunit that are noncovalently linked; thus, alterations in the interaction may happen. It is possible that, under some circumstances, the CD11a/CD18 dimer is disrupted and replaced with a different interaction, e.g., CD11a/CD40. This could constitute a more inflammation-driving receptor while CD11a/CD18 may be utilized more for interaction with ICAM and trafficking. Certainly, when we determined the surface expression phenotype of PBMC with and without Mn²⁺ activation, there was a decrease in surface-expressed CD18 in CD4-negative and Th40 cells while the CD40 intensity increased. We propose that, when Mn²⁺ activates the integrin to its erect state, CD18 can abandon the interaction and either recycle into the cell or be shed from the surface. CD40 intervenes to form an inflammatory receptor with CD11a/b (Fig. S4).

It is possible that CD11a(b)/CD18 and CD40 form a heterotrimer and that CD40 is cloaked by the integrin in its bent state. When we treated cells with Mn²⁺, CD40 surface expression increased without an increase in the percentage of cells expressing it. It is conceivable that the erect integrin uncloaks CD40, which then is available for more intense staining. Cloaking CD40 in an inactive cell state would prevent accidental signaling through this potent inflammation-driving molecule.

The dynamics of CD18 and CD40 were not evident in the CD4^hi, conventional CD4 T cells. While investigators often gate out the CD4^lo cells when analyzing data, we have previously shown that CD4^lo T cells (Th40) contain overall levels of CD3, CD4, and TCR similar to conventional CD4 T cells (40). The difference is that conventional CD4 T cells have those molecules available on the surface of the cells while Th40 cells have high internal levels suggesting an activated state.

Previously, we demonstrated that KGYY₁₅ immunoprecipitates CD40 (14). Our present findings further demonstrate that there is a direct interaction between the two molecules in solution phase assays (KinExA), with a K_d of 109.69 nM. Very interestingly, when KinExA was performed with a combination of CD40 and CD11b/CD18, the K_d of interaction with KGYY₁₅ was improved to 7.09 nM. This while CD11b/CD18 alone did not interact with KGYY₁₅. Conversely, a combination of CD40 and CD11a/CD18 negatively influenced the interaction somewhat, perhaps by altering the interaction site. CD11a/CD18 alone, like CD11b/CD18, did not interact with KGYY₁₅ in KinExA. These data, while performed in vitro, shed light on the possibilities that exist in vivo depending on the microenvironment, cell type, inflammatory status, etc. They also highlight that the ideal intervention in autoimmunity may not be to completely inhibit CD40–CD154 interaction; rather, a modulation of the signaling complex that is involved in the CD40–CD154 inflammatory nexus may be more advantageous therapeutically. Modulation away from inflammatory CD40 interactions rather than complete inhibition may also prevent immune suppression that can be associated with such a strategy (41).

Previous studies were unable to demonstrate an interaction between KGYY₁₅ and CD40 (15, 16). Those studies were done in solid phase assays or in CD154 cell reporter systems and used a chimeric protein of CD40 and the Fc-portion of IgG. The Fc-portion is larger than the CD40 portion, and it is unknown whether that distorts the CD40 protein or otherwise affects potential interactions. In one study, CD154 was overexpressed on embryonic kidney cells (16) that are wholly different from B or T cells and whether those cells express integrins CD11a/b and CD18 is unknown and clearly proves definitive for appropriate binding. Naturally expressed CD40 is uniquely glycosylated depending on inflammatory status, and CD40 can interact with several different proteins (42, 43, 44), including CD11a as shown here. Therefore, CD40 posttranslational modifications in the overexpressing kidney cells may not be representative of cells that are actually involved in the autoimmune processes in vivo, i.e., immune cells. These issues can account for the disappointing results in those studies (15, 16).

KGYY₁₅ immunoprecipitated CD11a and CD11b from both nonautoimmune and autoimmune conditions, but the patterns differed between the two. In these experiments, whatever number of the particular spleen cells that were purified from one spleen were used in the assays. This way, the splenic differences between NOD and BALB/c were taken into account. By cell count, Th40 cells are significantly more numerous in NOD than in BALB/c spleen, about five times as many (Fig. S5). BALB/c mice consistently have significantly greater numbers of MHCII⁺ cells per spleen compared with age-matched NOD mice, about 14 times as many (Fig. S5). The numbers of splenic CD8⁺ and CD4⁺ cells are statistically the same between the two mouse strains (Fig. S5). (PBMC numbers were kept the same for both strains in the experiments.) Interestingly, when BALB/c cells were activated with LPS prior to Mn²⁺ exposure and immunoprecipitation, some of the features present in NOD autoimmune cells were replicated, demonstrating that inflammatory activation impacts how the different proteins interact. In addition, despite immunoprecipitating CD11a/b from several cell populations, KGYY₁₅ was not capable of immunoprecipitating CD18 from all of those cells. This suggests that CD18 is not obligatorily associated with CD11a/b. It is possible that other molecules, CD40 in particular, can, during inflammatory activation, replace CD18 in the interaction (Fig. S4) and therefore when the KGYY₁₅ peptide interacts with CD11a/b, CD18 is not immunoprecipitated.

Some integrins, including CD11a and CD11b, have an I-domain in the headpiece of the protein. This domain contains a metal-binding region, is essential for the interaction with ligands, and is exposed when the integrin is in its activated, erect state (17). We were able to block some of the KGYY₁₅ binding to NAMALWA cells with antibodies to CD11a and CD11b. Often the epitope specificity of antibodies is not known; however, the specificity of the CD11a antibody (M17/4) is known, it binds in the I-domain of CD11a. Therefore, the finding that the M17/4 antibody blocked some of the KGYY₁₅ binding suggests that the KGYY₁₅ peptide binds in that domain. The I-domain is not exposed when the integrin is in its bent, inactive form, and thus we speculate that the peptide binds mostly to activated cells in vivo, making it specific for such cells. This would be advantageous since those are the cells that need to be targeted in autoimmunity.

As mentioned, CD154 interacts with CD11b/CD18 and some amino acids of CD154 that are critical in binding to CD40 (represented in KGYY₁₅) are involved in that interaction as well (19). CD154 homotrimerizes to contact CD40 at multiple domains (45). Therefore, the trimer can interact with more than one protein at a time even through the same domain. One CD154 molecule is capable of interacting with several proteins. For example, CD154 could interact with both CD40 and an integrin concomitantly when a cell is activated (Fig. S4). This may result in an inflammatory signal. We speculate that CD40 interaction with an integrin constitutes an inflammatory receptor for CD154 and that the KGYY₁₅ peptide interferes with signals generated through such a receptor by temporally modulating those interactions, whether CD40 or integrin interaction, or both. In many approaches to treating autoimmunity, immune suppression is a problem. KGYY₁₅ peptide may avoid this problem by simply modulating or altering the signaling complex rather than depleting the cell.

Experimental procedures

Mice

NOD and BALB/c mice from Taconic were housed at the University of Colorado Anschutz Medical Campus, AAALAC-approved facility. All animal experiments were performed under an IACUC-approved protocol (#00529) and adhered to the NIH Public Health Service Policy on Humane Care and Use of Laboratory Animals.

Reagents

Mouse KGYY₁₅ (VLQWAKKGYYTMKSN) was from AmbioPharm Inc. FITC-conjugated mouse KGYY₁₅ was from GenScript.

Anti-CD40 antibody 4F11 (27) was produced in-house.

Cell sorting antibody-conjugated microbeads (CD4, cat# 130-117-043; CD8, cat# 130-117-044: MHCII, cat# 130-052-401) were from Miltenyi Biotec.

For Western blots, anti-mouse CD11a (cat# LS-C331613 and LS-C316877) and anti-CD18 (cat# LS-C312785) were from LSBio Inc, and anti-mouse CD11b (cat# PA5-79532) and secondary anti-rabbit antibody (cat# 65-6120) were from Thermo Fisher Scientific. Anti-CD40 antibodies were from Santa Cruz Biotechnology, Inc and Sino Biological (cat# sc-975 and 101510-T32, respectively).

For blocking of cell staining, anti-CD11a (clone M17/4, cat# 16-0111-82) and anti-CD11b (cloneM1/70, cat# 16-0112-82) were from Thermo Fisher Scientific, recombinant CD11a/CD18 (cat# 3868-AV-050) and CD11b/CD18 (cat# 4047-AM-050) were from R&D Systems, and recombinant CD40 (cat# 10774-H08H) and CD28 (cat# 50103-M02H) were from Sino Biological.

For KinExA, Alexa Fluor 647–labeled Streptavidin (cat#405237), Alexa Fluor 647 anti-human CD11a antibody (cat# 301218), Alexa Fluor 647 anti-human CD11b Antibody (cat#301319), and Alexa Fluor 647 anti-mouse/human CD11b Antibody (cat# 101220) were from Biolegend. Recombinant CD11a/CD18, CD11b/CD18 proteins were as above. Recombinant biotinylated CD40 (10774-H27H-B) was from Sino Biological. Sepharose beads (cat# 17-0569-01) were from GE Healthcare.

For flow cytometry, anti-CD40 antibody (1C10) was produced in-house; anti-CD18 (cat# 553292) was from BD Biosciences; anti-CD11a (cat# 12-0111-82), anti-CD11b (cat# 17-0112-82), anti-CD69 (cat#17-0691-82), anti-CD154 (cat# 12-1541-81), and anti-CD25 (cat# 12-0251-83) were from Thermo Fisher Scientific; and anti-CD4 (cat# 130312) was from BioLegend. Isotype antibodies were from Thermo Fisher Scientific.

For cell purification, Lympholyte-Mammal and Lympholyte-M were from Cedarlane.

For immunoprecipitations, Dynabeads MyOne Tosylactivated (cat# 65501) from Thermo Fisher Scientific were conjugated according to manufacturer’s protocol.

All other reagents were from Sigma-Aldrich.

Cell purification and sorting

Running buffer (PBS containing 0.5% BSA and 2 mM EDTA) was used throughout manipulations unless noted otherwise. PBMCs were purified over Lympholyte-Mammal and were used without further purification. Splenic lymphocytes were purified over Lympholyte-M according to manufacturer protocol. Splenic cells were further purified by sorting using antibody-conjugated magnetic microbeads as described (21). Briefly, 30 μl of CD8 and MHCII microbeads were added per spleen and incubated, rocking at room temperature for 15 min. Cells were sorted on an AutoMACS Pro cell sorter (Miltenyi Biotec). CD8⁺/MHCII⁺ cells were collected, and the remaining cells were further sorted by adding 30 μl of CD4 microbeads per spleen with incubation as above. CD4⁺ cells were collected and considered CD4^hi, conventional CD4 T cells, and the remaining cells were considered CD4^lo, Th40 cells. Th40 cells are surface CD4^lo or negative, but we have demonstrated that they have similar levels of CD4, CD3, and TCR to conventional CD4 T cells, only those molecules are located intracellularly (21, 40).

Immunoprecipitations

For spleen, mouse equivalent numbers of splenic cells (Fig. S1) from one spleen were used rather than the exact same number per sample. That is, whatever total number of sorted cells were achieved per spleen was used for immunoprecipitation. This was to reflect that splenic cell numbers achieved are very different between the two mouse strains. For PBMC, 2 × 10⁶ cells were used per immunoprecipitation. Cells were suspended in Tris-buffered saline containing 0.5% BSA and treated with 2 mM MnCl₂ for 30 min, rocking at room temperature, then cells were pelleted. Cells were lysed in 1 ml lysis buffer (20 mM Tris pH 7.5, 137 mM NaCl, 0.1% Triton X-100, 1 μg/ml each of aprotinin and leupeptin, 0.2 mM PMSF, 0.4 mM sodium orthovanadate) and incubated for 5 min at room temperature, then debris was pelleted. Peptide- or antibody-conjugated magnetic beads were added to the lysate at 40 μg peptide or antibody per immunoprecipitation and incubated for 20 to 40 min, rotating at room temperature. Beads were washed twice with 1 ml lysis buffer each, then protein was eluted with 2 × 20 μl 0.1 M sodium citrate, pH 3.0.

Western blots

Nine microliters of 6× loading dye (0.35 M Tris pH 6.8, 10% SDS, 30% glycerol, 0.6 M DTT, 0.012% Bromophenol Blue) was added to 40 μl immunoprecipitation eluate, then samples were boiled for 2 min. A 15-μl sample was loaded per lane on 10% TGX gels (Bio-Rad Laboratories) and Western blots were run according to Laemmli (46).

Cell staining

NAMALWA cells (ATCC) are Burkitt’s lymphoma B cells that are known to express CD40 as well as integrins CD11a and CD11b. NAMALWA cells, 3 × 10⁴, were incubated with 3 μg (20 pmol) anti-CD11a or anti-CD11b antibody for 30 min at room temperature and were then stained with 30 ng (13.6 pmol) FITC-conjugated KGYY₁₅. Alternatively, 30 ng (13.6 pmol) FITC-conjugated KGYY₁₅ was incubated with 5 μg (25 pmol) recombinant CD11a/CD18, CD11b/CD18, or 2.8 μg (136 pmol) recombinant CD40, or 2.8 μg (68 pmol) recombinant CD28 for 30 min at room temperature, then the mixture was added to 3 ×10⁴ NAMALWA cells.

In other experiments, 1 to 3 × 10⁵ cells were stained with the indicated antibodies or isotype control antibodies.

Stained cells were run on a Cytoflex flow cytometer (Beckman Coulter), and data were analyzed by FlowJo software (BD Biosciences).

Kinetic exclusion assays

KinExA was performed by HTL Biosolutions Inc, using a KinExA 4000 instrument from Sapidyne Instruments and COOH coupled Sepharose beads.

In KinExA, the interaction of unmodified molecules in solution is assayed. In our case, KGYY₁₅ is titrated against a constant background of CD40, CD11a/CD18, and/or CD11b/CD18 and the two binding partners are allowed to reach equilibrium. The samples are then briefly exposed to a solid phase, in our case KGYY₁₅ peptide immobilized on Sepharose beads, and any free CD40, CD11a/CD18, and/or CD11b/CD18 is captured. The captured protein is then labeled using a fluorescently labeled antibody, anti-CD11a or anti-CD11b, or streptavidin in the case of CD40. The contact time with the solid phase is short and is less than the time needed for dissociation of the preformed peptide/CD40 and/or integrin complex in solution. Therefore, competition between the solution complex and the solid phase is “kinetically excluded” and the solution equilibrium is not altered during KinExA experiments. The signals generated from the captured CD40 and/or integrins, which are directly proportional to the concentration of free integrin in the equilibrated samples, are used to generate a binding curve, measured in a series. The KinExA method is a means to obtain true solution phase measurements. The peptide and the integrins in solution are unmodified, and the measurement process does not significantly alter the solution equilibrium. Therefore, affinity and kinetic results may more accurately reflect physiological binding interactions than solid phase methods do.

The assays were performed in Tris-buffered saline, pH 7.5, with 0.5% (5 mg/ml) BSA. In addition, the interactions were performed in the presence of 1 mM Mn²⁺ to activate the integrins.

Data analysis

Data analysis was done using Prism software (GraphPad Software).

Cartoon generation

Cartoon figures were created with BioRender software (BioRender).

Data availability

All representative data are contained within the article.

Supporting information

This article contains supporting information.

Conflict of interest

D. H. W. Jr is the founder and chief scientific officer of Op-T LLC.

Reviewed by members of the JBC Editorial Board. Edited by Peter Cresswell