Human β-defensin-3 structure motifs that are important in CXCR4 antagonism


  • Zhimin Feng,

    1. Department of Biological Sciences, School of Dental Medicine, Case Western Reserve University, Cleveland, OH, USA
    Search for more papers by this author
  • George R. Dubyak,

    1. Department of Physiology and Biophysics, School of Medicine, Case Western Reserve University, Cleveland, OH, USA
    Search for more papers by this author
  • Xun Jia,

    1. Department of Biological Sciences, School of Dental Medicine, Case Western Reserve University, Cleveland, OH, USA
    Search for more papers by this author
  • Jacek T. Lubkowski,

    1. Macromolecular Crystallography Laboratory, Center for Cancer Research, National Cancer Institute, Frederick National Laboratory, Frederick, MD, USA
    Search for more papers by this author
  • Aaron Weinberg

    Corresponding author
    1. Department of Biological Sciences, School of Dental Medicine, Case Western Reserve University, Cleveland, OH, USA
    • Correspondence

      A. Weinberg, School of Dental Medicine, Case Western Reserve University, 10900 Euclid Avenue, Cleveland, OH 44106-4905, USA

      Fax: 216-368-0145

      Tel: 216-368-6729


    Search for more papers by this author


Previously, we reported that human β-defensin (hBD)-3 can both antagonize CXCR4 function on T cells and promote receptor internalization in the absence of activation. In the present study, we explored the important structural elements of hBD-3 that are involved in blocking CXCR4 activation by its natural ligand, stromal-derived factor 1α (SDF-1α; CXCL12). Results from site-directed mutagenesis studies suggest that the ability of hBD-3 to inhibit SDF-1α–CXCR4 interaction, as assayed either by blocking SDF-1 binding to CXCR4 or antagonizing SDF-1-induced Ca2+ mobilization, is correlated with the presence of hBD-3 cysteine residues, specific surface-distributed cationic residues, and the electrostatic properties and availability of both hBD-3 termini. Specifically, hBD-3 activity against CXCR4 is reduced by: (a) replacing all six cysteines; (b) replacing the cationic residues with acidic ones in the N-terminus and C- terminus; (c) removal of the first 10 N-terminal residues; and (d) replacing the surface-exposed basic residues Lys8, Lys32 and Arg36 with neutral ones. The hBD-3–CXCR4 interaction has potentially wide-ranging implications for HIV-related biology, as well as for a host of CXCR4-dependent activities, including hematopoiesis, neurogenesis, angiogenesis, carcinogenesis, and immune cell trafficking. CXCR4 is highly expressed on T cells, monocytes, and epithelial cells. Therefore, understanding the structure–function relationship between hBD-3 and CXCR4 that accounts for the antagonistic interaction between the two molecules may provide new insights into HIV/highly active antiretroviral therapy-related pathology, as well as novel insights into the interaction between innate and adaptive immunity at mucosal sites.


calcium assay buffer


human β-defensin-3


recombinant human β-defensin-3


stromal-derived factor 1α


Human β-defensins (hBDs) are epithelial cell-derived, cationic peptides that are primarily recognized for their antimicrobial properties [1]. Recent findings indicate that hBD-3, the β-defensin exclusively expressed in the highly proliferating, nondifferentiated stratum basale of the oral mucosa [2], mediates important immunoregulatory functions, in addition to its antimicrobial properties. These include, but are not limited to, chemotaxis of myeloid cells via CCR2 [2-4], interaction with Toll-like receptor-1 and Toll-like receptor-2, which induces maturation of myeloid dendritic cells and monocytes [5], and nonactivating internalization of CXCR4 (the coreceptor of HIV-1), which leads to protection from HIV-1 X4 tropic viral infections [6, 7].

Previously, we reported that hBD-3 can: (a) rapidly antagonize activation of CXCR4 by its natural ligand, stromal-derived factor 1α (SDF-1α); and (b) promote CXCR4 internalization without activating the receptor [6]. This suggests that the hBD-3–CXCR4 interaction has potentially wide-ranging implications for HIV-related biology as well as for a host of CXCR4-dependent activities, including hematopoiesis, neurogenesis, angiogenesis, carcinogenesis, and immune cell trafficking [8-11]. CXCR4 is highly expressed on T cells, monocytes, and epithelial cells. Therefore, understanding the structure–function relationship between hBD-3 and CXCR4 that accounts for the dynamic interaction between the two molecules may provide new insights into HIV-related pathology, its modification by highly active antiretroviral therapy, and the interplay between innate and adaptive immunity at mucosal sites.

In the present study, we explored the structural elements of hBD-3 that are involved in blocking the productive interaction between CXCR4 and SDF-1α. Our findings suggest that the presence of cysteines and surface-exposed cationic residues is important for this antagonistic activity. In addition, the cationic/uncharged C-terminus and N-terminus also contribute to the CXCR4-blocking activity of hBD-3. Finally, and probably most importantly, truncating the N-terminus by removing the first 10 residues substantially reduces hBD-3 activity.


Generation of hBD-3 mutant proteins for analysis of antagonism of SDF-1α–CXCR4 signaling

Standard methods were used to generate recombinant hBD-3 (rhBD-3) and a series of 26 hBD-3 mutants with substitutions, deletions or insertions of specific residues or domains. The sequence of wild-type hBD-3 is shown in Fig. 1 as a template for the comparison with the mutant proteins, which included: (a) six constructs with single or multiple cysteine substitutions (C11S, C18S, C23G, C33I, C40G/C41Y, and C11S/C18S/C23G/C31I/C40G/C41Y); (b) five constructs with single or multiple substitutions in surface-exposed cationic residues (K8A, K32S, R36I, K8A/K32S, and K8A/K32S/R36I); (c) three constructs with C-terminal insertions of α-helical domains from the C-terminus of SDF-1α (EYLEKALN insert, EYLE insert, and EKGAYNEL insert); (d) six constructs with C-terminal substitutions or insertions of cationic residues (R42S/R43S, K44G/K45G, R42A/R43A/K44G/K45G, RRRK insert, R42E/R43E/K44S/K45S, and N4R/T5R/R42E/R43E/K44S/K45S; (e) three constructs with N-terminal substitutions or insertions (GIGDPVT substitution for wild-type residues 1–10, N4E/T5E/K8A, and N4R/T5R); and (f) three constructs with N-terminal and/or C-terminal deletions (Δ1–5, Δ1–10, and Δ1–5/Δ42–45). These hBD-3 mutants were compared for their relative abilities to target SDF-1α–CXCR4 signaling at the level of: (a) inhibition of direct SDF-1α binding to CXCR4-expressing CEM T cells; and (b) attenuation of SDF-1α-induced Ca2+ mobilization in CEM cells.

Figure 1.

Sequence of wild-type hBD-3. Residues are numbered 1–45 from the N-terminus, and respective cysteines are identified with roman numbers (I–VI).

Inhibition of SDF-1α binding by hBD-3

We previously reported that hBD-3 is a nonclassic antagonist of CXCR4; that is, hBD-3 can inhibit SDF-1α binding to, and activation of, CXCR4, but can also cause CXCR4 internalization without activating G-protein signaling responses. These results suggest that the binding of hBD-3 to CXCR4 induces conformational changes in that receptor that: (a) occlude the binding sites for SDF-1α; (b) suppress productive interactions with downstream G-proteins; and (c) facilitate productive interaction with downstream proteins (e.g. arrestins) involved in receptor internalization. We have now further studied the mechanism of hBD-3–CXCR4 interaction by characterizing the saturable binding of SDF-1α to CXCR4-expressing CEM cells in the absence or presence of hBD-3. The SDF-1α-binding assay was performed with 0.1 nm [125I]SDF-1α plus increasing concentrations of unlabeled SDF-1α in the absence or presence of 5 μg·mL−1 hBD-3. The amount of bound SDF-1α was calculated from the different specific activities (based on the ratios of [125I]SDF-1α concentration to unlabeled SDF-1α concentration). The amount of SDF-1α specifically bound to CEM cells progressively increased as the extracellular SDF-1α concentration was raised from 0.1 to 5 nm, and reached a plateau value at > 10 nm SDF-1α (Fig. 2). In the presence of 5 μg·mL−1 hBD-3, SDF-1α binding was decreased by approximately two-fold (or more) over the entire range of tested SDF-1α concentrations, indicating that hBD-3 acts to decrease the maximum number of SDF-1-binding sites. The SDF-1-binding assay illustrated in Fig. 2 was performed at 37 °C, but similar SDF-1α binding and suppressive effects of hBD-3 were observed at 4 °C (data not shown).

Figure 2.

Effects of hBD-3 on saturatable SDF-1α binding to CEM cells. The SDF-1α-binding assay was performed as described in 'Experimental procedures'. [125I]SDF-1α (10 nm) was added to each tube, together with 0–30 nm unlabeled SDF-1α in the absence or presence of 5 μg·mL−1 hBD-3. Cells were incubated at 37 °C for 20 min prior to termination of binding by rapid sedimentation and washing.

Role of disulfide bonds

We started to identify which regions of hBD-3 contribute to the molecule's ability to inhibit CXCR4 function by systematically replacing each cysteine with an uncharged residue. The rationale was based on the fact that hBD-3 has six conserved cysteines (Fig. 1) that form three intramolecular disulfide bonds that stabilize the molecule's distinctive β-sheet conformation [12]. Figure 3 shows that wild-type hBD-3 (20 μg·mL−1) completely blocked Ca2+ mobilization triggered by 10 nm SDF-1α. Replacing one or two cysteines in hBD-3 reduced, but did not totally abolish, this activity (Fig. 3B–F). However, replacing all six cysteines with uncharged residues was able to completely reverse the blocking effect of hBD-3 on SDF-1α-triggered Ca2+ mobilization (Fig. 3G). Similarly, the ability of this linearized cysteine-free mutant of hBD-3 to block high-affinity binding of SDF-1α to CXCR4 was markedly reduced as compared with the wild-type molecule or a saturating concentration (10 nm) of unlabeled SDF-1α (Fig. 3H).

Figure 3.

Effect of cysteine mutations on the activity of hBD-3: the ability of hBD-3 and its mutants to inhibit SDF-1α-dependent Ca2+ mobilization. (A) Wild type. (B) CM1 (C11S). (C) CM2 (C18S). (D) CM3 (C23G). (E) CM4 (C33I). (F) CM56 (C40G/C41Y). (G) CM1-6 (C11S/C18S/C23G/C31I/C40G/C41Y). (H) Inhibition of SDF-1α binding by CM1-6. For Ca2+ mobilization assays (A–G), 1 (black) or 2 (color) indicate control and experimental (pretreated with wild-type hBD-3 or mutants) Fura-2 fluorescence traces, respectively; arrows beneath the fluorescence traces indicate the time when respective reagents were added (black, control; color, experimental). Similar labeling is used for the Fura-2 fluorescence traces in all subsequent figures.

Effects of surface cationic residue mutations

Previous studies have indicated that electrostatic interactions between SDF-1α and CXCR4 play an important role in CXCR4 activation by SDF-1α [13, 14]. On the basis of the coordinates of SDF-1α (ID: 1SDF) and hBD-3 (ID: 1KJ6) available from the Protein Data Bank (, we modeled the structures of both proteins with the program ds viewerpro 6.0 (Accelrys, Figure 4A shows the SDF-1α (left) and the hBD-3 (right) structures, respectively. Three N-terminal surface cationic residues (Lys1, Arg8, and Arg12) are believed to be important for SDF-1α binding to, and activation of, CXCR4 [13]. Three similar cationic residues are present on the surface of hBD-3: Lys8, Lys32, and Arg36. We generated several mutants with changes to these three residues, and tested their function. As shown in Fig. 4B–E, single or double substitution(s) did not diminish the mutants' ability to block SDF-1α-triggered Ca2+ mobilization. However, replacement of all three cationic residues with neutral ones resulted in substantial reversal of this inhibitory hBD-3 function, i.e. ~ 50% (Fig. 4F). In contrast, this mutant effectively reduced high-affinity SDF-1α binding (Fig. 4G).

Figure 4.

Effect of mutation of surface-exposed cationic residues. (A) Comparison of surface-exposed cationic residues of SDF-1α (left) and hBD-3 (right). (B–F) The inhibition of SDF-1α-triggered Ca2+ mobilization by hBD-3 mutants. (B) K8A. (C) K32S. (D) R36I. (E) 8-32 (K8A/K32S). (F) 8-32-36 (K8A/K32S/R36I). (G) Inhibition of high-affinity SDF-1α binding by 8-32-36.

Effects of inserting the SDF-1α C-terminal α-helix into the C-terminus of hBD-3

Further comparison of the structures of hBD-3 and SDF-1α indicated that hBD-3 lacks the α-helix of SDF-1α (Fig. 5A). Cai et al. found that the C-terminal α-helix is important for SDF-1α-triggered CXCR4 signaling and chemotaxis, but is less critical for CXCR4 internalization [15]. Therefore, we inserted the C-terminal α-helix-forming residues of SDF-1α (EYLEKALN) into the C-terminus of hBD-3, and found that this mutant not only failed to activate CXCR4 (Ca2+ mobilization) independently, but was also unable to block SDF-1α-triggered Ca2+ mobilization (Fig. 5B). This suggests that the α-helix of SDF-1α, positioned in the C-terminus of hBD-3, abrogated the ability of hBD3 to antagonize CXCR4. Binding assay experiments indicated that inserting the helical structure only slightly diminished the molecule's ability to inhibit SDF-1α binding (Fig. 5E). To further investigate the role of the C-terminal α-helix insertion in reducing hBD-3 activity, we generated two additional mutants. In one, the C-terminus of hBD-3 was extended by four residues, EYLE (Fig. 5C), and in the second it was extended by eight residues, EKGAYNEL (Fig. 5D). Neither of these latter C-terminal extensions was predicted to have a helical structure, as determined with porter ( Both mutants retained the ability to block both SDF-1α-triggered Ca2+ mobilization and high-affinity SDF-1α binding (Fig. 5C–E). These data indicate that the insertion of an α-helix into, and not mere extension of, the C-terminus of hBD3 reduces its ability to block CXCR4 functions, suggesting that the C-terminus may be important for optimal hBD-3 interaction with CXCR4. Therefore, we generated more variants of hBD-3 with mutated C-termini.

Figure 5.

Activity changes resulting from the attachment of the α-helix from SDF-1α to the C-terminus of hBD-3. (A) Structural models of SDF-1α (left) and hBD-3 (right). (B–D) Inhibition of the SDF-1α-triggered Ca2+ mobilization by the respective mutants. (B) Helix (insert EYLEKALN at the C-terminus of hBD-3). (C) 4AAs (insert EYLE at the C-terminus of hBD3). (D) 8AAs (insert EKGAYNEL at the C-terminus of hBD3). (E) Inhibition of high-affinity SDF-1α binding by mutants Helix and 8AAs.

Effects of C-terminal mutations in hBD-3

hBD-3 is the most cationic of the four most widely expressed human β-defensins, bearing a net charge of +11 [12]. At the C-terminus, there are four consecutive cationic residues [RRKK)(42–45)] (Fig. 1). To test their involvement in hBD-3-induced CXCR4 antagonism, we generated several mutants by substituting uncharged or anionic residues for the positively charged ones. Replacing RR with SS (Fig. 6A), KK with GG (Fig. 6B) or all four with AAGG (Fig. 6C) did not reduce the ability of hBD-3 to inhibit SDF-1α-induced Ca2+ mobilization. Insertion of four additional cationic residues (RRRK) into the wild-type C-terminus did not abrogate hBD-3 activity (Fig. 6D). However, a modest loss of hBD-3 functional antagonism of CXCR4 was observed when the native tetrapeptide (RRKK) was replaced with EESS (Fig. 6E), and this was correlated with the reduced ability of this mutant to suppress high-affinity SDF-1α binding (Fig. 6G). We speculated that an overall cationic nature of the hBD-3 molecular surface is important for antagonism of CXCR4, and therefore designed a construct to compensate for the negative charge of the anionic EESS C-terminus by introducing two additional cationic residues at the N-terminus of hBD-3 (mutations N4R and T5R). Notably, this compensated mutant regained the ability to completely block both CXCR4 functional activation of CXCR4 by SDF-1α (Fig. 6F) and high-affinity SDF-1α binding (Fig. 6G). Taken together, these observations confirmed that the introduction of a negative charge at the C-terminus of hBD-3 significantly reduced the ability of this mutant to inhibit SDF-1α binding, suggesting an important role for the positively-charged C-terminus in hBD-3 antagonism of CXCR4.

Figure 6.

Effect of the C-terminal mutations on the activity of hBD-3. (A–F) Inhibition of SDF-1α-induced Ca2+ mobilization by several C-terminal hBD-3 mutants. (A) SSKK (R42S/R43S). (B) RRGG (K44G/K45G). (C) AAGG (R42A/R43A/K44G/K45G). (D) RRKKRRRK (insert RRRK at the C-terminus of hBD-3). (E) CA (R42E/R43E/K44S/K45S). (F) CA-NC (N4R/T5R/R42E/R43E/K44S/K45S). (G) Inhibition of high-affinity SDF-1α binding by several hBD-3 mutants.

Effects of N-terminal mutations in hBD-3

To examine the involvement of the N-terminus of hBD-3 in blocking SDF-1α-dependent Ca2+ mobilization and inhibiting SDF-1α binding, we took advantage of the fact that hBD-2 does not antagonize CXCR4 as does hBD-3 [6], and that the N-termini of hBD-2 and hBD-3 are highly divergent, with that of hBD-3 being cationic, and that of hBD-2 being anionic [12, 16]. We generated a chimera by replacing the first 10 residues of hBD-3 preceding the first cysteine (GIINTLQKYY) with the equivalent N-terminal sequence of hBD-2 (GIGDPVT). Figure 7A shows that this chimera substantially lost the ability to block SDF-1α-dependent Ca2+ mobilization. To further characterize the importance of N-terminal positive charges in hBD-3 interaction with CXCR4, we introduced three positive charge-neutralizing mutations (K8A, N4E, and T5E) into the hBD-3 N-terminus. This mutant (termed NA2) showed reduced activity in blocking Ca2+ mobilization (Fig. 7B). Furthermore, a mutant (NC) with additional cationic residues (N4R and T5R; Fig. 7C) and a mutant (K8A) with a neutral N-terminus (Fig. 7D) retained CXCR4 antagonistic activity. Binding assays also showed that both chimeras H2N and NA2 had reduced ability to inhibit high-affinity SDF-1α binding (Fig. 7E). In contrast, NC and K8A inhibited SDF-1α binding similarly to wild-type hBD-3 (data not shown).

Figure 7.

Effect of the N-terminal mutations on the activity of hBD-3. (A–D) Inhibition of SDF-1α-dependent Ca2+ mobilization by several N-terminal hBD-3 mutants. (A) H2N (Gly1–Tyr10 of hBD-3 replaced with GIGDPVT). (B) NA2 (N4E/T5E/K8A). (C) NC (N4R/T5R). (D) K8A. (E) Inhibition of high-affinity SDF-1α binding by mutants H2N and NA2.

Determining the minimal hBD-3 motif retaining CXCR4 antagonistic activity

We sought to identify the minimal hBD-3 structure that retains CXCR4 antagonistic activity. An N-terminal truncation mutant (6-45) was generated by deleting residues 1–5. This mutant was active in inhibiting both SDF-1α-dependent Ca2+ mobilization (Fig. 8B) and high-affinity SDF-1α binding (Fig. 8E). When this N-terminal truncation was combined with the truncation of the C-terminal RRKK cationic residues, the resulting mutant (termed 6-41) was less efficacious than the 6-45 mutant in suppressing SDF-1α-induced Ca2+ mobilization (Fig. 8D) but retained the ability to inhibit high-affinity SDF-1α binding (Fig. 8E). Notably, further truncation of the N-terminus by deletion of the first 10 residues generated a mutant (termed 11-45) with a markedly reduced ability to antagonize both SDF-1α-induced Ca2+ mobilization (Fig. 8C) and high-affinity SDF-1α binding (Fig. 8E). We also generated four mutants with even more extensive truncations (6-39, 6-37, 6-38, and 6-36), and found that none was active in blocking CXCR4 (data not shown).

Figure 8.

Determining the minimal hBD-3 structure needed to block CXCR4. Inhibition of SDF-1α-induced Ca2+ mobilization by three hBD-3 mutants, negative control (A), 6-45 (Gly1–Thr5 deleted) (B), 11-45 (Gly1–Tyr10 deleted) (C), and 6-41 (Gly1–Thr5 and Arg42–Lys45 deleted) (D), with various truncations of the N-terminal region. (E) Inhibition of high-affinity SDF-1α binding by these mutants.

The 6-45 mutant is as potent as wild-type hBD-3 in blocking CXCR4

The truncation studies revealed that, at 20 μg·mL−1, the efficacy of the 6-45 mutant as an antagonist of CXCR4 function was equivalent to that of wild-type hBD-3 (Fig. 8B,E). We further compared the potency of the 6-45 mutant (Fig. 9D–G) and that of wild-type hBD-3 (Fig. 9A–C,G) as an inhibitor of both SDF-1α-triggered Ca2+ mobilization and high-affinity SDF-1α binding. Inhibition of the Ca2+ mobilization response by hBD-3 began to diminish at 10 μg·mL−1 (1.94 μm), whereas the 6-45 mutant retained maximal antagonistic action at 7.5 μg·mL−1 (1.61 μm), but showed reduced activity at 3.75 μg·mL−1 (0.80 μm). The binding assay confirmed that the 6-45 muntant and wild-type hBD-3 are equally potent in suppressing high-affinity SDF-1α binding to CXCR4 (Fig. 9G).

Figure 9.

An N-terminal deletion mutant (6-45) is more active than wild-type hBD-3 in blocking CXCR4. Dose-dependent inhibition of SDF-1α-induced Ca2+ mobilization by wild-type hBD-3 and mutant 6-45. (A) 20μg/ml w/t hBD-3, (B) 10μg/ml w/t hBD-3, (C) 2.5μg/ml w/t hBD-3, (D) 15μg/ml 6-45, (E) 7.5μg/ml 6-45, (F) 3.75 μg/ml 6-45. The bar graph (G) shows dose-dependent inhibition of high-affinity SDF-1α binding by mutant 6-45.


Three disulfide bonds are central structural features of β-defensins. Through cysteine(s) substitution (Fig. 3), we found that a ‘no-Cys’ hBD-3 lost all activity as a CXCR4 antagonist, suggesting that retention of molecular conformation is needed for hBD-3 interaction with CXCR4. We also observed that replacing one or two cysteines did not change the activity of hBD-3, suggesting that hBD-3 maintains the required conformation needed to block CXCR4, even with fewer cysteine. As all defensins described in this study were generated with recombinant methods in a bacterial system (Escherichia coli), it is possible that they represent mixtures of variants with different topologies of disulfides. As previously reported by Wu et al., such a mixture results from the in vitro refolding of hBD-3 [17]. Further studies are necessary to assign specific CXCR4-blocking activity to particular topological variants of hBD-3. We are in the process of chemically synthesizing hBD-3 and its mutants by using orthogonal protection of cysteines [17], in order to provide more detailed insights into the structure–function characteristics of hBD-3–CXCR4 antagonism.

Because previous reports have indicated that three surface-exposed cationic residues of SDF-1α play a critical role in the molecule's interaction with its cognate receptor CXCR4 [13], we tested whether three similar surface-exposed basic residues, Lys8, Lys32, and Arg36, in hBD-3 were critical for its interaction with the same receptor (Fig. 4A). Interestingly, mutagenesis of all three positive residues, but not of one or two of them, altered the activity of hBD-3 against CXCR4 (Fig. 4), suggesting that the net surface charge of the molecule, rather than any individual cationic residue, is important for this activity. In contrast, the observation that replacement of four basic residues at the C-terminus did not change the antagonistic efficacy (Fig. 6C) suggests that the C-terminal basic residues are not critical for hBD-3 interaction with CXCR4.

Certain mutants, such as the 8-32-36 construct with reduced surface cationic charge (Fig. 4F,G), the SDF-1α helix insert construct (Fig. 5B,E), and the 6-41 truncation mutant (Fig. 8D,E), retained the ability to suppress high-affinity SDF-1α binding (assayed with 0.1 nm SDF-1α) but were unable to markedly attenuate the Ca2+ mobilization response triggered by a saturating (10 nm) concentration of SDF-1α. The differential efficacies of these mutants in the two assays suggest that, at the 20 μg·mL−1 test concentration, these hBD-3 variants effectively compete with 0.1 nm SDF-1α for presumably overlapping binding sites on CXCR4, but are unable to compete for these CXCR4 sites in the presence of 10 nm chemokine. Consistent with this scenario, the 8-32-36 construct was able to completely suppress the Ca2+ mobilization response triggered by 1 nm SDF-1α (data not shown).

Mutations targeting both termini of hBD-3 revealed that anionic residues in these segments substantially reduce the ability of hBD-3 to antagonize CXCR4 (Figs 6 and 7). Data for the mutants truncated at the N-terminus indicate that residues 1–5 are not necessary for interactions between hBD-3 and CXCR4 (Figs 8 and 9). However, removal of 10 N-terminal residues eliminated the activity of hBD-3, indicating a central role of this domain in hBD-3 antagonism (Fig. 8). Moreover, observations that either the insertion of an α-helix into the hBD-3 C-terminus (Fig. 5) or truncation of the last four residues of the C-terminus (Fig. 8) results in a loss of activity suggest that an accessible and unmodified C-terminus is important for interaction of hBD-3 with CXCR4.

Kofuka et al. suggested that SDF-1α binds to CXCR4 according to a two-step model [14]. In the first step, the β-sheet, 50-s loop and N-loop of SDF-1α bind rapidly to exposed extracellular loop regions of CXCR4, and the second, slower step leads to an interaction between the N-terminus of SDF-1α and the transmembrane segments of this heptihelical receptor. A similar two-step model was proposed by Wu et al., who suggested that electrostatic interactions between SDF-1α and the extracellular loop domains of CXCR4 are important for the initial binding [18]. Structural analysis indicates a number of similarities between the SDF-1α and hBD-3 molecules, which include cationic properties, a β-sheet head, and a flexible N-terminus. These similarities hint at possible factors that direct both hBD-3 and SDF-1α molecules to overlapping recognition epitopes on CXCR4, and are consistent with the more specific findings reported in this study. Further study is needed to elucidate whether hBD-3 disrupts the two-step SDF-1α–CXCR4 interaction.

CXCR4 is involved in many physiological processes, including hematopoiesis, neurogenesis, angiogenesis, cardiogenesis, immune cell trafficking, and cancer metastasis [8-11, 19]. It is also one of the coreceptors that HIV uses for cell entry [20]. Previously, we have shown that hBD-3 can effectively block CXCR4 function (i.e. its interaction with SDF-1α) by internalization of this receptor without its activation [6]. The structure–function analysis described in this study provides the beginning of a description of determinants of the interaction between hBD3 and CXCR4. Use of this information may aid in development of more potent antagonists of CXCR4, originating from an hBD-3 scaffold, with the goal of utilizing them for anti-HIV treatment or prevention, or in cancer therapy to control processes of angiogenesis and/or metastasis.

Experimental procedures


CEM X4/R5 cells (expressing both CXCR4 and CCR5) were obtained from the NIH AIDS Research and Reference Reagent Program, and maintained in RPMI-1640 supplemented with 5% fetal bovine serum and 400 μg·mL−1 Geneticin (Invitrogen, Carlsbad, CA, USA).

Cloning hBD-3 mutant plasmids

Site-directed mutagenesis was carried out to generate hBD-3 mutants. Primers were designed according to the Stratagene (La Jolla, CA, USA) QuickChange Site-Directed Mutagenesis Instruction Manual. The site find v4.0 program ( was used to design primers to generate mutants. Unless specified otherwise, expression plasmids encoding mutated variants of hBD-3 were created by PCR with predesigned primers and the Expand High Fidelity PCR System (Roche, Indianapolis, IN, USA), with pET-30c-hBD3 (from J. Harder and J. Schröder, Kiel University, Germany) [12] as the template. The PCR program was as follows: 95 °C for 3 min, followed by 20 cycles of 94 °°C for 1 min, 52 °C for 1 min, and 68 °C for 10 min, followed by one cycle of 94 °C for 1 min, 52 °C for 1 min, and 68 °C for 1 h. PCR products were kept at 4 °C until further use. A small aliquot of the PCR products was subjected to gel electrophoresis to examine the PCR efficiency. Purified PCR products (QIAquick PCR purification kit; Qiagen, Germantown, MD, USA) were used for transformation of competent BL21(DE3) cells (Novagen, Madison, WI, USA), and kanamycin-resistant clones were selected and sequenced to verify the mutations. Table S1 lists the mutations and associated primers used to generate plasmids encoding the mutants used in this study.

Generation of rhBD-3 and its mutants

Wild-type rhBD-3 and its mutants were expressed and purified as described previously [6, 12]. Briefly, His6-tagged fusion protein was expressed in BL21(DE3) cells by induction with 1 mm isopropyl thio-β-d-galactoside induction. Ni2+-affinity columns [Ni2+–nitrilotriacetic acid resin from Qiagen] and RP-HPLC (C-18 column; Phenomenex, Torrance, CA, USA) were used to isolate hBD-3/mutants fusion proteins, and this was followed by digestion with enterokinase (Novagen) to release mature peptides. The mature forms of rhBD-3 and/or mutated variants were purified with RP-HPLC. The molecular masses of recombinant proteins were verified by MS.

Ca2+ mobilization assay

SDF-1α-induced increases in cytosolic free [Ca2+] were determined with the Fura-2 fluorescent Ca2+ indicator dye. Briefly, CEM X4/R5 cells were suspended at 5 × 106 cells·mL−1 in calcium assay buffer (CAB: 130 mm NaCl, 5 mm KCl, 1 mm MgCl2, 1 mm CaCl2, 20 mm Hepes, 5 mm glucose, and 0.1% BSA). After addition of 1 μm Fura-2 AM (Molecular Probes, Eugene, OR, USA), cell suspensions were incubated at 37 °C for 30 min, pelleted by centrifugation at 1000 rpm for 5 min, washed twice with CAB, and resuspended in CAB at 1 × 106 cells·mL−1. Aliquots (1.5 mL) of Fura-2-loaded cell suspension were assayed in UV-transparent plastic cuvettes with continuous magnetic stirring and thermostatting at 37 °C. Excitation illumination was from a 75-W xenon arc lamp filtered through a 340-nm narrow bandpass optical filter. Fura-2 fluorescence emission through a 500-nm narrow bandpass optical filter was detected and quantified with an EMI photomultiplier tube. Unless noted otherwise, CEM cells were treated with 20 μg·mL−1 hBD-3 or hBD-3 mutant (or vehicle control) for 2–3 min prior to stimulation with 10 nm SDF-1α. The SDF-1α-induced transient changes in Fura-2 fluorescence were recorded for another 2–3 min prior to plasma membrane permeabilization by addition of 33 μg·mL−1 digitonin and release of the cytosolic dye into the extracellular CAB. The 1 mm Ca2+ of the CAB saturates the high-affinity Ca2+-binding sites of the released Fura-2 to provide a maximum Ca2+-saturated Fura-2 fluorescence signal (Fmax). The permeabilized cell suspension was then supplemented with 16.6 mm EGTA (to chelate Ca2+ bound to the Fura-2) and 66.6 mm Tris base (to alkalinize the medium to approximately pH 8.5 for increased affinity of the Ca2+–EGTA complex); this results in a minimum Ca2+-free Fura-2 fluorescence signal (Fmin). The resulting Fmax and Fmin signals were used to calibrate in the intracellular Fura-2 fluorescence signals (F) recorded prior to permeabilization according to the equation [Ca2+] = 224 nm × (F − Fmin)/(Fmax − F), where 224 nm is the Kd for the Ca2+–Fura-2 complex at the [Mg2+], pH and ionic strength that define the cytosolic milieu. The calculated [Ca2+] values corresponding to the intracellular Fura-2 fluorescence were used to define the y-axes on the representative SDF-1α-induced Fura-2 fluorescence transients illustrated in the figures.

[125I]SDF-1α-binding assays

As described previously [6], CEM X4/R5 cells were washed and resuspended in CAB. Cells (106 in a final volume of 300 μL) were aliquoted into microfuge tubes, and supplemented with 0.1 nm [125I]SDF-1α (PerkinElmer NEN). For analysis of saturation binding (Fig. 2), unlabeled SDF-1α (0–30 nm) was also with or without 5 μg·mL−1 hBD-3. Cell suspensions were incubated at 37 °C for 20 min, and rapidly centrifuged (2500 rpm for 3 min) to pellet the cells. Following aspiration of the supernatant, the cell pellets were washed by resuspension in NaCl/Pi and recentrifugation (2500 rpm for 3 min). [125I]SDF-1α in the final washed cell pellet was quantified by γ-counting. For analysis of high-affinity SDF-1 binding in the presence of hBD-3 or hBD-3 mutants (Figs  3H, 4G, 5E, 6G, 7E, 8E, and 9G), CEM cells were incubated for 20 min with 0.1 nm [125I]SDF-1α in the absence or presence of 20 μg·mL−1 wild-type hBD-3 or mutant hBD-3, or 10 nm unlabeled SDF-1α, prior to sedimentation, washing, and quantification by γ-counting. For each experiment, the amount of SDF-1α bound to the cells in the presence of hBD-3 mutants was normalized to the amount bound in the absence of any competing ligands (considered to be 100%), and compared with the blocking efficacies of wild-type hBD-3 or saturating unlabeled SDF-1α. All experiments were performed in triplicate.


This study was supported by NIH/NIDCR PO1DE019759 (A. Weinberg) and by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research (J. T. Lubkowski).