WT human CXCL12 was bacterially expressed using a codon\optimized cDNA (Genscript, Piscataway, NJ, USA) as a thioredoxin\His\tagged fusion protein from the pET\32(+) vector with an enterokinase cleavage site at the N\terminus. exploration of the chemokine/galectin interactome is warranted. value as obtained (Appendix?Table?S1). On the other hand, N160 lies on PD173955 the opposing, non\interacting S\face in Gal\3 and is known to contribute to carbohydrate binding (Fig?2F, right panel). For empirical validation of our model, we used site\directed mutagenesis to produce several mutants and assess effects on heterodimer formation. Q220E in NMR and N160A in ligand blots and SPR were used as controls. Three Gal\3 RICTOR CRD mutants (Q220E, Q220K, and H217A) were selected to assess their effects on HSQC spectra of 15N\labeled CXCL12 when examining mixtures. Even though all HSQC spectra look highly similar, analysis of the data could reveal distinct differences. Figure?EV3 shows chemical shifts of 15N\labeled CXCL12 with each of these Gal\3 CRD mutants. These maps show the same trends as observed with WT Gal\3 CRD. PD173955 Although this indicates that WT Gal\3 CRD and its mutants interact with CXCL12 in the same way, the magnitudes of ? changes are different. Compared to WT Gal\3, Q220E ? values are slightly PD173955 increased (Fig?EV3A), whereas those for Q220K and H217A are decreased (Fig?EV3B and C). CXCL12 sequence\averaged ? values are 0.0061?ppm for WT Gal\3 CRD, 0.0073 for Q220E, 0.0036 for Q220K, and 0.0048 for H217A. Smaller chemical shift changes usually indicate weaker intermolecular interactions 39. Here, average ? values suggest slightly stronger binding between CXCL12 and Q220E, and weaker binding between CXCL12 and Q220K and H217A. These trends parallel those observed in our MD\based free energy calculations, which yielded ?values of ?50?kcal/mol for WT Gal\3 CRD, ?58?kcal/mol for Q220E, ?31?kcal/mol for Q220K, and ?38?kcal/mol for H217A (Appendix?Table?S1). Open in a separate window Figure EV3 1HC15N chemical shift maps for 15N\labeled CXCL12 with WT Gal\3 CRD and its mutants ACC Values plotted vs. the amino acid sequence of CXCL12 are shown for 30?M 15N\enriched CXCL12 in the presence of 500?M label\free Gal\3 CRD mutants (A) Q220E, (B) Q220K, and (C) H217A (black). The results of the experiment with WT Gal\3 CRD from Fig?2A are overlaid PD173955 in gray. Densitometric analysis of CXCL12 binding to variants of Gal\3 (Appendix?Fig S7A) and Gal\3 CRD (Appendix?Fig S7B) demonstrates that residues N222 and E185 are indeed involved in the interaction with CXCL12, whereas N160 is not. Similarly, the affinity of CXCL12 injected over sensor chips with immobilized Gal\3 mutants R168A, E185A, H217A, and Q220K was reduced, whereas the affinity of N160A was not (Appendix?Fig S8ACH). In addition, Gal\3 mutant binding to the N\glycans of a common galectin binder, i.e., the glycoprotein asialofetuin (ASF), was only impaired in the case of N160A that showed no significant effect on heterodimer formation (Appendix?Fig S8ICL), which supports our findings that CXCL12/Gal\3 heterodimer formation is not significantly affected by glycan binding (Appendix?Fig S1A and B). Gal\3\mediated inhibition of CXCL12\induced leukocyte migration We next investigated the functional consequences of CXCL12:Gal\3 heterodimerization. Initially, we examined whether Gal\3 affects CXCL12\induced migration of Jurkat T cells, and discovered that both Gal\3 and Gal\3 CRD inhibited chemotaxis in a dose\dependent manner (Fig?3A and B). In contrast, Gal\1 only inhibited migration at 1?M (Fig?3C). These results with Jurkat cells were replicated using primary cells, i.e., activated human CD4+ T cells (Fig?3D). PD173955 A bell\shaped chemotaxis curve was observed upon increasing the concentration of CXCL12, with the height of the curve being significantly reduced in the presence of 0.1?nM Gal\3 CRD (Fig?3 E and F). Consistent with our concept, the CXCL12 chemotaxis curve with Gal\1 remained unchanged even in the presence of 100?nM.