Tetrameric assembly disruption impairs CXCL4 chaperon ability for DNA and type I interferon immune amplification

Abstract

Chemokine (C-X-C motif) ligand 4 (CXCL4), an antimicrobial chemokine with multiple effector functions, is a known biomarker in systemic sclerosis (SSc), but it also plays important roles in inflammatory bowel diseases, atopic dermatitis, psoriasis, systemic lupus erythematosus and other chronic inflammatory diseases. In SSc, the presence of CXCL4 correlates with severe disease progression: CXCL4 contributes to the interferon-(IFN)-I signature by forming pro-inflammatory liquid crystalline complexes with self-DNA, which have the capacity to activate TLR9 in a multivalent manner and potently induce IFN-α in plasmacytoid dendritic cells. Here, we show that in order to chaperone DNA to amplify TLR9 responses effectively, CXCL4 needs to retain its ability to assemble into tetramers, in addition to forming the optimal liquid crystalline structure. By comparing various mutated CXCL4 derived peptides with altered cationic charge and with ablated capacity to from stable tetramers, we find the surprising result that the physiological pre-assembly of CXCL4 into tetramers, before DNA interaction, is important for strong amplified IFN-I secretion. Most importantly, small molecules that destabilize CXCL4 tetramerization block the plasmacytoid dendritic cells IFN-α response, suggesting new pharmacological interventions in SSc, and possibly in other autoimmune conditions characterized by the presence of high CXCL4 and CXCL4-DNA complexes expression.

Molecular analysis and functional data shed light on the crucial role of CXCL4 tetramerization in immune amplification mediated by CXCL4-DNA complexes in SSc and in other autoimmune diseases characterized by platelets activation

Introduction

The antimicrobial chemokine (C-X-C motif) ligand 4 (CXCL4) (also called PF4, platelet factor 4), plays a direct role in exacerbating inflammation in systemic sclerosis (SSc), an autoimmune disease characterized by fibrosis of the skin and/or internal organs^1,2. By forming CXCL4-DNA complexes, which can be detected in both SSc blood and the skin, CXCL4 amplifies. Toll like receptor (TLR9)-mediated type I interferon (IFN-I) production in plasmacytoid dendritic cells (pDCs), the main IFN-I producing cells^3–5. The spatially periodic structures formed by CXCL4-DNA complexes have inter-DNA lattice spacings well-optimized for TLR9 activation, a TLR expressed by pDCs, which recognizes double strand (ds)DNA^3,4. The emerging picture from this result and cognate results from other peptide-nucleic acid complexes^3,4,6–9 indicate that peptides and proteins can modulate interactions between immune ligands (such as nucleic acids) and their cognate receptors (TLR), by forming supramolecular structures. For example, antimicrobial peptides (AMP), bacterial toxins, and amyloids can all form supramolecular scaffolds to present spatially-ordered DNA to TLR9 to amplify immune responses in various inflammatory diseases. This activation is strongly influenced by the inter-DNA spacing between adjacent parallel DNA ligands. Inter-DNA spacing close to the steric size of TLR9 allows for optimal multivalent binding of columnar DNA lattices to clustered TLR9 arrays^3,4,6–9.

It is well-known that the peptide cationic charge is a necessary condition for organizing strongly anionic nucleic acids such as DNA into geometrically ordered complexes capable of amplified TLR9 stimulatory ability. It is equally well- known that peptide cationic charge is not a sufficient condition since not all cationic molecules can amplify TLR9 activation⁶. For example, even strongly cationic molecules such as the TAT cell penetrating peptide do not lead to immune activation, since they organize DNA complexes with close-packed rather than open structures. Clearly, the complex interplay between electrostatics, structure, and the concomitant processes that lead to the necessary processes of cell entry, endosomal access, and TLR9 activation are important for CXCL4. Here, we identify an unanticipated characteristic that leads to strong TLR9 stimulation by CXCL4-DNA complexes: We find that the ability of CXCL4 to assemble into stable tetrameric structures at physiological conditions⁹ is critical for strong immune amplification.

To dissect the functions of the modular CXCL4 structure we studied various mutated or truncated CXCL4 derived peptides with altered cationic charge (the carboxyl-terminal modules of CXCL4 have more cationic charge than full-length CXCL4) and with ablated capacity to form stable tetramers^11–13. We also used small molecules that interfere with tetramer formation. Our data indicate that the C-terminal domain of CXCL4 contributes to DNA interactions, which is stronger in the presence of a higher cationic charge. However, CXCL4 tetramerization is crucial for strong TLR9 stimulation. We hypothesize that CXCL4 tetramerization alters the effective steric size and multivalency of the fundamental cationic unit that interacts with anionic DNA, which influences the state of ‘overcharging’ of the complex^12,14,15, and thereby its size to promote cell entry and endosomal access^12,14,15. These data are relevant for SSc pharmacological interventions, but also for the treatment of any autoimmune and autoinflammatory disorder in which CXCL4 is up-regulated (and dysregulated) in blood and tissues.

Results

Peptide CXCL4-47-70 organizes DNA in complexes into a liquid crystalline structure apparently suitable for TLR9 multivalent binding

CXCL4 contains a cationic alpha-helical structure in its C-terminal domain, involved in the binding to the anionic heparin^10–13, which suggests that the same domain mediates DNA binding. At present, it is unknown whether bioactive metabolites of CXCL4 preserve this ability to amplify TLR9 activation via self-assembly with DNA. If the stimulatory ability of CXCL4 were solely due to the cationicity of its C-terminal domain, shorter terminal domain peptides with increased cationic charge might be even more efficient at stimulating TLR9 in complex with DNA. We chose to study first a 24mer peptide called CXCL4-47-70, described as the minimal sequence of CXCL4 that mediates functional effects, namely its anti-angiogenic activity^12,13. In parallel, we tested a correspondent peptide with reduced cationicity, called CXCL4-L1-47-70, mapping to the COOH-terminal part of the CXCL4-L1 chemokine, the non-allelic variant of CXCL4^12,13. CXCL4-47-70 (NGRKICLDLQAPLYKKIIKKLLES) is a 24-mer C-terminal peptide of CXCL4; CXCL4-L1-47-70 (NGRKICLDLQALLYKKIIKEHLES), the variant, presents three amino acid substitutions (in bold), which lower its cationic charge. The C-terminal component of CXCL4 has an alpha-helical secondary structure^13,15. At neutral pH, CXCL4-47-70 has a net cationic charge of +4.0, compared to +3.1 for CXCL4; +2.0 is the charge of the CXCL4-L1-47-70 peptide. We expect both CXCL4-47-70 and CXCL4-L1-47-70 to form ordered structures with DNA based on their cationic charge. Using small-angle X-ray scattering (SAXS), we characterized the corresponding peptide-DNA structures, using DNA with differing degrees of fragmentation (Table 1)³. SAXS measurements were performed on equilibrium phase-separated liquid-crystalline peptide–DNA complexes in aqueous solution (140 mM NaCl + 10 mM HEPES, pH 7.4, 37 °C).

Table 1.

CXCL4-47-70 organizes DNA into liquid-crystalline supramolecular complexes at an inter-DNA spacing optimal for amplifying TLR9 activation

Condensing agent	Human DNA low degr	Human DNA med degr	Human DNA high degr
CXCL4	Disordered columnar q1 = 1.56 nm-1 d = 4.03 nm	Disordered columnar q1 = 1.55 nm-1 d = 4.05 nm	Disordered columnar q1 = 1.55 nm-1 d = 4.05 nm
CXCL4-47-70	Disordered columnar q1 = 1.64 nm-1 d = 3.83 nm	Disordered columnar q1 = 1.70 nm-1 d =3.70 nm	Disordered columnar q1 = 1.64 nm-1 d = 3.83 nm
CXCL4-L1 47-70	No signal	No signal	No signal

Values of the “q1” and “a” (distance between each DNA ligand) for CXCL4, CXCL4-47-70 and CXCL4-L1-47-70 in the presence of three different preparations of human DNA at low (DNA sonicated for 30 s), intermediate (DNA sonicated for 60 s) and high (DNA sonicated for 150 s) fragmentation (see Methods and ref. ³).

Data are from one representative experiment of two performed.

CXCL4-47-70 recapitulated the characteristics of full-length CXCL4, and actually seemed to possess slightly improved DNA-organizing property. Figure 1 and Table 1 show that CXCL4-47-70 forms complexes with inter-DNA spacings in the range of d ~ 3.7–3.8 nm (q₁ = 0.164–0.170 Å^-1). This result predicts strong TLR9 activation and IFN-α production. In comparison, CXCL4-DNA complexes have spacing in the d ~ 4.0 nm range (q₁ = 0.155 Å^-1) as previously shown³. The structures formed were not strongly sensitive to the degree of DNA fragmentation. Interestingly, CXCL4-L1-47-70 was unable to form ordered structures with DNA under similar conditions, predicting a diminished or null ability to activate TLR9. These results suggest that the cationic charges of the short C-terminal domain of CXCL4 are indeed important for forming structures that can stimulate TLR9.

Fig. 1. CXCL4-47-70, unlike CXCL4-L1-47-70, organizes DNA into liquid-crystalline supramolecular complexes at an inter-DNA spacing optimal for amplifying TLR9 activation.

SAXS data of CXCL4-47-70 and CXCL4-L1-47-70, and control CXCL4, bound to human DNA: The first peak positions q1 is indicated, where present. The fragmentation state of DNA (the highest in DNA sonicated for 150 s, the lowest in DNA sonicated for 30 s, see Methods and ref. ³) does not significantly affect the ability to form complexes with DNA suitable for TLR9 triggering. Data are from one representative experiments of two performed.

CXCL4-47-70 peptide has a poor stimulatory ability of pDCs in complex with DNA

According to the results presented above, we expected to see that CXCL4-47-70-DNA complexes were efficient TLR9-stimulators and CXCL4-L1-47-70-DNA complexes were not. To confirm this, each peptide was mixed with fragmented (sonicated, see Methods) DNA. As control, we also used full-length CXCL4 alone or in complex with DNA. CXCL4 alone, as expected³, was not stimulatory for plasmacytoid dendritic cells (pDCs), whereas in complex with DNA (CXCL4-DNA) was highly stimulatory. CXCL4-L1-47-70 and CXCL4-47-70 alone were also unable to stimulate IFN-α, like CXCL4 alone. As expected from the SAXS data, even in complex with DNA, CXCL4-L1-47-70 was ineffective at stimulating IFN-α in pDCs. Unexpectedly though, the stimulatory ability of CXCL4-47-70 in complex with DNA was also poor, significantly lower than that of the full-length CXCL4 (Fig. 2; supplementary data set 1). This, despite CXCL4-47-70 condensed the DNA into the proper structure needed for TLR9-stimulation, with optimal distance among the DNA ligands bound by the molecule. These results suggest that the ability of the CXCL4-terminal domain to impart to DNA ligands a regular spacing, suitable for TLR9-triggering, is not sufficient for a proper TLR9 signaling and IFN-α release.

Fig. 2. CXCL4-47-70 does not properly activate pDCs.

PDCs were stimulated with human DNA (see Methods) in complex with CXCL4-47-70 or CXCL4-L1-47-70 or CXCL4 at different µMolar concentrations (µM), or with the same peptides in the absence of DNA (the human DNA alone gave values of zero pg/ml). IFN-α production was quantified by ELISA after 24 h. Cumulative results are presented in a dispersion diagram as individual aligned replicates, median-connected; P-values refer to different stimulatory ability of CXCL4-DNA compared to CXCL4-47-70-DNA and are from Mann–Whitney’s t test (one-tailed); N = 5 (CXCL4-47-70-DNA), N = 6 or N = 7 or N = 8 (for CXCL4-DNA depending on the dose, 1 μM, or 2 μM, respectively) from independent experiments at different doses.

A 54mer CXCL4-C-terminal domain-derivative efficiently stimulates TLR9 and IFN-α production in pDCs

The data reported above suggest that other CXCL4-portions may be also necessary for the efficient TLR9-stimulatory ability of CXCL4-DNA complexes. To address this, we tested a C-terminal domain of CXCL4 with a more extended amino acid sequence. We selected a form composed by 54 amino acids (CXCL4-17-70). CXCL4-17-70 is interesting in that it can be proteolytically generated from full-length CXCL4 during a mixed lymphocyte reaction (MLR) in vitro, a condition that mimics inflammation¹⁶. Intriguingly, Western blots performed in denaturing conditions in a limited number of SSc plasma, showed the presence of monomeric forms of CXCL4, including, in some SSc patients, the visualization of a shorter form of CXCL4. The latter may be compatible with the presence of slightly shorter form of CXCL4 (Fig. S1). CXCL4-17-70 is more cationic than CXCL4, having a charge of +7.1, double than that of full-length CXCL4. In an electrophoretic mobility shift assay (EMSA), CXCL4-17-70 appeared very proficient in delaying DNA migration (Fig. 3a), indicating a stronger DNA binding ability at lower doses, compared to full-length CXCL4 (Fig. 3b). Functional data of pDC-stimulation assays showed that CXCL4-17-70-DNA complexes were more effective at stimulating IFN-α than the full-length CXCL4-DNA complexes, when used at the lowest μMolar concentrations (Fig. 3c; Supplementary data set 2). Figure 3d Supplementary data set 2), and confirmed that also CXCL4-17-70-DNA complexes stimulated TLR9, as bafilomycin, hydroxycloroquine (HCQ) and the TLR9-inhibitors A151 and ODN2088, all significantly abolished the effect of such complexes on pDCs. As further controls, we used two inhibitors of stimulator of interferon genes (STING), acting at two distinct STING pathway steps, called H-151¹⁷ and G-140¹⁸. Both H-151 and G-140 diminished the pDCs activation. The same inhibitors were used during stimulation of pDCs with full-length CXCL4-DNA complexes with similar results (Fig. S2). This data confirmed that the cationic charges of the C terminal domains of CXCL4 are important to ensure the DNA binding ability but suggest that a more extended part of CXCL4 is required for optimal TLR9-stimulation, and likely for STING-driven stimulation.

Fig. 3. CXCL4-17-70 is a better DNA binding peptide and is highly stimulatory for pDCs via TLR9.

EMSA assay performed on a 10% agarose gel, which shows the delay of DNA migration in the presence of bound CXCL4-17-70 (a) or full-length CXCL4 (b) peptides at scalar doses (from 0.75 to 24 µg, as indicated in the Figure). Representative experiments of two performed. c PDCs were stimulated with huDNA (DNA) in complex with different µMolar concentrations (indicated on X-axis, µM) of CXCL4-17-70 and full-length CXCL4 or with the two peptides without DNA. IFN-α production was quantified by ELISA after 24 h. Cumulative results are presented in a dispersion diagram as individual aligned replicates, median-connected; P-values refer to different stimulatory ability of CXCL4-DNA compared to CXCL4-17-70-DNA and are from Mann–Whitney’s t test (two-tailed). N = 8 to N = 10 (CXCL4-17-70-DNA), N = 6 to N = 10 (for CXCL4-DNA depending on the dose), from independent experiments at different doses. d PDCs were stimulated with: DNA in complex with the CXCL4-17-70 peptide (at 1 µM), or with CXCL4-17-70 alone, in the absence or presence of bafilomycin (Bafilo) or A151 or ODN2088 (inhibitors of TLR9), hydroxycloroquine (HCQ), to confirm involvement of TLR9 in IFN-α production by pDCs induced by the CXCL4-17-70-DNA complexes. H-151 and G-140 (inhibitors of the STING cytosolic pathway) were also used in the assay. IFN-α production was quantified by ELISA after 24 h. Horizontal bars are the mean, vertical bars are standard error of the mean; P values from Wilcoxon’s test (one-tail). Five independent experiments for A151, bafilo, HCQ, and G-140 inhibition, n = 10 and n = 14 independent experiments for ODN2088 and H-151 inhibition, respectively.

Preventing CXCL4 tetramer formation drastically lowers the potency of CXCL4-DNA complexes stimulation of pDCs

The results presented above strongly suggest that TLR9 activation requires not just structure but other processes to occur. One of these processes could be the formation of tetramers. Indeed, results from the scientific literature indicate that CXCL4 mainly exists as tetramers in physiological conditions^{10–13,19–21}. The amino acid residues that determine this tetrameric structure reside in the NH2-/middle-portions of the CXCL4 molecule^19–21. Both full-length CXCL4 and peptide CXCL4-17-70, unlike the CXCL4-47-70^{11, 19-10} peptide, comprise the amino acid residues (aa 28 and 50) that allow tetramers formation^11,19–21. To address whether the tetrameric structure of CXCL4 molecule was important for the TLR9 stimulatory capacity of CXCL4-DNA complexes, we took advantage of mutated forms of CXCL4 (and CXCL4-17-70), in complex with DNA. The two peptides, called CXCL4-K50E and CXCL4-17-70-K50E present a mutation at the amino acid residue 50, where a lysine (K) has been substituted with a glutamic  acid (E). This substitution impairs the capacity of CXCL4 to form a tetrameric structure^11,19–21. The results in Fig. 4 show that the mutated forms of both CXCL4 (Fig. 4a; Supplementary data set 3) and CXCL4-17-70 (Fig. 4b; Supplementary data set 3), were incapable of stimulating pDCs when complexed with the DNA. The K-to-E substitution also lowers the cationic charge of the two mutated peptides: the charge of CXCL4-K50E drops to +1.1, which might explain less efficient stimulation; however, the cationicity of CXCL4-17-70-K50E remains high (+5.1), still much higher than that of full-length CXCL4. Thus, we verified whether CXCL4-K50E and CXCL4-17-70-K50E peptides still bound to DNA. Both peptides were able to delay DNA migration in a EMSA assay, as shown in Figs. S3 and S4, respectively (Fig. S3b and Fig. S4b), which indicates their binding to DNA. Overall, these results suggest that an amino acid substitution that impairs tetramerization abolishes immune amplification ability on IFN-α release via TLR9 for both CXCL4 and its derivative CXCL4-17-70 peptide.

Fig. 4. Mutated versions of CXCL4 and CXCL4-17-70, which do not form tetramers, do not activate pDCs in complex with DNA.

Different µMolar concentrations of CXCL4 and CXCL4-K50E (a), or CXCL4-17-70 and CXCL4-17-70-K50E (b) peptides, were pre-mixed with DNA to form complexes. Complexes of each kind of peptides were added to the pDCs. IFN-α production was quantified by ELISA after 24 h. Production of IFN-α in response to CXCL4, CXCL4-17-70, CXCL4-K50E, CXCL4-17-70K70E alone or DNA alone was zero. Cumulative results are presented in a dispersion diagram as individual aligned replicates, median-connected. P-values refer to differences between the wild type peptide and the mutated peptide in both (a) and (b), and are from Mann–Whitney’s test (two-tailed). N = 6 to N = 9 replications for CXCL4-DNA compared to N = 10 replications for the CXCL4-K50E-DNA and N = 7 to N = 9 compared to N = 6 to N = 9 replication for CXCL4-17-70-DNA and CXCL4-17-70-K50E-DNA, respectively.

The tetrameric CXCL4 structure disruption by small molecules significantly reduces the ability to stimulate pDCs and IFN-α by CXCL4-DNA complexes

To confirm that the ability to form tetramers plays a role in regulating the strength of TLR9-stimulation by CXCL4 after forming complexes with the DNA, we took advantage of the use of small molecules called inhibitor A3101 and A3104, respectively, both previously designed to disrupt CXCL4-tetramer formation²¹ in heparin induced thrombocytemia (HIT)²². The presence of these inhibitors during the formation, in vitro, of CXCL4-DNA or CXCL4-17-70-DNA complexes determined a dose dependent and significant decrease in the capacity of these complexes to stimulate IFN-α production by the pDCs, as shown in Fig. 5a, b (and Supplementary data set 4), respectively. To confirm the specificity of the observed inhibition, and exclude that inhibition was due to an intrinsic toxicity of the two small molecules, we tested their effect on pDCs stimulated with CpGA. Inhibitor A3101 did not significantly affect IFN-α production in response to CpGA (Fig. 5c; Supplementary data set 4); the highest doses of inhibitor A3104 lowered the IFN-α production by pDCs (Fig. 5d; Supplementary data set 4). These control experiments suggested that the small molecule A3101 blocked CXCL4-DNA and CXCL4-17-70-DNA complexes driven pDCs stimulation in a specific manner and not because of its intrinsic toxicity. The inhibitor A3104 seemed to show some toxicity, but only when it was utilized at the highest dose (500 μM). To further exclude toxicity, we performed lactate dehydrogenase (LDH) assay using the concentrations of 100 and 250 μM of both inhibitors. To this end, we cultured pDCs in the presence of IL-3 (a stimulus that does not usually induce IFN-I production but ensures survival of the cells)²³ in the presence/absence of the two inhibitors. Culture supernatants were harvested after a 24 h culture and tested in LDH assay (see Methods). Both A3101 and A3104 were not toxic, although a tendency to more toxicity was seen in the presence of A3104, but differences observed were not statistically significant (Fig. S5). We verified by immune blot that A3101 truly interfered with tetramer formation. This was the case for both CXCL4 and CXCL4-17-70 peptides (Fig. S6). Moreover, we also checked whether the tetramer inhibitor A3101 reduced CXCL4 complex formation with DNA. To this end, we made an EMSA assay by mixing a fixed dose of CXCL4 with increasing doses of A3101 and, afterwards, we made a mixture with DNA (Fig. S7). The migration on a gel of the DNA was consistently delayed in the presence of CXCL4 (as expected), indicating good binding to the DNA. The addition of graded doses of A3101 did not drastically change DNA migration, although more DNA migrated into the gel. This suggests that, in part and especially at the highest doses of A3101, less DNA bound to CXCL4.

Fig. 5. Disruption of the tetrameric structure diminishes the effect of CXCL4- and CXCL4-17-70 in complex with DNA.

CXCL4 (a) and CXCL4-17-70 (b) peptides were pre-mixed with the inhibitor A3101 or A3104 (at three scalar µMolar concentrations: 100, 250 and 500). After 10 min, DNA was added to the mixture. The mixtures containing each molecule bound to DNA in the presence or absence of each inhibitor was administered to the pDCs. IFN-α production was quantified by ELISA after 24 h. Horizontal bars are the mean, vertical bars are standard error of the mean; P-values from Wilcoxon’s test (two-tails). N = 9 or N = 11 (for A3101 inhibition) from different experiments. Inhibitor A3101 (c) or A3104 (d) were given, (at the same concentrations used in panel a–b), to pDCs stimulated with CpGA to address toxic effects. Horizontal bars are the mean, vertical bars are standard error of the mean, P-values from Mann–Whitney’s test (one-tail). N = 8 (N = 7 for A3101 inhibition) in Panel (c) and N = 6 (N = 5 for A3104 inhibition) in Panel (d).

This data indicates and confirms that the tetrameric structure of CXCL4 is an important determinant of a successful CXCL4-DNA complexes mediated stimulation of TLR9 in pDCs, leading to immune amplification of the IFN-α response.

The small molecule A3101 efficiently blocks the interferogenic effect of natural CXCL4-DNA complexes derived from SSc blood

We have previously shown that natural plasma-derived CXCL4 behaves like synthetic CXCL4. Indeed, the addition of exogenous DNA to SSc plasma containing high amounts of CXCL4 significantly increased IFN-α production by pDCs³.

To set up an in vitro system, which mimics the in vivo situation, we selected SSc plasma with a high CXCL4 content (the range concentration in CXCL4 of “CXCL4-positive plasma” was 39,189.4 ±  8,726.74 pg ml − 1; the range content of “CXCL4-negative plasma” was 71.7 pg ml − 1  ±  32.7), and measured, by homemade ELISA³ the presence of CXCL4-DNA complexes (see Methods). Figure 6a (and Supplementary data set 5), shows that results of ELISA test for detection of CXCL4-DNA SSc plasma as compared to HD by home-made ELISA test (detail is Methods). We used a 1:50 dilution of several CXCL4-DNA positive plasma to stimulate pDCs. While almost all CXCL4-DNA complexes positive plasma stimulated pDCs (20 out of 23 plasmas stimulated variable amounts of IFN-α release by the pDCs, 87%), only 3 out of 12 (25%) of the negative plasma induced some IFN-α (Fig. 6b; Supplementary data set 5). Figure 6c (and Supplementary data set 5), shows that the pDCs stimulation by the CXCL4-DNA positive plasma was lowered or inhibited by the addition of neutralizing anti-CXCL4 antibodies (see methods), suggesting that CXCL4 was involved in the stimulation. Control antibodies (irrelevant rabbit IgG or rabbit anti-MPO antibodies) did not induce a significant inhibition of pDCs activation.

Fig. 6. A3101 efficiently blocks the interferogenic effect of natural CXCL4-DNA complexes derived from SSc blood.

a Plasma of SSc patients (N = 95) and healthy donors (HD, N = 16) were tested for presence of CXCL4-DNA complexes as described in Methods. Number of SSc plasma positive for CXCL4-DNA expression were 39 out of 95 (41%). Horizontal bars are the means, vertical bars are standard errors of the means, P value is from a Mann–Whitney test, one-tail. Background level of complexes in HD is also reported on the graph as dotted line (OD = 0.36, mean of HD ± 2 standard deviations). b PDC were treated with plasma, that were positive (N = 32) or negative (N = 11) for CXCL4-DNA complexes; P value is from a Mann–Whitney test, one-tail. c CXCL4-DNA positive plasma were incubated with blocking anti-CXCL4 antibodies or with the small molecule inhibitor A3101 (100 µM) or with irrelevant antibodies IgG rabbit or anti-MPO (see methods). The cells were cultured for 24 h and the IFN-α was measured in the culture supernatants by ELISA. In the panels, horizontal bars are the mean, vertical bars are standard error of the mean, P values by Mann–Whitney’s test (two-tails). In the panel c plasma are N = 32, repetitions with anti-CXCL4 are N = 28, with A3101 are N = 15, for rabbit irrelevant Abs is N = 10.

Most importantly, addition to the stimulatory plasma of inhibitor A3101 significantly abolished pDCs activation and IFN-α release. These results suggest that it is possible to interfere with the natural CXCL4-DNA complexes ability to stimulate IFN-α by disrupting the CXCL4-tetrameric structure. Also in these kind of experiments we performed a LDH assay using culture supernants harvested after 24 h of culture. PDC were stimulated with different SSc plasma containing CXCL4-DNA complexes in the presence of A3101 (we also added A3104, as further control). The two molecules did not induce a particular cytotoxicity, although A3104 induce a little more toxicity compared to A3101 (Fig. S8); however, this difference was not statistically significant.

These results reinforce the observations reported above that A3101 suppressed pDCs production of IFN-α by disrupting tetramers and not by a mere toxic effect (intrinsic cytotoxicity) on the cells.

Disruption of CXCL4 tetramerization inhibits the stimulatory properties of CXCL4-DNA complexes in another autoimmune disease

The data shown above suggest that tetramerization is important for the interferogenic properties of CXCL4 in complex with DNA. Such complexes can be present in other autoimmune/ autoinflammatory diseases characterized by platelets activation (see below). We have previously shown that systemic lupus erythematous (SLE) patients have a high concentration of CXCL4 in their blood³. CXCL4 is considered a renal disease biomarker in SLE and is found up-rgulated in SLE urine²⁴. We addressed the presence of CXCL4-DNA complexes in plasma of a SLE cohort. Of a cohort with high CXCL4 expression, we found that 8 SLE patients (out of 115, 7%), harbored in blood CXCL4-DNA complexes (Fig. 7a; Supplementary data set 5), as compared to 12 HD plasma.

Fig. 7. A3101 efficiently blocks the interferogenic effect of natural CXCL4-DNA complexes derived from SLE blood.

a Plasma of SLE patients (N = 115) and healthy donors (HD, N = 14) were tested for presence of CXCL4-DNA complexes as described in Methods. Number of SLE plasma positive for CXCL4-DNA were 8 out of 115 (7%). Horizontal bars are the means, vertical bars are standard errors of the means, P values from a Mann–Whitney test, one-tail. Background levels of complexes in HD is also reported on the graph as dotted line (OD = 0.12, mean of HD ± 2 standard deviations). b CXCL4-DNA positive SLE plasma were incubated with a blocking anti-CXCL4 or an irrelevant antibody (rabbit irrelevant antibody), or with the small molecule inhibitor A3101 (100 µM). The cells were cultured for 24 h and the IFN-α was measured in the culture supernatants by ELISA. In all Panels, horizontal bars are the mean, vertical bars are standard error of the mean, P values by Wilcoxon’s test (one-tail). Plasma N = 7 and inhibition with A3101 is N = 7. Inhibition with anti-CXCL4 or irrelevant Abs are both N = 5.

We used the SLE CXCL4 positive plasma (1:50 dilution) to stimulate pDCs and we found that they were stimulatory in a CXCL4 dependent manner (a blocking anti-CXCL4 diminished IFN-α production). The inhibitor A3101 was able to block pDCs activation by the CXCL4-DNA positive plasma in a significant manner (Fig. 7b; Supplementary data set 5). These results confirm, in a disease setting other than SSc, that tetramerization is an important determinant of the interferogenic ability of CXCL4 in complex with DNA.

Discussion

In this study, by using mutated CXCL4 versions of different lenght and small molecules that destabilize CXCL4 oligomerization²¹, we dissect the fine mechanism of the CXCL4 chaperone ability and capacity to organize DNA to stimulate TLR9. We report the unanticipated result that the ability of CXCL4 to assemble into stable tetrameric structures at physiological conditions is critical for TLR9 activation and strong pDC-IFN-α release. Strikingly, we find that the ability of CXCL4 to organize DNA into a liquid crystalline form is a necessary but not sufficient condition for strong activation, as revealed by experiments on CXCL4 C-terminal domains of different lengths showing different functions. Indeed, a short peptide, CXCL4-47-70, organizes DNA into a structure commensurate with TLR9 steric size (in a manner similar to CXCL4 and LL37^3,4,7,9), but produced only negligible pDCs stimulation. Of note, this short peptide lacks the necessary amino acids residues required for tetramer assembly, as an interaction between the amino acids 28 and 50 is important for tetramerization^10,19–21. Thus, it appears that it is the retention of the ability to aggregate into a tetrameric state that is crucial for the CXCL4-DNA complexes to stimulate TLR9 and IFN-α in pDCs. This is supported by the data on the mutated form of CXCL4 (CXCL4-K50E), incapable to form tetramers and, in fact, virtually devoid of pDCs stimulatory ability in complex with DNA. It could be argued that CXCL4-K50E peptide has lowered its cationic charge due the applied substitution (new charge: +1.1 instead of +3.1). However, the mutated CXCL4 was still able to bind the DNA in EMSA assays although will lower strength. Despite this, a reduction of the cationic charge cannot, solely, justify complete drop of the interferogenic potential, especially if considering the results obtained with peptide CXCL4-17-70-K50E. Indeed, the latter retains a consistent positive charge (+5.1), higher than that of full-length CXCL4 ( + 3.1), and a high DNA binding ability. Despite this, it failed to stimulate pDCs release of IFN-α in complex with DNA.

Altogether, these findings indicate that the liquid crystalline structure at an inter-DNA distance near 3.5–4.00 nm^3,4 in the DNA sub-phase does not represent a sufficient condition for ensuring strong stimulation. This stimulation involves not just local structure for TLR9 binding, but efficient cell entry and endosomal access^14,15. Indeed, the CXCL4-DNA complex is stabilized by entropically modulated electrostatic interactions between anionic DNA and a cationic macroion. The process of tetramer formation changes the cationic macro-ion from the CXCL4 monomer (+3 charge) to a CXCL4 tetramer that is more multivalent (+12 charge) and also sterically larger. Intuitively, one can expect that the electrostatic complexes between CXCL4 and DNA are neutral, but that is not the case, as the overall charge of the electrostatic complex can be non-neutral, and is a function of the macro-ion multivalence and steric size^8–10,14,15. These ‘overcharging’ effects are expected to impinge on the size of the complex by suppressing van der Waals attraction and thereby promote cell entry and endosomal access of the complexes. We hypothesize that this hierarchical assembly of CXCL4 leads to non-stoichiometric complexes and favors cell entry and endosomal access.

The same overcharging effect can explain the findings that the tetramer inhibitor A3101 seems to decrease the ability of CXCL4 to bind the DNA, an aspect that deserves further investigation. Indeed, this is in line with the possibility that tetramers reach a cationicity of +12, as compared to the single monomer (+3). However, again, a partial drop of the DNA binding in the presence of A3101 is anyway difficult to reconcile with the drastic reduction of pDCs activation by A3101 seen with CXCL4-DNA positive plasma from both SSc and SLE.

Noteworthy, the present data cement the role of the endosomal pathway and TLR9 involvement³ in the induction of IFN-α by CXCL4-DNA complexes, as two different TLR9-inhibitors, bafilomycin and HCQ all efficiently block pDCs activation mediated by these complexes. HCQ could be actually a safe way to inhibit these complexes in SSc and is already used to treat SLE²⁵. However, a significant inhibition of pDCs activation was also seen in the presence of the STING inhibitors H-151 and G-140, suggesting that this pathway is also involved in CXCL4-DNA stimulation of IFN-I. Since the small molecules inhibitors, as well as the mutated peptides that cannot form tetramers almost completely abolish IFN-α secretion by pDCs, it is likely that tetramerization is involved also in the pathway that is STING mediated. It will be interesting to check the effect of complexes on monocytes, and assess whether the use of mitochondrial or bacterial DNA in complex with CXCL4 has a deeper or lower effect on the cytosolic pathway activation. The STING pathway of DNA recognition has an important role for activation of IFN-I in SSc monocytes²⁶ as well SSc fibroblasts, affecting the cross-talk between BAG3 and the IFN-dependent gene IFITM-2¹⁸. Thus, CXCL4-DNA (or even CXCL4-RNA) complexes could activate these cells in SSc patients with a role for the CXCL4 tetrameric form.

The small molecules A3101 as well as A3104, had been previously selected for their capacity to disrupt tetramers of CXCL4 in complex with heparin²¹ (the latter are complexes that play a role in HIT)^21,22. The authors discussed the possibility to use these small molecules to block induction of anti-CXCL4 antibodies that are heparin-dependent and mediate damage in HIT²¹. In that case, the important goal to reach by disrupting tetramers of CXCL4 was to lower the size of the complexes formed with heparin to diminish their immunogenicity. Indeed, the largest it is the size of molecular complexes, the highest is the immunogenicity of such complexes²¹. Thus, this study²¹ and our findings, suggest that, in SSc, the utility of these small molecules in the inhibition of CXCL4 tetramer formation could be two-fold. At one end, this inhibition can block excessive IFN-α production by pDCs, which is a harmful phenomenon^26–37. At the other end, since CXCL4 becomes a target of auto-antibodies^38–40 in SSc, this effect may also limit the immunogenicity of circulating CXCL4 and CXCL4-DNA complexes. The complexes formed by CXCL4 with heparin could also play a role in SSc, as we demonstrated that autoantibodies to CXCL4 in SSc patients were also, in part, heparin-dependent⁴⁰. It is conceivable that anti-CXCL4 antibodies play pathogenic roles in SSc, as they tend to correlate with selected disease features and with plasma IFN-α; in addition, they may enhance the up-take of CXCL4-DNA complexes by the immune cells^38–40.

An additional interesting finding of the present study is that a slightly shorter peptide of CXCL4, CXCL4-17-70 is very efficient in forming complexes with DNA, which are more stimulatory for TLR9. It is possible that this form is present in vivo, in case of an excessive inflammation. Indeed, CXCL4-17-70 was originally detected in vitro during a mixed-lymphocyte reaction (MLR), a phenomenon that mimics an inflammatory milieu¹⁶. Although our immune blot of CXCL4 in SSc plasma detected a monomeric form of CXCL4 with a size compatible with that of CXCL4-17-70, at present our results are not sufficient to state that this form is generated in vivo during inflammation and is present in blood or tissues. A systematic analysis of the CXCL4-17-70 form in plasma and tissues (if present), and its relationship with disease status, warrant future investigations.

In conclusion, this study represents a dissection of the interferogenic capacity of CXCL4 in complex with DNA, which suggests a novel approach to block the CXCL4-inflammatory activity: the disruption of the CXCL4 tetrameric structure. This disruption translates into a drop of the pDCs stimulatory effect via TLR9, and via STING, in the presence of DNA. It is interesting that a recent study also suggested that a ways to target CXCL4 in inflammatory diseases could be inhibiting its oligomerisation⁴¹, which had been shown to be crucial for GAG binding and recruitment of inflammatory leukocytes⁴².

Thus, our data are in line with these findings and indicate that disrupting CXCL4 can block additional mechanisms that fuel inflammation. Among these, there is also the reported ability of CXCL4 to block the activation of B-cells to inhibit their tolerance induction capacity⁴³ via a redistribution of TLR9 ligands (CpG) into these cells.

Finally, it is of great interest to address, at this point, whether the tetramerization is also important for the capacity of CXCL4-RNA complexes to stimulate other TLRs, such as TLR7 and TLR8, in pDCs and other cell types, namely myeloid DC and B-cells^38–40,44. In SSc, the IFN-I pathways up-regulation can have deleterious effects^26–37. At disease onset, the IFN-I signature associates with complications at later times³³. The same appears true for CXCL4 up-regulation, as the highest CXCL4 expression at the disease onset results in more frequent later stages complications³¹. Our own data show that memory B-cells are stimulated to become antibody secreting plasma cells in response to both CXCL4-DNA and CXCL4-RNA complexes, whereas myeloid DC are stimulated by CXCL4-RNA complexes to produce TNF-α^38–40,44. Thus, early intervention with oligomerization blockers could restrain the activation of immune cells by CXCL4 in complex with nucleic acids.

Finally, relevant for clinicians, this new knowledge offers a tool for the development of novel therapeutic approaches that limit the interferogenic/pro-inflammatory capacity of CXCL4 not only in SSc, but also in other pathological conditions. Indeed, CXCL4, and even IFN-I, are also upregulated in different pathological conditions in which platelets play major roles^45–52. Here, we have shown, for instance, that CXCL4-DNA complexes can be present in SLE blood, and can be efficiently blocked by the inhibitor A3101, as observed for SSc. Importantly, CXCL4 is considered a biomarker of renal disease in SLE²⁴. The amounts of A3101 used in the present study to inhibit pDCs activation are  in line with the amounts used by Sachais et al.²¹. These authors show an inhibition of more than 50% of tetramers formation at a concentration between -4M and -3M, corresponding to concentrations between 100 and 1000 μM (therefore, we used a range between 100 and 500 μM and we obtained a significant inhibition already at 100 μM in the presence of full-length CXCL4, Fig. 6). Moreover, for natural CXCL4-DNA complexes present in plasma of both SSc and SLE a dose of 100 μM was sufficient to drastically block IFN-α production by the pDCs. The present data, obtained with A3101, are certainly a tool to prompt studies in pre-clinical animal models, given that LDH assays showed that the cell toxicity of this molecule is negligible. For the moment, we wish to provide to physicians the “proof of concept” that specific inhibitors could be design to block tetramerization and, therefore, the inflammatory effect of CXCL4-DNA complexes. Furthermore, the data on SLE plasma indicate that the oligomerization of CXCL4 is not context dependent (or disease dependent) and may be a characteristic of CXCL4 in different pathologic conditions, which indicates that interventions with tetramer disruptors can be beneficial in other clinical settings. These findings can pave the way for more effective tetramers blockers design by pharmacologists to obtain the best effect with the lowest possible dose of an inhibitor as this approach is not limited to SSc and SLE. For example, CXCL4 level correlates with disease duration in Sjögren’s syndrome (pSS) and with microvascular impairment. Among pSS patients, those with Raynaud’s phenomenon (RP) had abnormal CXCL4 levels more frequently as compared to patients without RP⁵⁰. CXCL4 is elevated in the synovial tissues of early RA patients and may contribute to chronic inflammation by increasing the immune cells trafficking to the site of inflammation⁴⁵. In both atopic dermatitis (AD) and psoriasis, CXCL4 can be expressed at high level in blood. In both diseases, CXCL4 is a marker of platelets activation, and CXCL4 levels correlate with disease activity in both AD and psoriasis⁴⁷. Finally, in inflammatory bowel diseases (IBD), CXCL4 marks active disease^46,51. IFN-I can play a role also in IBD, although its exact role is not completely understood⁵³. Overall, our results can be helpful in a strategy to contrast the immune-stimulatory and pro-inflammatory effect, as well as the antigenicity, of CXCL4 bound to nucleic acids or heparin or, in general, to poly-anionic molecules.

As a limitation of this study, we cannot currently forecast side effects linked to the use of small inhibitors of CXCL4 tetramerization in vivo. An animal model could possibly address this issue⁵⁴. Anti-CXCL4 antibodies are important in the normal anti-infectious responses when CXCL4 binds to bacterial products, for example, and allows opsonization phenomena⁵⁵. Thus, our study cannot predict whether inhibition of tetramer formation will lead to a deficiency of the normal protective immune response to pathogens.

Methods

Patients

This study was approved by the ethical committee of the institutions of the Sapienza University and of ISS, and was conducted according to the Declaration of Helsinki. Informed, written consent was obtained from SSc and SLE patients giving blood. All ethical regulation relevant to human research participant were followed. Patients satisfied the ACR/EULAR 2013 classification criteria for SSc⁵⁶.

Reagents

CXCL4, CXCL4-K50E, CXCL4-17-70 and CXCL4-17-70-K50E and CXCL4-47-70 and CXCL4-L1-47-70 were from Biomatik (Kitchener, ON, Canada).

Bafilomycin A was from Sigma Aldrich (USA), and was used at 10 μM. The TLR9 inhibitor ODN TTAGGG (A151) was used at a concentration of 3 µM and purchased from Invivogen. The inhibitor of TLR9 ODN 2088 was used at 3 µM, hydroxycloroquine (HCQ) was used at 1 μM; the STING inhibitors H-151 and G-140 were used both at 3 μM. All the latter inhibitors were purchased from Invivogen (San Diego, CA,USA). Human recombinant IL-3 was purchased from Peprotech and was used at 10 ng/ml.

DNA (huDNA) was extracted from peripheral blood mononuclear cells (PBMCs) (from buffy coats, see below) by using the Qiagen kits mericon DNA Bacteria Kit and DNA blood Maxi Kit (Germany). DNA fragmentation was obtained as² Five milligrams of DNA in a volume of 300 µl were fragmented by using the Sonics Vibra Cell sonicator (Sonics & Materials Inc.), with the following settings: 2, 4, and 10 sonication cycles (30 s ON, 30 s OFF in ice) to obtain the DNA fragment size between 100 and 1000 bp. We controlled the resulting size distribution on 2% agarose gel by electrophoresis. Small molecules A3101 (STK278345) and A3104 (STK275857) were purchased from Vitasmalab.biz.

Isolation and stimulation of blood pDCs

For isolation of human peripheral blood pDCs, blood buffy coats of healthy donors (HD) were obtained from the Blood Center of Policlinico Umberto I, Rome, IT. For PBMC separation we used Ficoll centrifugation and pDCs were purified as described³ by using Diamond Plasmacytoid Dendritic Cell Isolation Kit (Miltenyi Biotec). Cell purity was evaluated by staining the cells with the following antibodies: Isotype control (ISO) FITC BD 556649, Lineage FITC, eBioscence 22777872; ISO APC, Invitrogen 17471482; ISO percp cy5.5, BD550795; anti-HLA DR percp cy5.5, BD 552764; APC anti-BDCA2, BD 566428, followed by flow cytometry acquisition, and was ≥ 97% (Fig. S9). FITC-conjugated antibodies were used 1 to 10 diluition and the others at 1 to 25 dilutions. Purified pDCs were seeded into 96-well round-bottom plates at 300 × 10³ cells per ml. CXCL4, or the other peptides, were premixed with total huDNA (15 µg/ml) and added to the pDC cultures after 15-min incubation at room temperature. Stimulation of pDCs was also performed with plasma (1:50 dilution, in culture media) from SSc patients. We stimulated with plasma in the presence of the following antibodies: neutralizing rabbit polyclonal anti-human-PF4 at 3 µg/ml (Abcam; ab9561) or a control antibody directed to human myeloperoxidase (MPO, Abcam, rat-anti-human MPO; ab300651) at 3 µg/ml, or, as further control, polyclonal rabbit IgG control (Rabbit polyclonal IgG control (Southern Biotech, cat. No. 0111-14, used at 3 µg/ml). All these antibodies are azide free. Antibodies were added to the plasma 15 min before adding them to pDCs. In some experiments, the A3101 inhibitor was added to the plasma prior the incubation of the plasma of pDCs or prior the formation of the complex between CXCL4 or the other peptides and DNA.

ELISA assays for IFN-α production

Supernatants of stimulated pDCs were collected after overnight culture and tested by using ELISA kits for detection of human IFN-α (MabTech, Sweden).

Measurement of CXCL4–DNA complexes

CXCL4–DNA complexes were identified using a modified methods previously published³. As capturing antibody for CXCL4, we used the rabbit anti-human CXCL4 antibody (abcam ab9561) at 2 µg/ml was. We coated 96-flat well plates (100 µl) overnight at 4 °C. After blocking in PBS 2% BSA (200 µl), plasma (100 µl diluted to 1:100 in PBS) was added and incubated for 2 h at room temperature (RT). After incubation, wells were washed three times with 200 µl of 0.1% Tween 20 in PBS, and we added the anti-DNA antibody (mouse anti-dsDNA, Novus NB 100-662) at 1:300 dilution in PBS 1X. After 1-h incubation, we washed three times with 0.1% Tween in PBS. After washing, we added the HRP-conjugated anti-mouse (Sigma SAB3701214-1MF IN GOAT) at 1:1000 dilution for 1 h at room temperature. After four washings, as above, the chromogenic substrate 3.3’, 5.5’-tetramethylbenzidine (TMB) was added and incubated in the dark. The absorbance was measured at 450 nm after stopping the reaction by 2 N HCl. Plasma were considered positive when OD was above an established cutoff, which was calculated as the mean plus two times the standard deviation of OD values obtained with HD plasma.

Electrophoretic mobility shift assays (EMSA)

EMSA with human DNA was performed by mixing various μM concentrations of peptides with appropriate concentration of DNA (the same that induced IFN-α production in pDCs). DNA alone and the mixtures were run on a 2% agarose gel to evidence the delay in migration of the DNA due to the binding to the tested proteins. DNA on the gel was visualized by coloring with SYBR Safe DNA gel staining (ThermoFisher Scientific). For analysis of binding of DNA in the presence of A3101, the same dose of CXCL4 was premixed with different concentration of A3101 (corresponding to 50, 100, 200 and 400 μM) before adding 1 μg of human genomic DNA in each well of the 10% agarose gel.

Cytotoxic assay

PDC were cultured in the presence of IL-3 or were stimulated by SSc plasma (1:50) in the presence/absence of inhibitors for 24 h. The possible toxicity was assess using the culture supernatants harvested after 24 h culture and assessed for LDH expression by using the LDH-Cytotoxicity Assay Kit II ab65393 by Abcam, according to the manufacturer’s instructions. The cytotoxic effects was calculated with the following algorithm: test sample minus low control/high control minus low control. The LDH positive control in our tests was 85% + 5 (percent of toxicity with the highest standard of the LDH expression).

Western blot analysis

Plasma samples were diluted in RIPA buffer 1X (Tris HCl pH7.4, 25 mM; 150 mM NaCl; 1.0%Triton X-100; 0.1% SDS; 1% sodium-deoxycholate; Protease Inhibitor Cocktail), dissolved in Tricine SDS Sample Buffer (Novex) and heated for 10 min at 90 °C. Proteins of plasma simples were separated on 10–20% tricine/SDS-PAGE (Novex, Invitrogen by Thermo Fisher Scientific). For each lane were run 20 μl of the sample buffer, or 8 μl of molecular weight markers (Spectra Multicolor Low Range Protein Ladder, Thermo Fisher Scientific). The proteins were transferred at 4 °C to nitrocellulose membranes at 70 V, and the membranes were blocked at room temperature in 3% BSA. Membrane were then incubated overnight at 4 °C with the polyclonal anti-CXCL4 antibodies (Abcam). The immunoreactive bands were detected by chemiluminescence, coupled to peroxidase activity (ECL kit, Thermo Scientific, Rockford, IL, USA) and imaged with a Bio-Rad ChemiDoc XRS system (Bio-Rad Laboratories Inc., Hercules, CA, USA).

Photo-induced cross-linking of unmodified proteins (PICUP) and tricine SDS-PAGE gel characterization

Aiming to obtain an electrophoretic characterization (Tricine/SDS-PAGE) of the aggregation state of CXCL4, and CXCL4-COOH molecules, samples were stabilized in order to prevent molecules misfolding in a reducing environment. We therefore followed a previous protocol⁵⁶. Briefly, for each sample, we prepared a 40 μl volume containing 10 µg molecules (CXCL4, CXCL4-L1 and CXCL4-COOH), 50 μM Tris (2,2 bipyridyl) dichlororuthenium (II) hexahydrate (RUBPY) and 1 mM ammonium persulfate (SIGMA). Cross-linking reaction occurred irradiating samples for 2 s with a 100 W white lamp in a dark room and reaction was quickly quenched adding 20 μl of reducing sample buffer containing 5% β-MeOH; the samples were heated at 95 °C for 5 min and were analysed by 10-20% Tricine/SDS-PAGE (Novex, Invitrogen by Thermo Fisher Scientific). For each lane were run 40 μl of the sample buffer, or 8 μl of molecular weight markers (Spectra Multicolor Low Range Protein Ladder, Thermo Fisher Scientific). Then, the gels were stained by silver procedure (Pierce Silver Stain for Mass Spectrometry, Thermo Fisher Scientific).

SAXS methods

For SAXS experiments, human DNA with different degrees of fragmentation was ethanol precipitated and suspended in physiological buffer (140 mM NaCl + 10 mM HEPES, pH 7.4) to 5 mg/mL. DNA was used fragmented by sonication (as described above). Self-assembled protein–DNA complexes were formed by incubating the CXCL4 or CXCL4-47-70 or CXCL4-L1-47-70 with DNA at isoelectric peptide-to-DNA charge ratios (P/DNA  =  1/1) in micro centrifuge tubes, as shown in prior work^3,6,9. Complexes were vortexed at low speeds (900 RPM) for 1 h or until strong precipitates formed. After thorough mixing and centrifugation, precipitated complexes were transferred to 1.5-mm quartz capillaries (Hilgenberg GmbH, Mark-tubes) and hermetically sealed using an oxygen torch. The structures of CXCL4–DNA and the related complexes were solved using SAXS. Experiments were performed at the Stanford Synchrotron Radiation Lightsource (SSRL, Beamline 4-2) using monochromatic X-rays with an energy of 9 keV. A Rayonix MX225-HE detector (pixel size 73.2 μm) was used to measure the scattered radiation. Independent identical samples were prepared and measured over multiple separate experiments (n  ≥  3). 2D powder diffraction⁵⁷ patterns were azimuthally integrated using the Nika 1.76 package for Igor Pro 7.04 and FIT2D^58,59. SAXS data were analyzed by plotting scattering intensity I(q) against the momentum transfer q in Mathematica.

Structures of the peptide-DNA complexes were solved by measuring the q-positions of the first peak position. Peak positions were accurately measured by fitting diffraction peaks to a squared-Lorentzian. SAXS patterns with a single major peak at position q1 corresponds to a liquid crystalline DNA structure with short-range order. The lattice parameter d indicates the inter-DNA spacing between parallel DNA columns and is calculated from the first peak position by the formula q1 = 2π/d^3,6,60.

Statistics and reproducibility

Differences between mean values were assessed by Student’s t test for unpaired samples or, more often, by Mann–Whitney and Wilcoxon’s test (for sample size equal or higher than 5), to compare the effects of different complexes with DNA or SSc plasma, on purified pDCs or effects of inhibitors. Statistical significance was set at P  ≤  0.05. Replicates were referring to experiments performed with different donors (either plasma or buffy-coats).

Reproducibility of the data is ensured by several experiments performed with different donors (buffy-coats), to overcome the problem of human donor immune cells variability of the response. Sample size of patients was higher than 70–80 to ensure differences with HD in non-parametric tests.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

G.O., A.M., R.P., R.L., and L.F. performed experiments on pDCs and ELISA tests. L.G. performed Western blot analysis. I.P. performed EMSA. V.R., K.S. and F.R.S. provided SSc and SLE patients plasma and record of their clinical data. E.Y.L. and G.C.L.W. directed and performed experiments regarding SAXS analysis. L.F. wrote the original draft with contribution by G.C.G.W. and R.L. All authors contributed in reviewing text and Figures. L.F. conceived the research, acquired funding, directed and supervised this study.

Peer review

Peer review information

Communications Biology thanks Douglas P. Dyer and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Laura Rodríguez Pérez. A peer review file is available.

Data availability

Source data behind the graphs can be found in the Supplementary data set 1–5.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10176-1.