Fusion proteins (FP) are a group of proteins in which two or more protein domains are integrated into a single-chain molecule (Chen et al., 2013, Yu et al., 2015). They can occur naturally but are often designed for biotechnological applications such as biosensors, enzymatic engineering, protein purification, and medical treatments (Aroul-Selvam et al., 2004, Czajkowsky et al., 2012, Kobe et al., 2015, Schmidt, 2009, Yu et al., 2015). These FPs are composed of a protein of interest fused with a tag, which can be a peptide or, in the case of this study, an entire protein (Fig. 1). The addition of the tag increases the expression and stability, improves the solubility, and may protect the protein of interest from proteolysis (Terpe, 2003). Also, some tags facilitate purification allowing affinity chromatography, reducing the purification steps, and obtaining purities of up to 90 % (Smyth et al., 2003).

Protein crystallography is a field where protein purification is crucial, and the use of FP before crystallization improves protein production and allows to increase its purity (Chayen and Saridakis, 2008, Corsini et al., 2008, Derewenda, 2004, Holcomb et al., 2017, Krauss et al., 2013, McPherson and Gavira, 2014). It is also common to remove the tag before proceeding with the crystallization (Derewenda, 2004, Terpe, 2003). The impact of leaving the tag for crystallization is controversial. While some authors state that the tag can affect the crystallographic structure, or even the capacity to obtain crystals, others argue that tags influence is insignificant (Holcomb et al., 2017). It has been also proposed that a tag, with a non-defined three-dimensional structure, such as the histidine tag (HisTag), reduces the possibility of obtaining crystals increasing the sample heterogeneity (Kobe et al., 2015). There have also been reports where HisTag removal favors crystal formation, increase resolution, and reduce mosaicity (Bucher et al., 2002, Huh et al., 2014, Kim et al., 2001, Sugawara et al., 2009). Furthermore, there are various reasons for not removing a tag; the Protein Data Bank (PDB) has hundreds of examples in which the HisTag is present in the construction. Still, the HisTag is not visible in the electron density map nor the fold of the protein of interest is affected (Burley et al., 2021; Carson et al., 2007). Other authors have shown the utility of maintaining protein tags for crystallization, mainly in proteins of interest difficult to crystallize (Wild and Hothorn, 2017). In several cases, authors declare that the tag presence facilitates the crystallization (Holcomb et al., 2017, Kobe et al., 2015, Momin et al., 2019, Waugh, 2016). Additionally, there are examples were the presence of protein tags in a FP does not affect the protein of interest structure (Waugh, 2016). FPs were also useful in determining protein structures by expanding the phases from the known to the unknown segment of a complex (Corsini et al., 2008, Kobe et al., 2015, Liu et al., 2003).

Determining the phases of a protein is a significant challenge in crystallography. The choice of the phase-determining method is key to the structural determination success and the selection depends on the available structural information of the protein, its amino acid sequence, and other biochemical characteristics (Smyth and Martin, 2000, Wlodawer et al., 2013). In general, phases can be obtained through molecular replacement (MR), isomorphous replacement (MIR, SIR), anomalous scattering (MAD, SAD), or a combination of these methods (Taylor, 2010). In certain situations, partial phases are used as a starting point for determining a crystallographic structure, and more recently also in MR (Hauptman, 1997, McCoy, 2007). This approach was used in several FP structures, determined by expanding the phases from the protein tag to the protein of interest (Smyth et al., 2003, Waugh, 2016). In these scenarios, manual model construction has been performed until a complete model was obtained (Wild and Hothorn, 2017). Additionally, in those cases, software such as PHENIX AutoBuild (Terwilliger et al., 2007) and Buccaneer (Cowtan, 2006), and ARP/wARP (Perrakis et al., 2001) have been used for automated model building (AMB) (Langer et al., 2008).

In this study, the PDB was mined to obtain an up-to-date list of the FP crystallographic structures some of the most used protein tags: maltose binding protein (MBP), thioredoxin (TRX), glutathione transferase (GST), green fluorescent protein (GFP) and the small ubiquitin-like modifier protein (SUMO). Among them, 116 PDB entries were chosen to evaluate the effectiveness of using partial MR with the protein tag, followed by AMB to improve phases and construct the protein of interest. Our purpose is to determine statistically the characteristics needed to successfully expand phases from the protein tag to the protein of interest in a FP using AMB.