The study of protein–protein interactions (PPIs) is crucial for understanding how proteins function within a cell. PPIs underlie a wide range of biological processes, from cellular organization and immunological response to metabolic and developmental control, and have become a cornerstone of system biology (Rao et al., 2014; Akbarzadeh et al., 2024). Today, several approaches for detecting PPIs are routinely performed both biochemically and in cellular models. For living cells, some of the commonly implemented approaches include the yeast two-hybrid (Y2H) system, fluorescence-resonance energy transfer (FRET), bioluminescence resonance energy transfer (BRET), the synthetic lethality method, and APEX proximity labeling (Wu et al., 2007; Lobingier et al., 2017). There are also numerous biochemical approaches, such as tandem affinity purification, affinity chromatography, surface plasmon resonance (SPR), protein microarrays, dot blots, co-immunoprecipitation (co-IP), pull-down assays, protein fragment complementation, phage display, X-ray crystallography, and NMR spectroscopy (Wu et al., 2007; Rao et al., 2014).

The advent of these approaches has facilitated the study of PPIs for a wide range of proteins. However, the study of PPIs in cytoskeletal proteins remains a major challenge. Cytoskeletal proteins are present in all three domains of life and involved in various functions, including cell shape, division, movement, apoptosis, differentiation, and signaling (Ramaekers and Bosman, 2004; Wickstead and Gull, 2011). In humans, many diseases have been associated with abnormalities in cytoskeletal proteins, including cardiovascular disease, neurodegeneration, cancer, liver cirrhosis, pulmonary fibrosis, and skin disorders (Ramaekers and Bosman, 2004). The application of PPI tools in the study of the cytoskeleton is limited by some of the intrinsic properties of the proteins that comprise it (Table 1), such as their tendency to form highly stable, insoluble structures that aggregate with many proteins, hindering the identification of direct binding partners (Choi et al., 2014).

Western blots (WBs), first developed in the 1970s, are now among the most widely used assays in biological research for quantifying individual proteins of interest. Far Western blots (FWBs), developed in the 1980s, are an adaptation of WBs for the quantification of PPIs (Wu et al., 2007). FWBs are particularly useful when working with structural proteins, where proximity ligation or other cellular assays can be uninformative, due to individual proteins being tethered to the highly integrated cytoskeletal proteome (Fletcher and Mullins, 2010).

Both WBs and FWBs begin with the loading of protein samples (that have been quantified; see “Protein purification and sample preparation” below) and running them on an SDS-PAGE gel for size separation. The samples are then transferred to a membrane (e.g., nitrocellulose or polyvinylidene fluoride (PVDF); see “II. Transfer: step 8”). In WBs, a primary antibody recognizes the protein of interest on the membrane. In contrast, in FWBs, membrane-bound “prey” proteins denatured during SDS-PAGE separation are detected, and the same blot is then processed to renature the adherent proteins. This renaturation step ensures that the proteins adopt their native conformations, which are essential for bait–prey protein interactions (see “Denaturing/renaturing of prey proteins”). After renaturation, the blot is incubated with the “bait” protein, which is subsequently detected with an antibody against the bait (Figure 1). After application of the secondary antibody, the resultant FWB band reflects the binding of the bait protein to the prey protein on the membrane.

Importantly, the strength of the bait–prey binding affects the intensity of the resultant band. One specific application of FWBs is determining how a mutation in a protein of interest (prey) affects its affinity for a binding partner (bait). In this case, mutant and wild-type proteins are run on separate lanes in the same gel. After transfer, the membrane is incubated with the bait protein, allowing for competitive binding between the mutant and the wild-type prey proteins. The resulting band intensities in the mutant vs. wild-type lanes reflect the relative prey–bait binding affinity. As with other PPI approaches mentioned above, FWBs are limited by the strength of the interaction between the prey and bait proteins, generally requiring moderate to strong interactions. Studying weak or transient protein interactions may require more sensitive approaches, such as confocal scanning (PPI-CONA), a mechanically transduced immunosorbent assay (METRIS), fluorescence polarization (FP), or massively parallel sequencing (MP3-seq) (Du, 2015; Petell et al., 2021; Baryshev et al., 2024; Koszela et al., 2024).

This protocol has been adapted to quantify the relative binding strength of interaction for challenging proteins such as cytoskeletal proteins. The incorporation of a WB on the same membrane later used for a FWB increases the sensitivity of the assay beyond that offered by traditional membrane staining, expanding the application of FWBs to proteins that are not commercially available.

To illustrate the application of far Western blotting, we present results from a binding affinity comparison between the wild-type (WT) keratin 32 (KRT32^WT) and a mutant protein containing a missense variant p.Thr99Ile (KRT32^T99I), to keratin 82 (KRT82). The variant p.Thr99Ile was identified as a likely pathogenic heterozygous variant in a large kindred with loose anagen hair syndrome (LAHS) in successive generations (autosomal dominant), and it segregated appropriately with the phenotype. Keratin 32 is a type I hair keratin expressed in the cuticle layer of the hair shaft, known to heterodimerize with type II hair keratin 82 (Melo et al., 2025).

Keratins are cytoskeletal proteins, predominantly expressed in epithelial cells, that are essential for mechanical stability, cellular integrity, and intracellular signaling. Pathogenic variants in keratin genes have been implicated in several human diseases, including skin and liver disorders (Toivola et al., 2015). There are 54 human keratins, which are further divided into type I and type II (Moll et al., 2008). High sequence homology between keratins prevents the development of antibodies specific enough to discriminate between different keratins. Type I and type II keratins form heterodimers that are precursors to intermediate filaments (IFs) (Jacob et al., 2018). Keratins are extremely promiscuous, with type I proteins able to bind to any type II and vice versa (Hatzfeld and Franke, 1985). These properties, along with the high frequency of disulfide bonds, often result in type I and type II keratins co-purifying from cell lysates. Thus far, in vitro PPI studies have been limited by the unavailability of specific antibodies for many keratins and the difficulty of purifying these proteins (Table 1).

For this study, KRT32^WT and KRT32^T99I were used as prey proteins, while KRT82 was used as bait. HEK293 cells were transfected with the c-Myc-KRT32^WT, c-Myc-KRT32^T99I, or HA-KRT82^WT plasmid. The tagged proteins were purified from the lysates using anti-c-Myc magnetic beads (Pierce™ anti-c-Myc magnetic beads, Thermo Fisher Scientific) or anti-HA magnetic beads (Pierce™ anti-HA magnetic beads, Thermo Fisher Scientific) for KRT32 and KRT82, respectively. Purified samples underwent Western blotting to verify the presence of the protein of interest in each sample, as confirmed by molecular weight. Incorporating the c-Myc and HA tags in the plasmid design allowed for the use of precoated magnetic beads in the purification step and highly specific anti-c-Myc and anti-HA primary antibodies.

Developed membrane images showed similar loading of the c-Myc-KRT32 prey proteins (∼60 kDa), as revealed with an anti-c-Myc WB (Figure 3A). This was followed by complete stripping of the secondary antibody for anti-c-Myc (Figure 3B). Next, the membrane was incubated with purified HA-KRT82, and an anti-HA primary antibody was applied. After applying a secondary antibody for anti-HA, the resulting FWB membrane showed bands at ∼60 kDa that reflect binding between KRT32 and KRT82. This band is substantially weaker for KRT32^T99I than for KRT32^WT (Figure 3C), revealing that the threonine-to-isoleucine substitution at amino acid 99 in keratin 32 lowers its binding affinity for keratin 82. Four biological replicates of this assay were performed, showing a 20.7% reduction in band intensity for the c-Myc-KRT32^T99I compared to the c-Myc-KRT32^WT band (Melo et al., 2025). Bands were quantified using the “analyze > gels > plot lanes” function in ImageJ, which allows for band selection and signal quantification. Bands from bait probing (Figure 3C) were normalized to their respective bands from prey probing (Figure 3A), and the KRT32^WT and KRT32^T99I lanes were compared (Table 5).

Several hurdles were encountered during the adaptation of this protocol. Most issues occurred in the initial steps of the assay (protein purification through membrane stripping). Table 6 includes some of the most common issues, potential causes, and solutions.

While various techniques have advanced PPI research, cytoskeletal proteins pose unique challenges to the use of many PPI research methods due to their tendency to form large, stable, and insoluble aggregates. This study highlights the potential of FWB as an effective tool for studying PPIs involving cytoskeletal proteins. FWB is particularly useful for studying structural proteins that are difficult to analyze using other methods, offering a streamlined approach by detecting binding partners based on molecular weight. This is especially valuable when the cellular context or fixed intracellular location limits other techniques.

Additionally, FWB is useful for studying the effect of gene variants on protein interactions, providing insights into diseases caused by disruptions in cytoskeletal protein interactions. By comparing wild-type and variant proteins, as illustrated with the keratin case study, FWB helps researchers understand how subtle changes in protein structure can contribute to disease phenotypes.

In conclusion, FWB provides a powerful and broadly applicable approach for PPI analysis in laboratories with standard biochemical capabilities. Beyond its utility for cytoskeletal proteins, it is broadly applicable to structural or insoluble proteins and to proteins lacking specific antibodies. As such, FWB represents a valuable addition to the biochemical research toolbox for any laboratory studying protein biology.