Mechanisms and biotechnological applications of transcription factors

Abstract

Transcription factors play an indispensable role in maintaining cellular viability and finely regulating complex internal metabolic networks. These crucial bioactive functions rely on their ability to respond to effectors and concurrently interact with binding sites. Recent advancements have brought innovative insights into the understanding of transcription factors. In this review, we comprehensively summarize the mechanisms by which transcription factors carry out their functions, along with calculation and experimental-based methods employed in their identification. Additionally, we highlight recent achievements in the application of transcription factors in various biotechnological fields, including cell engineering, human health, and biomanufacturing. Finally, the current limitations of research and provide prospects for future investigations are discussed. This review will provide enlightening theoretical guidance for transcription factors engineering.

Graphical abstract

1. Introduction

Transcription is a crucial component of the central dogma of molecular biology (DNA-RNA-protein) [1], serving as a bridge that translates genetic information into diverse forms at the individual level. Transcription factors (TFs) play a pivotal role in regulating the transcription of target genes by selectively recognizing and binding specific DNA regions known as TF binding sites (TFBSs), typically located in the promoter region [2]. This mechanism allows for the differential expression of genes throughout the genome. TFs can exert control over processes that specify the developmental patterns of plant and animal cells, thereby driving cell differentiation [3]. In the case of microbial cells, TFs play a systematic role in regulating the finely tuned expression of numerous genes within the metabolic network, ensuring efficient growth and reproduction [4]. In terms of its impact on human health, dysfunction of TFs often results in misregulated gene expression [5,6], leading to various diseases and syndromes that pose a risk to human well-being [7]. Consequently, a comprehensive understanding of TFs and their effective utilization holds significant potential for advancing both the production of high-value bio-products and human healthcare.

The study of TFs has been a subject of considerable interest, and there is a desire to transition from relative unfamiliarity to their efficient application [8]. The general methodology employed to achieve this goal can be succinctly summarized as follows: Firstly, TFs are screened and their fundamental characteristics and functions, including subcellular localization and interacting proteins [9,10], are comprehensively elucidated [11]. Subsequently, based on this valuable information, potential application domains are explored where TFs can be efficiently customized to exert high biological functionality, such as contributing to human health or biomanufacturing [12]. In the subsequent development, it is imperative to focus on the development of novel technologies for identifying and characterizing TFs. Moreover, a broader range of cross-cutting areas in biology should be explored and adapted to fully unleash the competitive potential of TFs as a versatile tool.

Although TFs can control the transcription of target genes through various mechanisms [13], these mechanisms fundamentally rely on their interactions with TFBSs. Therefore, the identification of TFBSs is a crucial research concern. In order to fulfil this task, many highly innovative techniques have been rapidly developed. These include both specific experimental [[14], [15], [16], [17]] and calculation methods [[18], [19], [20]]. The latter can be described in more detail as a bioinformatics-based strategy for predicting TFBSs. This is accomplished by integrating a vast amount of experimental data into databases and employing big data algorithms for analysis and calculation. These two strategies can collaborate to facilitate the identification of TFs and TFBSs [21], particularly with significant improvements in accuracy and throughput.

Recently, there has been no systematic review on the universal mechanisms of TFs, including the recent achievements in the field of TFs and TFBSs identification. Additionally, the potential contribution of TFs to biotechnological applications has not been highlighted. In this review, these concerns are systematically summarized and discussed. The objective is to provide a comprehensive perspective on TFs research to meet the current research needs and inspire valuable insights for future research directions.

2. Transcription initiation

The transcription of genes is catalyzed by the holo-RNA polymerase (RNAP), which contains five subunits (α₂ββ’ωσ) (Fig. 1) [13]. Among these components, α₂ββ’ω plays a catalytic role, facilitating enzymatic activity, while the σ factor is responsible for recognizing promoters and directly binding to its conserved −10 and −35 regions [22].

Fig. 1.

Simplified schematic overview of the transcriptional complex reconstruction at 3.9 Å resolution and class I/II transcription activation mechanisms.

Taking the transcription of sugar metabolism-related genes in bacteria as an example, transcription initiation requires the assistance of cyclic adenosine 3′,5′-monophosphate (cAMP) and its receptor, catabolic gene activator protein (CAP). The active CAP binds to the promoter DNA upstream of the RNAP to activate transcription [23,24]. CAP-dependent promoters can be classified into two activation models based on the different binding positions of CAP on the promoter. In the case of class I, CAP binds to the promoter DNA located in the −61.5 region. Transcription activation can be achieved through the “recruitment” mechanism, whereby CAP interacts with the C-terminal domain of the α subunit (αCTD) of RNAP [25]. In the case of class II, the CAP binding site is situated at the −41 region, which coinciding with the −35 region of the promoter. CAP interacts with the σ subunit of RNAP, inducing conformational changes in RNAP and facilitating the process of transcription activation [26]. Recently, the structures of both class transcriptional complexes were successfully resolved using high-resolution cryo-electron microscopy. As a result, several authentic microscopic phenomena have been accurately depicted, such as the spatial conformation of the opened promoter DNA [27], the bending angle of the DNA strand in the transcriptional complex and the interacting residues in the CAP-αCTD region [28,29].

TFs actively engage in the transcriptional complex, exerting their regulatory influence on transcription intensity by either activating or repressing it.

3. TFs control the transcription

TFs usually contain at least two core structural domains, namely the DNA binding domain (DBD) and the effector domain (ED) (Fig. 2a) [21]. The DBD, which often includes helix-turn-helix (HTH), helix-loop-helix, zinc finger or leucine zipper motifs [30], is primarily responsible for specifically recognizing and binding to their respective TFBSs. On the other hand, the ED serves as the regulatory domain involved in signal sensing. It can bind various intracellular metabolites, including: CoA [31,32], cofactors [NADP(H), NAD(H) [33]], sugar metabolites [pyruvate [34], glucosamine6-phosphate [35], fructose-1,6-diphosphate [36]], amino acids [lysine [37]]. Additionally, changes in the external environment (Fig. 2b), such as pH [38], temperature [39], light [40], dissolved gas [[41], [42], [43]] or cell density [[44], [45], [46]], can also act as signals for induction. The induction signal can trigger the binding or dissociation of a TF to a specific DNA region. Since this process mainly occurs in the promoter region, it often impacts the efficiency of RNAP in carrying out its transcriptional function. In other words, the TF, consisting of two structural domains with differentiated functions, forms a unified entity that cooperatively regulates gene transcription by responding to intra- or extracellular signals.

Fig. 2.

Work mechanism of TFs. A Structure of TFs. Effector domain receives the corresponding signal that allows the TF to bind or dissociate from the binding site located at the promoter of its regulon gene. B Some examples of signaling molecules. C Effectors contribute to the function of TFs. D Mechanisms of activation and repression of regulated genes by TFs.

TFs can be classified into multiple families based on their unique structures and functions. Table 1 provides examples of TF families along with their typical characteristics. These TFs regulate various pathways within the metabolic network and can act as preferential repressors or activators, or even display dual regulatory roles. Importantly, the presence of distinct TFBSs for each family ensures a high level of orthogonality in regulating their respective pathways, preventing random cross-interference.

Table 1.

Some TFs families.

Family	Example	Action	Some regulated functions	Characteristics of TFBSs	DBD position	Reference(s)
TetR	MexZ, QacR, AcrR.	Repressor	Biosynthesis of antibiotics, efflux pumps, osmotic stress, etc.	Inverting palindrome sequences. e.g. 5′TACATACATTTGTGAATGTATGTA -3’.	N-terminal	[105,106]
GntR	FadR, McbR, GabR.	Repressor	General metabolism	Inverted or direct repeat sequences. e.g. 5′TNG(N)nCNA-3′.	N-terminal	[107]
LysR		Activator/repressor	Carbon and nitrogen metabolism	Interrupted palindrome sequences. e.g. 5′ATC-N9-GAT-3′.	N-terminal	[108]
AraC		Activator	Carbon metabolism, stress response and pathogenesis	Asymmetrical and highly AT-rich sequences. e.g. 5′GATATAAN3TAN-3′.	C-terminal	[109,110]
MerR	SoxR, BltR, BmrR.	Activator	Resistance and detoxification	Dyad symmetrical sequence. e.g. 5′ACCTCAAGTTTGCTTGAGGT-3′.	N-terminal	[111,112]
CRP	CAP, RedB, FNR.	Activator/repressor	Global responses, catabolite repression and anaerobiosis	Two inversely-repeated sequence. e.g. 5′TGTGANNNNNNTCACA-3′.	C-terminal	[113,114]

While the specific mechanisms of action may vary among different families of TFs, the fundamental nature of these signaling molecules as effectors remains consistent. Their essential role is to modulate the binding affinity between the TF and its associated TFBS DNA, either promoting binding or facilitating dissociation. Effectors achieve this objective through various pathways (Fig. 2c), including direct interactions or signal transduction cascades. For instance, effectors can directly interact with TFs, as seen with the malate-responsive TF, MalR [16], or activate activators for TFs, such as the glutamine-responsive TF, GlnR [47]. Effectors can also truncate redundant sequences within the TFs, as demonstrated by the pH-responsive TF, PacC [48,49]. Furthermore, effectors can induce oligomerization, as observed in the temperature-sensitive TF, Hsf [50], or phosphorylate TFs, like the glucose-responsive TF, CcpA [51].

4. TFs adopt a unique mode of operation

TFs are typically classified into activating and repressing types based on their functions, which have highly positive or negative effects on transcription initiation, respectively. Considering the dual-functional impact of effectors on the interaction between TFs and their TFBSs (promoting binding or dissociation), Fig. 2d focuses on the “promoting binding” to illustrate the relevant mechanism. In the case of transcriptional activation, this can be achieved through the recruitment of RNAP by TFs. It can also occur due to the improved spatial conformational adaptation of the promoter's double-stranded DNA to RNAP resulting from TFs binding [52]. On the other hand, transcriptional repression seems to have more diverse mechanisms compared to activation. These mechanisms include: (1) Steric hindrance. The TFBS overlaps with the RNAP-binding site, creating direct competition between the TF and RNA polymerase for the cis-acting locus (the conserved −10 or −35 promoter element) [53]. (2) Roadblock. The binding of a TF and its TFBS located downstream of the transcription initiation region forms a roadblock for transcription elongation, hindering the progression of RNAP [54]. (3) Deformation. Repression can occur when the repressor simultaneously binds to motifs located upstream and downstream of the promoter, leading to significant deformation of the promoter DNA. This deformation can cause a mismatch between the promoter and RNAP, inhibiting transcription [55]. (4) Anti-activation. Repression can be achieved through the TF acting as an “anti-activator”. This occurs when the repressor competes with an activator due to the overlap of their TFBSs or interacts with the activator, thereby inhibiting transcription [56]. (5) Difficult promoter clearance. Repression may occur by preventing promoter clearance due to the interaction of the TF with RNAP, hindering the initiation of transcription [57].

TFs typically function as homodimers or multimers rather than in a monomeric state, which is a distinct characteristic of their mode of operation. This can perhaps be attributed to the fact that the DNA sequences recognized by TF monomers do not contain enough information, making it challenging to precisely extract regulated genes from the vast background of randomly occurring sequences in the genome. In contrast, oligomerized TFs can greatly enhance the affinity and specificity of transcriptional regulatory binding, as the cis-regulatory sequences they recognize are palindromic repeats with double the length. This may represent the most effective strategy evolution has retained to achieve precise regulation by multiple TFs for multiple genes with a high level of orthogonality. It is worth noting that TFs can also function as monomers to regulate target genes. For example, the TF, Gcr1^U regulates glycolytic genes in Saccharomyces cerevisiae, Spx regulates oxidative stress-related genes in Bacillus subtilis [58,59]. Additionally, there are TFs that function as tetramers, such as MYC2 derived from plants [70]. The oligomeric complex of MYC2 mediates the formation of DNA looping to enhance transcription of target genes. Considering that the vital biological functions of TFs rely on their interactions with specific TFBSs, the identification of them has emerged as a significant research topic. The notable accomplishments in this field warrant comprehensive summarization and discussion.

5. Identification of TFs and TFBSs

In general, two strategies can be employed to identify TFs and TFBSs as depicted in Fig. 3a. Each strategy has its own unique characteristics. Experiment-based strategies offer objective and reliable results but often demand a considerable amount of labor-intensive work. On the other hand, calculation-based strategies, relying solely on computer programs and algorithms, can greatly reduce manual labor but may compromise to some extent on accuracy.

Fig. 3.

Prediction of TFs and their matching TFBSs. A Strategies based on calculation and experimentation are used to identify TFs and TFBSs. B Workflow scheme for TFs and target regulon gene prediction.

5.1. Experimental methods for identification of TFs and TFBSs

Accurate identification of TFs and TFBSs is a prerequisite for their practical application in the field of biotechnology. To achieve this goal, numerous ingenious experimental methods have been developed in recent years. It is noteworthy that these experiments need to be carefully designed, taking into account the distinction between the known and the unknown entities.

5.1.1. Gene-centered experimental methods

Yeast one-hybrid is a classical method used to identify the specific TFs that transactivate a particular gene. It is also the only method that allows researchers to identify upstream regulators of a specific biological process [60]. The main objective of yeast one-hybrid can be summarized as determining which TFs specifically bind to a given TFBS.

Although this technique has proven to be valuable in problem-solving and has undergone gradual improvements to enhance its desirability, there are certain limitations to be aware of. It is possible for an endogenous yeast expression activator to bind to the bait DNA binding site and activate the reporter gene, leading to a potential omission of the corresponding target gene fragment. Additionally, the inserted target element may interact with the yeast endogenous TF, resulting in the loss of DNA binding ability in the DBD region of the target element and consequently omitting the target gene. To address this challenging issue, a refined yeast one-hybrid system can be achieved by using two independent bait sequences [61] or two reporter genes [62]. This approach efficiently suppresses the occurrence of false positives. However, the application of this technique is still limited in certain cases, such as the inability to retrieve heterodimeric TFs or the challenge of identifying low-abundance TFs in cDNA libraries. Therefore, it is necessary to further improve this technology to address these issues or develop a novel technique that is urgently needed.

5.1.2. TF-centered experimental methods

In contrast to the uniqueness of the strategies mentioned above, there are numerous alternative approaches available for identifying TFBSs of specific TFs.

5.1.2.1. Differential expression of TFs for identifying regulatory genes

Overexpression or silencing of specific TFs can induce changes in the transcript levels of their target genes, which can be observed through transcriptomic data analysis. Gene ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis can then be utilized to identify the target genes. These genes with altered transcript levels are considered to be regulated by the specific TF (Fig. 4a). Additionally, this approach can provide insights into the activation or repression functions exerted by the TF on specific genes within relevant pathways. For instance, overexpressing the TF RamA in Corynebacterium glutamicum [63], or PsrA in Serratia marcescens [12] can yield valuable information. In addition to endogenous TFs, exogenous TFs can also be expressed. For instance, Li et al. [64] overexpressed the TF KmHSF1 and KmMSN2 from thermotolerant Kluyveromyces marxianus in Saccharomyces cerevisiae, resulting in significant changes in the expression levels of 50 and 55 genes, respectively.

Fig. 4.

Strategies for identifying TFBSs. A The genes regulated are identified by comparing transcriptomic changes in specific TF upon overexpression and silencing. B Chromatin immunoprecipitation-based identification strategy. C Schematic diagram of the CUT&RUN strategy.

Although convenience and efficiency are significant advantages of this approach, there is an obvious limitation that arises from it. It becomes challenging to interpret which genes are directly controlled by the TF through its binding to the regulatory elements of the gene, and which genes are influenced indirectly through cascade effects on other regulatory proteins. In other words, the different patterns of gene expression resulting from the overexpression or lack of expression of a TF cannot distinguish between its direct and indirect effects. This ambiguity can lead to confusion regarding the sequence of events triggered by TF activity, making it difficult to accurately piece together the sequence of regulatory events involved in a specific biological process. In light of this, methods based on immunological techniques can be instrumental in identifying direct protein-DNA interactions.

5.1.2.2. Chromatin immunoprecipitation (ChIP) and its derivative technology

ChIp is the most commonly used technique to address the crucial question of where TFs bind in the genome. The standard ChIP assay typically begins with the formation of a covalent linkage between the DNA-binding protein and the bound DNA using a chemical cross-linking reaction, usually with formaldehyde. This cross-linking step is followed by shearing of the chromatin into protein-DNA complexes of a specific length, often achieved through ultrasound. The sheared chromatin is then incubated with an antibody that specifically recognizes and binds to the target protein (TF). Subsequently, the cross-links are reversed, the DNA is purified, and finally, the enriched DNA fragments are subjected to high-throughput sequencing [65]. (Fig. 4b). ChIP has proven to be an effective method for identifying potential TFBSs within the genome and has been widely utilized in various species. For instance, it has been used to discover TFBSs for a large population of TFs in yeast [66,67], identify direct targets TFBSs of the Caenorhabditis elegans TF, DAF-16 [68], and identify the STAT3's TFBSs in human B cell [69].

The advantage of the ChIP technique is that it provides clear evidence that the binding between proteins and DNA is actually occurring within the cell, since the binding takes place in the in vivo environment. Additionally, by comparing the sequencing results to the annotated whole genome, the distribution and specific location of TFBSs across the entire genome can be determined. However, while these techniques can identify TFBSs, they do not reveal the specific regulatory function performed by the TFs on target genes (activation or repression). To determine the exact regulatory function, subsequent differential expression analysis of TFs is required. This analysis is distinct from the differential expression analysis of TFs without ChIP, as it involves an independent experiment on a small set of candidate target genes identified based on the sequencing results.

To address certain limitations of ChIP, such as the requirement for a large number of starting cells for chromatin DNA purification and the lower signal-to-noise ratio resulting from sequencing whole genome fragments, several highly innovative and remarkable approaches have been developed based on the ChIP method. One such example is the Cleavage Under Targets & Release Using Nucleas (CUT&RUN) technique, which was developed by Peter et al. [70]. In this technique (Fig. 4c), a TF-specific antibody fused with a calcium-inducible MNase, this protein complex can bind to the TFBSs within intact cells and cleave the adjacent DNA adjacent to them. The cleaved DNA fragments are then carried by the protein complex and released from the cell for subsequent sequencing. This method differs from ChIP, where the entire genome is initially fragmented, resulting in background noise in the sequencing of the entire genome. In CUT&RUN, DNA fragments are only cleaved around the binding sites, effectively reducing background noise during sequencing while retaining the advantages of the in situ approach. Moreover, this method allows for the use of a lower number of starting cells and can be automated using magnetic beads. While ChIP-based techniques such as CUT&RUN are powerful tools for studying TF binding, they have limitations. One major limitation is that they are endpoint measurement-based techniques, which means they cannot capture dynamic changes in TF binding in real-time. Additionally, the construction of specific antibodies that successfully bind to TFs can be challenging and may introduce experimental variability. To overcome these limitations, non-immune-dependent techniques that do not rely on antibodies have been developed for the efficient identification of TFBSs.

5.1.2.3. Un-immunoprecipitation dependent technology based identification

The un-immunoprecipitation dependent technique offers an alternative to the challenges associated with obtaining high-quality specific antibodies, as it operates on a fundamentally different principle. One such technique, called “DamID” (DNA adenine methyltransferase identification), utilizes the recognition ability of DNA adenine methyltransferase (DAM) by fusing it with a TF [71] (Fig. 5a). In this method, when the fusion protein is expressed in vivo, DAM preferentially methylates specific base sequences (GATC) near the TFBSs. The methylated regions can extend thousands of bases away from the binding site due to the diffusion of DAM. Moreover, since the base sequence GATC occurs at a relatively high frequency in the genome, almost every region can be detected using this technique. Subsequently, the genomic DNA is digested with DpnI (only cuts methylated GATC) or DpnⅡ (only cuts no-methylated GATC). This is followed by southern blot or PCR-based quantification of methylation frequencies, enabling the mapping of TFBSs across the entire genome. DamID offers several advantages, such as the detection of TF-TFBS interactions in living cells and the ability to achieve reliable results even with low levels of expression of the Dam fusion protein. This reduces the likelihood of interference with the function of the endogenous protein or its target. However, DamID does have limitations. The method's resolution is relatively low, limited to approximately 1 kilobase (kb), which restricts its utility for precise mapping of TFBSs. Additionally, extensive methylation can be toxic to cells, posing a challenge for experiments that require extended periods of methylation.

Fig. 5.

Un-immunoprecipitation method to identify TFBSs. A DAMID strategy based on methylation of sequences adjacent to the binding sites and determination of the binding sites based on the methylation frequency. B 3D-seq strategy based on introduction of base-specific mutations adjacent to the binding sites and determination of the binding sites based on mutation frequency. C “calling card” strategy based on transposition technology.

In response to the aforementioned limitations, researchers have made significant advancements in developing innovative tools with alternative approaches. For instance, Larry et al. have introduced a novel tool called “3D-seq,” which utilizes nucleic acid-targeting deaminases. This class of proteins is capable of inducing mutations in both DNA and RNA molecules [72]. In this technique (Fig. 5b), a modified cytosine deaminase called DddA, which preferentially acts on double-stranded DNA, is fused downstream of the TF. When the TF binds to its target TFBSs, DddA induces base pair substitutions (G-C to A-T) in the vicinity of the TFBSs. Through sequencing comparisons with appropriate controls, the average frequency of these conversion events can be determined, providing valuable information about TFBSs. Additionally, the system's controllability is enhanced through the introduction of auxiliary ligands regulated by the arabinose promoter. However, it is important to acknowledge the potential limitations of this technique. The resolution of the method is dependent on the frequency of cytosines found within the context of DddA-preferred sequences. While these sequences occur, on average, once every 10 base pairs, their frequency may vary within specific regions of the genome. This variability can significantly reduce the resolution of the technique in those particular regions.

Transposon technology has also emerged as a valuable approach for TFBSs identification [73]. For instance, Wang et al. developed a novel method called the “calling card” method (Fig. 5c). The underlying principle can be summarized as follows: When a transposase is fused to a TF, it induces the integration of a transposon into the genome near the TFBSs. Following transposition, cells that have undergone transposition are selected, and their genomic DNA is extracted. The transposon-containing fragments are then cleaved near the transposon ends using a restriction enzyme, and the resulting fragments are ligated to circularize them. The genomic DNA adjacent to the transposon end is subsequently amplified through reverse PCR, and the sequence of the reverse PCR product, representing a specific genomic DNA sequence near the TFBSs, is identified through sequencing. The obtained sequences are then compared to the original genome to determine the insertion locations, thereby revealing the TFBSs locations. To summarize, the “calling card” method utilizing transposon technology allows the transposase, guided by the TF, to leave a distinct marker or “calling card” in the proximity of TFBSs within the genomic regions. This “calling card” can be identified and serves as a record of the genomic regions visited by the TF. This approach offers advantages over the limitations associated with DamID and 3D-seq techniques mentioned earlier. However, it is important to note that the transposition efficiency in this method is relatively low. This means that not all TF binding events may result in successful transposon integration and “calling card” formation. Consequently, the method may not capture the complete repertoire of TFBSs.

As evident from the given examples, it becomes apparent that there may not be a single universal solution that can address all limitations. Instead, trade-offs must be considered to select the optimal strategy based on specific requirements. While the aforementioned in vivo techniques offer valuable insights into TF regulation targets within their natural context, they do encounter challenges in achieving substantial enrichment of binding fragments in samples and conducting biophysical characterization and quantification of protein-DNA interactions. To address these limitations, relevant in vitro-based strategies such as electrophoretic mobility shift assay (EMSA) [74], surface plasmon resonance (SPR) [75], DNA microarray-based interaction technology (MITOMI) [76], or protein-binding microarrays (PBM) [15] can be utilized.

5.2. Calculation methods for prediction of TFs and TFBSs

With the rapid development of bioinformatics, molecular biology databases now offer a vast array of nucleotide and amino acid sequences. Analyzing and computing these sequences has emerged as the fastest and most effective method for discovering new genes and predicting protein structures. Among the crucially conserved regions of TFs, the DBD and ED exhibit minimal alterations during the extensive process of evolution, ensuring the accurate execution of their regulatory functions. Consequently, these conserved structural domains enable the prediction of TFs and TFBSs through bioinformatics methods that rely on sequence homology analysis [77].

Considering the molecular mechanisms through which TFs exert their regulatory functions primarily involves the interaction between TFs and their corresponding TFBSs, it becomes possible to make predictions about one based on the knowledge of the other. In other words, two strategies are employed (Fig. 3b): the TFs-centered method, which aims to predict TFBSs, and the gene-centered method, which aims to predict TFs. Despite their distinct approaches, both strategies are integrated into the same workflow model of “input-align-output”.

Given the labor-intensive nature of experimentally searching for targets across a wide range of proteomes and genomes, calculation-based validation approaches using databases have become the preferred choice. The field of bioinformatics has witnessed rapid development, leading to significant enrichment of TFs-related databases and improved accuracy of predicted results. These databases cater to specific species, such as mouse, human, plant, E. coli, and more. Furthermore, there are comprehensive databases that integrate common species across taxonomies, encompassing not only the aforementioned species but also others like Drosophila melanogaster, Caenorhabditis elegans, birds and reptiles (Table 2).

Table 2.

TFs related databases.

Species	Database
Human	hTFtarget [115]
Plant	PlantTFDB [116], PlantPAN [117]
Mouse	TFdb [118]
E. coli	ERMer [119]
Comprehensive Database	AnimalTFDB [120], JASPAR [121]

The calculation-based approach described above, while providing a reasonable level of accuracy in most cases, encounters a significant limitation when attempting to predict proteins that lack homology with the reported TFs. This limitation poses a challenge for calculation that rely on homology alignment. In essence, these databases may be unable to answer the question of whether a protein is a TF or not when it lacks homology with known TFs.

6. Applications based on the pair TFs and TFBSs

The successful identification of TFs, target gene maps, and their corresponding TFBSs not only enables a comprehensive understanding of regulatory mechanisms at a genome-wide level but also holds significant potential for various applications, including disease treatment, drug development, and microbial production. By maximizing their appropriate utilization, these findings can make substantial contributions in these fields.

6.1. Induced directed differentiation in stem cell engineering

Stem cells have an extraordinary ability to expand limitlessly and differentiate into diverse cell types. However, the lack of dependable protocols for generating cell lines and the need for lengthy time periods present challenges to their efficient application. Current protocols primarily rely on external signals, often mediated by TFs, to regulate the genetic programs that drive specific cell types. Nevertheless, this method necessitates prolonged stimulation with external signal factors, and achieving multiple lineage cell variations within a single culture is often challenging.

Thus, in theory, cell differentiation can be directly induced by activating one or a subset of TFs through genetic manipulation using molecular biology techniques, bypassing the need for specific external environmental signal factors like pharmacological agents or bioactive proteins (Fig. 6a). By directly activating specific TFs, the time required for cell conversion can be shortened. Previous studies have demonstrated the powerful potential of TF induction in stem cell differentiation and reprogramming somatic cells into pluripotent cells. For instance, the combined expression of TFs, Brn2, Ascl1, and MytL1 can convert human pluripotent stem cells (hPSCs) into functional neurons [78]. Additionally, the expression of a single TF, Ngn2, can induce the development of embryonic stem cells (ESCs) into mature neuronal cells in less than 2 weeks [79]. Conversely, the expression of specific TFs can also induce somatic cells into pluripotent stem cells, as seen in the induction of pluripotency in mouse embryonic and adult fibroblasts [80].

Fig. 6.

Application of TFs in stem cell engineering. A Direct genetic manipulation of TFs genes for directed differentiation of pluripotent cells. B Workflow for determining the role of specific TF in determining the direction of cell differentiation.

However, in order for this strategy to be applicable to humans, a crucial question arises: among the approximately 1600 TFs in the human body that have not been comprehensively tested, which of them can lead to targeted cell differentiation? This essentially raises the question of how to accurately select the appropriate target TF. To address this challenge, a computer-guided calculation-based strategy called IRENE has recently been developed to identify specific and more efficient TFs while maintaining a higher level of confidence [81]. Additionally, several other similar systems have been proposed [82]. Furthermore, innovative experimental methods with higher theoretical confidence have been employed (Fig. 6b). It begins with constructing a comprehensive TF library, followed by transducing the library into hPSCs and inducing their overexpression. Different cell types are then sorted based on the specific TF overexpression, and transcriptome sequencing is performed. Finally, the differentially expressed TF genes are analyzed using clustering and fitness estimation. This approach allows for the mapping of the gene module network and the identification of the specific TFs that contribute to determining the direction of cell differentiation. In doing so, hPSCs have been successfully programmed into neurons, fibroblasts, oligodendrocytes and vascular endothelial-like cells [83]. Furthermore, KLF4 and KLF5 were identified to promote the development of epithelial cells [84]. These findings suggest that TF-based programming for cell engineering can be systematically investigated and holds great promise for various applications.

The aforementioned strategy indeed holds potential in elucidating the role of TFs in cell differentiation. However, it is worth noting that lentiviral vectors carrying TF libraries have a limited infectivity in stem cells, which may lead to unintentional omissions in the search for functional TFs. Therefore, further research should concentrate on exploring the extent to which combinatorial TF expression can generate specialized cell types. Additionally, it is essential to develop orthogonal programming strategies that can fulfill the requirements of diverse applications based on this approach.

6.2. Contributions to human health

In the human genome, there are more than 1600 TFs, and approximately 19% of them have been closely associated with at least one disease [77], including cancer. Transcriptional dysregulation plays a crucial role in the development of numerous diseases, and even a dysregulation in a single TF can have profound consequences, leading to changes in cell fate [85]. TFs are integral components of the transcriptional program and have direct regulatory effects on target genes, making them more specific for disease regulation compared to upstream signaling proteins like kinases [86]. This higher specificity makes TFs highly significant in targeted therapy for diseases.

The regulatory balance of the human metabolic network is fundamental for maintaining the health and normal functioning of the organism, and human TFs play a crucial role in this dynamic balance. Abnormalities in the expression levels or functions of TFs directly impact human health (Fig. 7). Certain diseases are associated with overexpression of specific TFs. Examples include T cell acute lymphoblastic leukaemia [87], atherosclerosis [88], systematic lupus erythematosus [89]. Targeting these TFs through strategies such as inhibiting their transcription or introducing inactivating mutations, and targeted degradation can significantly disrupt the disease progression. Conversely, diseases caused by loss-of-function mutations in TFs, such as congenital heart, maturity-onset diabetes of the young [90], and IPEX syndrome [91], may benefit from therapeutic strategies involving the targeted use of agonists against related TFs to restore overall function.

Fig. 7.

The principle of TFs can contribute to human health. The balanced network of transcription factors maintains a healthy state, while their dysregulation causes the corresponding disease.

Furthermore, the use of small molecules to modify TFs and treat diseases caused by TF abnormalities is an attractive option [92]. However, TFs are generally considered “undruggable” because they often lack binding pockets for small molecules. The specialized DBD of TFs, such as zinc fingers, are typically optimized to recognize specific DNA sequences and stabilize TF-DNA interactions. This structural and functional characteristic make TFs unsuitable for accommodating small molecules [93]. Nevertheless, significant progress has been made in recent years in the engineering of small molecule modulators for TFs, thanks to the development of interdisciplinary fields such as molecular biology, structural biology, and virtual screening. Considering the mechanism of TF action, the ED can serve as an intrinsic control point for modulating transcriptional activity, and it has been traditionally regarded by medicinal chemists as the most promising target for the development of effective TF modulators. A notable accomplishment in this regard is the development of the androgen receptor antagonist enzalutamide [86]. Its pharmacology can be succinctly described as competitively inhibiting the binding of androgens to their receptors, which are TFs. Consequently, enzalutamide effectively hinders the receptors from binding to DNA [94]. This drug received FDA approval in 2019 due to its demonstrated efficacy in treating castration-resistant prostate cancer (CRPC).

It is crucial to acknowledge that aforementioned are based on the assumption that we already possess knowledge about the aberrantly expressed TFs. However, precisely determining them is a challenging task. Current large-scale studies on TFs primarily rely on RNA sequencing (RNA-seq). Nevertheless, predicting TF expression levels based on quantitative transcriptome data is difficult due to their low correlation coefficient [95]. Similarly, at the proteomic level, the accurate identification and quantification of TFs pose challenges due to their low expression abundance and the lack of relevant enrichment techniques [96]. Fortunately, a recent development involves a method to identify outlier TFs at the proteomic level [97]. This strategy utilizes synthetic DNA containing a concatenated tandem array of consensus TF response elements to enrich TFs. Following pulldown, coupled with mass spectrometry (MS) techniques for analysis, it ultimately enables the identification of aberrantly expressed outlier TFs. This approach undoubtedly deepens our understanding of how TFs perform their regulatory functions in tissues and provides a theoretical foundation for designing drugs that target TFs.

6.3. Contributions to microbial cell factory

With the advancements in metabolic engineering and synthetic biology, engineering TFs has emerged as a crucial strategy to enhance the performance of microbial cell factories [63]. In recent years, significant progress has been made in developing efficient metabolic engineering components based on the working mechanisms of TFs. These innovations have found applications in the biosynthesis of high-value-added products (Fig. 8). Several noteworthy applications deserve mention.

Fig. 8.

Applications of TFs engineering in microbial cell factories. Designed as a logic gate for application to genetic circuits; developed into biosensors as high-throughput screening tools; transcriptional modifications based on hybrid promoters to construct toolkits that can be used for fine tuning; introduction of environmental response elements for dynamic regulation.

Logic gates play a crucial role in controlling the expression of outputs by receiving and processing various input signals. The interactions between TFs and TFBSs form the basis for signal integration and calculation. There are three fundamental types of logic gates: “AND” gate, “OR” gate, and “NOT” gate. By selectively cascading these basic gates, more advanced forms can be obtained to meet higher requirements. The assembly of logic gates and pathway genes can create ideal genetic circuits that enhance the yield of biosynthesis. For instance, improved production of N-acetylglucosamine [35] or 2′-fucosyllactose [98] has been achieved through the implementation of such genetic circuits. A comprehensive review on the contribution of genetic circuits to metabolic engineering has been described by Seong et al. [99], providing detailed insights into this topic.

Biosensors have emerged as powerful intracellular detection tools for specific metabolites, offering high efficiency, accuracy, and simplicity. They operate on the principle that the binding of a specific metabolite to the ED of a relevant TF can modulate the regulation of its target gene, resulting in changes in transcription and expression levels. These changes exhibit a positive correlation with the metabolite concentration. By detecting alterations in the expression of target genes, such as fluorescent proteins or screening markers, biosensors can provide a reflection of the intracellular metabolite concentration. The use of biosensors eliminates the need for laborious cell fragmentation and offers the significant advantage of high throughput. Recently, several biosensors have been developed with exceptional performance for metabolites such as malate [16], erythritol [100], niacin [101], among others. These biosensors can be utilized as a strategy for high-throughput screening of high-yielding strains from large sample mutation libraries. However, challenges still exist. For instance, not all metabolites have specific response TFs associated with them, given the vast number of metabolite species. Alternatively, some biosensors may have impractical response thresholds or exhibit poor linear correlation, limiting their practical utility.

By adjusting the position of the TFBS within the promoter, it is possible to alter the transcriptional strength of downstream genes. This strategy can be competitively applied to achieve precise regulation of specific pathways. In the study conducted by Xu et al. [34], a library of promoters with varying initiation abilities, ranging from 0.7 to 36 times compared to the wild-type promoter, was obtained through such transcriptional modifications. Similarly, Zhu et al. [36] and Yu et al. [98] achieved desirable initiation capabilities of metabolic engineering elements using a similar approach. However, predicting the effect of adjusting the position of the TFBS on the initiation ability of obtained promoters remains challenging. To enhance predictability, statistical-guided prediction methods can be employed, as demonstrated by Tian et al. [102] in their investigation of the effect of N-terminal coding sequences on gene expression. Alternatively, deep learning methods, as employed by Ding et al. [103] to fine-tune the dynamic range of biosensors by focusing on cross-ribosome-binding sites, can also provide predictive capabilities.

By leveraging the responsiveness of transcription factors to signaling molecules, dynamic regulatory systems can be designed with innovation. Microbial cell factories incorporating such dynamic regulatory systems can effectively sense external signals, such as light, pH, temperature, dissolved oxygen, and autonomously modulate the flux of internal metabolic pathways to achieve optimal metabolic balance. Dynamic regulation, as opposed to simple gene overexpression or knockout, offers the advantage of minimizing metabolic burden, thereby maximizing productivity and performance [99]. For instance, the production yield of 2′-fucosyllactose was significantly enhanced by utilizing different temperatures as effectors to trigger dynamic regulation in a Bacillus cell factory [98]. Similarly, the introduction of light-controlled dynamic regulatory strategy has been applied to improve isobutanol biosynthesis [104]. However, one common challenge in dynamic regulation is the occurrence of regulatory delays due to longer response times. This can result in the differentiation of cell subpopulations with varying productivity due to uneven sensing of signals. To address this issue, the introduction of a population quality control system would be an ideal solution.

7. Conclusions and future perspectives

TFs are of great interest due to their essential physiological functions, and their rational application, based on a comprehensive understanding of their mechanisms of action, can have significant implications for human health and industrial production. Recent advancements in this field have further strengthened our confidence to pursue further research on TFs. However, it is crucial to clarify the future focus of research in this area. Firstly, it is important to recognize that our understanding of transcriptional regulation remains fragmented and incomplete. For instance, proposed mechanisms based on in vitro binding experiments, need to be validated in their true endogenous environment. Testing their function solely in vitro or in heterologous systems may not fully reflect their actual regulatory role in native hosts. Secondly, regulators should not be studied in isolation. The interplay of multiple TFs, following specific regulatory logic, forms a complex network of actions that maintains the intracellular environment's homeostasis. While ensuring high orthogonality in the specific application of particular TFs is essential to avoid global perturbations in non-target pathways within the metabolic network. Lastly, the design of artificial TFs offers opportunities to overcome limitations associated with natural TFs. Natural TFs may have constraints such as limited versatility or weak binding to signaling molecules or DNA. Artificial design of high-performance TFs can be achieved through strategies like structural domain substitution or engineering specific effector domains that can bind specific signals.

Although TFs play crucial roles in both eukaryotes and prokaryotes, research on TFs has been more extensively conducted in eukaryotes, particularly in humans and plants. This discrepancy can be attributed to the larger genome size and greater diversity of TF families in eukaryotes, as well as the availability of shared databases for eukaryotic TF research. However, considering the significant industrial applications of prokaryotes, such as high-value biomanufacturing and environmentally friendly bioremediation, it is essential to allocate resources and research efforts towards studying TFs in prokaryotes. Furthermore, prokaryotes offer a relatively simpler genetic background, which makes it easier to unravel complex phenomena and mechanisms. The insights gained from studying TFs in prokaryotes can potentially yield generalizable mechanisms that can be applied to eukaryotes. This mutually beneficial approach can lead to a win-win situation where knowledge and discoveries from prokaryotic TF research can be flexibly applied to eukaryotes, enhancing our understanding and applications in both domains.

CRediT authorship contribution statement

Hehe He: Conceptualization, Validation, Investigation, writing original draft; Mingfei Yang: Supervision; Siyu Li: Validation, Investigation; Gaoyang Zhang: Investigation; Zhongyang Ding: Writing, Editing; Liang Zhang: Writing, Editing; Guiyang Shi: Supervision, Project administration; Youran Li: Supervision, Project administration, Conceptualization. All authors read and approved the final manuscript.

Declaration of competing interest

The authors have declared no conflict of interest.