Understanding the regulation of protein synthesis under stress conditions

Abstract

Protein synthesis regulation primarily occurs at translation initiation, the first step of gene translation. However, the regulation of translation initiation under various conditions is not fully understood. Specifically, the reason why protein production from certain mRNAs remains resistant to stress while others do not show such resilience. Moreover, why is protein production enhanced from a few transcripts under stress conditions, whereas it is decreased in the majority of transcripts? We address them by developing a Monte Carlo simulation model of protein synthesis and ribosome scanning. We find that mRNAs with strong Kozak contexts exhibit minimal reduction in translation initiation rate under stress conditions. Moreover, these transcripts exhibit even greater resilience to stress when the scanning speed of 43S ribosome subunit is slow, albeit at the cost of reduced initiation rate. This implies a trade-off between initiation rate and the ability of mRNA to withstand stress. We also show that mRNAs featuring an upstream ORF can act as a regulatory switch. This switch elevates protein production from the main ORF under stress conditions; however, minimal to no proteins are produced under the normal condition. Because, in stress, a larger fraction of 43S ribosomes bypasses the upstream ORF due to its weak Kozak context. This, in turn, increases the number of scanning ribosomes reaching the main ORF, whose strong Kozak context can convert them into 80S ribosomes, even under stress conditions. This switching allows an efficient use of cellular resources by producing proteins when they are required. Thus, our computational study provides valuable insights into our understanding of stress-responsive translation-initiation.

Significance

Protein production decreases in most transcripts under stress conditions, but in some mRNAs it remains unchanged or even increases. We explain this phenomenon using a Monte Carlo simulation model of protein synthesis and ribosome scanning. We find that mRNAs with strong Kozak contexts exhibit minimal reduction in translation initiation under stress. In addition, we show that mRNAs with an upstream ORF can act as a regulatory switch, increasing protein production from the main ORF under stress while producing minimal to no proteins under normal conditions. We also show that a weak uORF and a strong mORF Kozak context are necessary for upregulating protein synthesis under stress.

Introduction

Protein synthesis is a vital biological process that is tightly regulated at multiple levels (1,2,3,4,5,6). It involves decoding the genetic information present in messenger RNA (mRNA) to produce functional proteins. This process occurs through a sequential series of steps: initiation, elongation, and termination. Initiation is the process of the formation of a fully assembled ribosome at the start codon. During elongation, the ribosome moves along the mRNA, adding amino acid (aa) subunits as per the genetic instructions carried by the mRNA. Protein synthesis terminates when the ribosome reaches the stop codon. Among these three steps, initiation plays the most crucial role in determining the overall rate of protein production (3,7,8,9). As the rate-limiting step of protein synthesis, it sets an upper bound on the overall production of proteins. The translation initiation rate depends upon several mRNA sequence features and varies significantly from one gene to another (3,7,9). Initiation rate can also vary substantially under different cellular conditions, particularly in response to stress (10,11,12,13,14,15). Stress conditions present a formidable test for the translational machinery for the efficient use of resources and ensure cellular survival (13,14,15,16,17). Cellular stress, whether arising from external factors, such as nutrient deprivation, environmental perturbations, or intracellular stresses like endoplasmic reticulum stress, poses substantial challenges for the survival and proliferation of a cell. In response to stress, cells exhibit a remarkable ability to reprogram their translational landscape through various means, including modifications in translation initiation rate (14,18,19,20). This reprogramming prioritizes the synthesis of key proteins that aid in stress management and survival, partly achieved through appropriate modifications in the translation initiation rate (18,19,20).

Translation initiation involves a multifaceted orchestration of factors and events that ultimately result in the assembly of a translationally competent 80S ribosome at the start codon (13,14,21,22,23,24). Central to this process is the formation of the eIF2-GTP-Met-tRNA ternary complex (TC), an essential component for the preinitiation complex (PIC). The PIC comprises the small 40S ribosomal subunit, TC, and some essential initiation factors (IFs). The start codon is selected through a scanning mechanism, with the ribosome 43S subunit moving along the mRNA in search of the start codon (25,26). The interplay between IFs guides the scanning process, ensuring fidelity in AUG recognition. The choice of the translation start site in eukaryotes is also influenced by the context of nucleotides surrounding the start codon (13,27,28). Notably, a consensus sequence, 5ʹ-[(A/U)A(A/C)A(A/C)AAUGUC(U/C)]-3′, known as the Kozak sequence, has been identified as the preferred motif for start codon recognition, underscoring the regulatory role of this sequence (9,13,14). During translation initiation in eukaryotes, the 43S ribosomal complex scans the 5′ untranslated region (UTR) until it identifies a start codon within a suitable Kozak context (13,14,25,29). The basepairing between the Kozak sequence and the 43S ribosomal subunit slows down 43S scanning near the start codon (30,31,32,33). Then, a stable base pairing of the codon-anticodon on the identified start codon initiates structural rearrangement, leading to GTP hydrolysis in the TC and the subsequent release of inorganic phosphate from the GTP bound eIF2 molecule (13,21,22,23,29,34). At this point, scanning comes to a halt, and the 43S subunit becomes trapped on the identified AUG codon. This recognition of the AUG codon leads to the release of eIF2-GDP, followed by the joining of the 60S ribosome subunit and the dissociation of the IFs (13,23,29,34). The resulting 80S ribosome now starts synthesizing a protein from a 5′ to 3′ direction on the mRNA molecule.

Translation initiation not only depends on the Kozak context but also on the stress levels, specifically the phosphorylation of the eIF2 molecule (13,14,21,29,35,36,37,38,39). Due to this phosphorylation, eIF2_p lose the ability to form the TC, thus significantly reducing the number of 43S ribosome subunits available for scanning and subsequent formation of 80S ribosome (13,14,29,36,38,39,40,41). In addition to this, it has been observed in mouse CHOP mRNA that the phosphorylation of the eIF2 molecule reduces the recognition of the start codon by the 43S ribosome complex, leading to increased bypassing at the start codon (14,21,35,36,38). Both of these experimental observations suggest a decrease in the translation initiation rate under stress conditions (13,14,18). However, studies have demonstrated an upregulation in the synthesis of proteins involved in managing stress conditions when the cells are subjected to stress (13,18,19,42,43,44,45). The factors leading to such behavior are not fully understood (13,18,42,43). In addition to that, it has also been shown that the transcripts coding proteins that help manage stress conditions tend to have an upstream open reading frame (uORF) with a weak Kozak context (13,18,42). However, the exact role of Kozak context in the upstream as well as in the main ORF (mORF) have not been fully understood (13,18,46). Moreover, transcripts with uORFs encoding stress management proteins from mORF that may or may not have any functional use. Yet, the question of how this metabolic burden is optimized remains unanswered. To elucidate the mechanisms underlying increase in the relative protein synthesis during stress conditions, we develop a model of ribosome scanning and protein synthesis. The model is simulated using the Monte Carlo simulation approach (7,8,47,48). Using the model, we find that strong Kozak context makes an mRNA resilient against stress conditions, maintaining consistently high protein production. However, a significant decrease in the initiation rate is observed in transcripts with a weak Kozak context. In addition, we find that the factors influencing the scanning speed of the 43S ribosomal subunit (e.g., specific sequence motifs and mRNA secondary structure) can significantly modulate initiation rates (49,50). Moreover, a reduced scanning speed in uORF-less transcripts equips them to better endure stress conditions; however, sacrificing the overall protein production. We also show that the presence of a weak Kozak context in a uORF acts as a regulatory switch, enabling higher protein production from the mORF under stress conditions, while limiting protein production under normal conditions. This phenomenon arises due to reduced recognition of the start codon of uORF under stress conditions. This leads to an increase in the 43S scanning ribosome reaching the mORF. We then systematically varied Kozak strengths of uORF and mORF, and identified the regime that allows an efficient increase in protein production from mORF under stress conditions. Furthermore, when ribosome reinitiation is introduced, we observe a nonlinear increase in the translation initiation rate as the nucleotide sequence length between the uORF and mORF (intercistronic length) increases. This finding aligns with the previous studies (51,52,53). In conclusion, the results reported in this paper underscore the crucial roles of scanning speed and Kozak sequence strength in governing translation initiation rates and highlight the adaptive nature of cellular translation dynamics.

Methods

Simulation model of 43S ribosome scanning and protein synthesis

We employ an extension of the totally asymmetric simple exclusion process (TASEP) model with two distinct particle types to simulate 43S ribosome scanning and protein synthesis (54,55,56,57). In this model, an mRNA is conceptualized as a one-dimensional lattice, while the 80S and 43S ribosomes are represented as extended particles, with each covering 30 nucleotides (Fig. 1) (55,56,57,58,59). The mRNA comprises the coding sequence, which is responsible for encoding the protein, and is flanked by the UTRs on both sides. The 5′ UTR of mRNA can host uORFs, which are also capable of synthesizing protein molecules (21,46,60). The protein synthesis begins with the binding of the PIC (i.e., TC-43S complex) at the 5′ end of the mRNA molecule (Fig. 1). The rate at which this step occurs is denoted by α_B, depends not only on the eIF2 concentration but also on several other IFs. Therefore, we assume the relation between α_B and the availability of eIF2 concentration follows the Hill function: α_B = k_b × f_H([eIF2]) (see supporting material for details), where k_b is a basal factor. Following the binding of the PIC at the 5′ end, The 43S scans one nucleotide at a time in the UTR to identify the start codon in the appropriate Kozak sequence context (13,14,18,61). The scanning of nucleotides in the UTR is unidirectional and occurs at a rate of ω_s (25,26,29). Upon reaching the appropriate start codon, the 43S scanning speed slows down by a factor f (hereafter referred to as Kozak strength factor [KSF]) due to its interaction with Kozak sequence (30,31,32). This means that the translocation speed on the Kozak sequence becomes ω_k = fω_s, thus increasing the chances of the formation of 80S ribosomes, which occurs with rate k_80S. Moreover, the recognition of the start codon also depends upon the phosphorylation status of eIF2 molecules. The increased phosphorylation of eIF2 results in reduced recognition of the start codon, thereby enhancing the bypassing of the 43S ribosome subunit (13,21,35,36,38). We incorporate this experimentally observed phenomenon in our model by assuming that 80S formation rate k_80S = k_r × f_H([eIF2]), where k_r is a basal factor for the recognition rate. Depending on the level of this recognition rate, the 43S ribosomal subunit may start the translation process if the start site is recognized in the uORF region or it may bypass the uORFs and continue downstream scanning until it finds the appropriate start site (Fig. 1). Once the initiation site is recognized, the ribosome proceeds to translocate one codon at a time at a rate ω_t. The movement persists until the ribosome reaches the stop codon, where it releases the polypeptide chain at a rate β for both uORF and mORF (Fig. 1). In our simulation model, we set ω_t = 5 aa s⁻¹ and β = 50 s⁻¹ (see Table S1 for the details of parameters used in our model).

Figure 1.

Schematic representation of translation process of an mRNA with one uORF. The yellow color represents the noncoding regions of the mRNA, the blue color uORF and the red color represent the mORF. Initiation of protein synthesis occurs at the start codons highlighted in dark gray, whereas protein synthesis is terminated at the stop codon of uORF and ORF, highlighted in black.

In addition to this, in some mRNA molecules, such as GCN4 and ATF4, it is observed that, after termination of protein synthesis at the stop codon of the uORF, the 40S subunit may remain attached to the mRNA. The TC can then attach to the scanning 40S subunit and start translation at the downstream start codon (14,52,53). The rate k_TC at which TC attaches to the 40S subunit also depends on the TC availability, which is influenced by the available unphosphorylated eIF2 concentration (21,52,53,62,63). We incorporate this phenomenon into our model by assuming TC binding rate has a Hill function-like dependence on eIF2 concentration: k_TC = k_tc × f_H([eIF2]), where k_tc is the basal factor for the TC binding rate.

We simulate 43S ribosome scanning and protein synthesis using the Monte Carlo simulation method, as used in previous studies (8,47,48,55,64,65). In this model, we denote the position of the 43S subunit of ribosome by the 17th nucleotide in the UTR and the 6th codon in the coding sequence, which is the position of the A site (3,9). The unidirectional translocation of 43S and 80S ribosomes on the lattice adheres to exclusion, which means that they cannot advance to the next site if they are obstructed by other downstream 43S and 80S ribosomes. In our simulations, we generate a uniformly distributed random number r between 0 and 1 to determine the occurrence probability of a specific transition. If, in a given state of mRNA translation, n transitions are possible, then we generate n such different random numbers for each of the possible transitions. To determine whether a specific transition will occur, we compare the probability of a specific transition in time dt with the random number r. Here, dt is the simulation time step and is kept fixed at 0.05 s, which is very small in comparison with the timescale at which different events, such as initiation, 80S formation, and codon translation, take place in our simulation model. If the random number r for a specific transition is less than the probability then this transition will occur, otherwise it does not. For example, the probability of binding of the 43S ribosome subunit to the mRNA’s 5′ cap in time dt is α_B × dt. Therefore, if the random number r is less than or equal to α_B × dt, the 43S subunit will bind to the 5′ end of the mRNA, provided that the first 30 nucleotides from the 5′ end are unoccupied by any other 43S subunits. Similarly, during the scanning process, if a 43S subunit is at the j-th nucleotide position, it will translocate to the (j + 1)-th position if r is less than or equal to ω_s × dt, given that no ribosome is present at the (j + 30)-th position. We repeat this procedure with all possible transitions and update the time and state of the system accordingly. Note that, depending on the probability of each transition, multiple transitions can occur within a single time step of dt. This generates a trajectory of the state of the translation system as a function of time, which we use to compute relevant quantities by putting several counters at different positions. For example, to calculate the initiation rate of the mORF, we count the number of ribosomes converting from 43S to 80S in a specific time duration. Dividing this number with time duration gives the initiation rate α for the mORF.

Results

Strong Kozak and low ribosome scanning rate makes an mRNA resilient against stress conditions

The Kozak consensus sequence is a stretch of nucleotides that surrounds the start codon in eukaryotes and plays a significant role in determining the translation efficiency of an mRNA molecule (13,27,28,66,67). The interaction of this sequence with the 43S ribosomal subunit helps in facilitating the recognition of the start codon, leading to the formation of a stable 80S complex (13,21,35). This interaction modulates the scanning rate of the 43S ribosome subunit in the proximity of the start codon (30,31,32). A strong Kozak context leads to a stronger decrease in the scanning speed compared with a weaker sequence. In our model, we represent this modulation in scanning speed by a factor f. The scanning speed of the 43S ribosome on the Kozak sequence therefore can be expressed as ω_k = fω_s where ω_s is the UTR scanning speed.

Recent studies have found that the variations in the Kozak sequence within mRNA influence translational efficiency distinctly (13). Transcripts with weak Kozak context show a stronger decrease in translation efficiency under stress conditions, whereas a minimal decrease in translation efficiency is observed in transcripts with strong Kozak context (13). We first assessed whether this experimentally observed behavior in protein synthesis can be reproduced using our simulation model under varying stress levels. For this, we set the KSF, f, for strong and weak Kozak to 0.2 and 0.9, respectively. The scanning speed ω_s was set to 11 s⁻¹ (for reference see Table S1). We compute the protein synthesis rate using our simulation model by varying eIF2_p level for mRNA with strong and weak Kozak contexts. We find that as we increase the phosphorylation levels of eIF2 molecules from 0 to 50%, protein synthesis rate in mRNA with a strong Kozak sequence decreases by 16%. However, for mRNAs with a weaker Kozak sequence, this decrease was 3.8-fold higher (Fig. 2 A). This result shows that the transcripts with strong Kozak context are more resilient against stress conditions, consistent with a previously published experimental study (13). We also test the robustness of these results by varying the half saturation constant of Hill function (see Eq. 3, Fig. S1, and methods) and found similar outcomes (see Fig. S2).

Figure 2.

Phosphorylation levels of eIF2 influence protein synthesis. (A) Protein synthesis rates from mRNA are depicted in a bar plot at 0, 20, and 50% of eIF2_p levels. The translation initiation rates are plotted against eIF2_p levels at different Kozak strengths and scanning speed in (B) and (C), respectively. In (B) 43S scanning speed is fixed at 11 s⁻¹, whereas in (C) Kozak strength factor was set to 0.2.

To further test the robustness of these conclusions, we also varied the stress level and measure the initiation rate for different KSFs, ranging from 0.1 to 0.9 (Fig. 2 B). We make three important observations. First, the initiation rate of mRNA having strong Kozak context is always higher than the weaker ones. Second, the initiation rate of mRNAs decreases monotonically with the increasing eIF2_p levels. Third, under stress conditions, there is a minimal reduction in initiation rate (up to a specific eIF2_p level) for sequences with a strong Kozak context. In contrast, this reduction is significant for transcripts with a weak Kozak context, aligning with the observations in Fig. 2 A.

The strength of the Kozak sequence and the availability of eIF2 is responsible for the observed phenomena. The first observation stems from the reduced interaction of the 43S subunit with the weaker Kozak sequence, consequently leading to a significant fraction of 43S subunits bypassing the start codon. Conversely, mRNAs featuring a strong Kozak sequence lead to the higher residence time of the 43S subunit at the canonical start codon, thus giving more opportunity for the formation of the 80S ribosome (13,14,21). The second observation can be explained by the lesser availability of TC due to the phosphorylation of the eIF2 molecule, leading to a reduced formation of PICs (13,14,29,36,38,39,40,41) (see model section for details). The reduction in the abundance of PIC molecules subsequently lowers the rate at which the 43S subunit binds to the 5′ termini of mRNA molecules. In addition to that, increased phosphorylation also increases the bypassing rate (13,21,35,36,38), therefore leading to a further decrease in the initiation rate. The rationale behind the third observation is: when the mRNA has a strong Kozak context and conditions are normal, the timescale of 80S formation at canonical start codon is much faster than the 43S bypassing. Therefore, a minimal decrease is observed in the initiation rate as the stress level increases (note well, the rate of 80S formation is k_80S = k_r × f_H([eIF2]), where [eIF2] is unphosphorylation level of the eIF2 molecule). However, a further increase in the phosphorylation of eIF2 levels leads to a sharper decrease in initiation rate, because a greater decrease in k_80S makes it much slower than the rate of ribosome bypassing. In contrast to that, in the mRNAs with weak Kozak context, phosphorylation leads to a sharper decrease, even at low stress conditions, because the timescale of 80S formation is comparable with or slower than the timescale of ribosome bypassing. Next, we also varied the scanning rate in the mRNA molecules from 5 to 17 s⁻¹, in steps of 3 s⁻¹, for the mRNAs containing the strong Kozak sequence (f = 0.2) and computed the initiation rate using our simulation model. With the increase in scanning speed, the initiation rate of transcripts with strong Kozak context increases (Fig. 2 C). This is because faster scanning allows 43S ribosomes to quickly reach the start codon, thus increasing the initiation rate. Interestingly, we also find that the initiation rate remains robust, even at higher phosphorylation levels, when the scanning rate is slower. This implies that a slower scanning rate equips the transcript better to withstand stress-induced translational challenges. The resilience at low scanning speed, where the translation initiation rate remains unaffected by the phosphorylation status of the eIF2 molecule, occurs because the 43S ribosome flux on mRNA is in a scanning rate-limiting regime. However, when the phosphorylation levels of eIF2 surpasses this regime, the 43S binding becomes the rate limiting step. This transition from the scanning rate-limiting to binding rate-limiting regime is equivalent to maximal current to the low density phase of the TASEP model (55,68,69). It is important to note that, when the scanning rate is lower, this transition occurs at a lower 43S binding rate, meaning at higher stress levels; thus making the transcript more resistant against strong stress conditions. These results show that making a transcript more resilient against stress conditions results in lower initiation under normal conditions, suggesting a trade-off between translation efficiency and stress resilience.

In summary, we show that mRNAs with strong Kozak sequences display resilience against stress conditions and consistently maintain higher initiation rates when compared with their weaker Kozak counterparts at all stress levels. These findings underscore the critical roles of scanning speed and Kozak sequence strength in governing initiation rates, emphasizing the resilience of transcripts featuring low scanning speed and robust Kozak sequences to stress-induced challenges.

Role of uORF and Kozak sequences in translation initiation during stress and normal conditions

Integrated stress response in eukaryotes phosphorylates the eIF2 protein, resulting in a decrease in translation efficiency of the majority of mRNA molecules (10,11,12,13,14,15,18). However, the expression of proteins that help manage stress and their translation efficiency increase during stress conditions (14,18,18,19,20). Such transcripts tend to have one or more uORFs with a weaker Kozak context (13,14,18,18,19,20). To understand this interesting observation in mRNAs with a single uORF (14), we introduced a uORF with a length of 118 nucleotides in the UTR (see methods). The 5′ end of this uORF is located 35 nucleotides away from the N-terminus of the mRNA molecule. We set the scanning speed ω_s = 8 s⁻¹ and assigned a KSF of 0.9 and 0.2 to uORF and mORF, respectively. With these parameters, we simulated protein synthesis from this mRNA and computed the protein synthesis rate at varying levels of eIF2 phosphorylation. We find that the protein production from the mORF is minimal under normal conditions and increases as the stress level rises (Fig. 3 A) (14,18,61), thus suggesting that uORF performs a dual role within the mRNA; under normal conditions it functions as a hindrance, potentially impeding the translation process. However, under stress conditions, this uORF exhibits a distinct role by upregulating mORF translation. This observation is consistent with the experimental outcomes of Wek and co-workers in mouse CHOP mRNA (14), again validating our simulation model. It is important to note here that the uORF has a weak Kozak context. A strong Kozak context in the uORF leads to a minimal to no increase in the protein production from the mORF under stress conditions (see the later part of this subsection for details). This again confirms the role of Kozak sequence in regulating the protein synthesis under stress conditions.

Figure 3.

(A) Protein synthesis rate of the mORF is plotted at different eIF2_p levels. Translation initiation of the mORF is plotted against the eIF2_p levels at varying Kozak strength of uORF and mORF in (B) and (C), respectively. In (B) the KSF of the mORF is kept fixed at 0.2, whereas in (C) the KSF of the uORF is set to 0.9. (D) Translation initiation of mORF is plotted at different scanning speeds of 43S ribosome against eIF2_p levels.

To further understand the role of Kozak strength, we simulated protein synthesis on an mRNA molecule containing a uORF. Then, using those simulation trajectories, we computed the initiation rate of the mORF as a function of eIF2 phosphorylation at uORF Kozak strength varying from 0.1 to 0.9. In these simulations, we do not change the Kozak strength of the mORF and fix it at 0.2. We find that a decrease in the Kozak strength of the uORF results in an increase in the initiation rate of the mORF. This pattern remains consistent at all eIF2_p levels (Fig. 3 B). This phenomenon arises due to substantial bypassing of the 43S subunit at the uORF initiation site when the uORF possesses a weak Kozak sequence (14,18,61,70). For more details, see the results section “strong Kozak and low ribosome scanning rate makes an mRNA resilient against stress conditions”. We also find that the initiation rate of the mORF increases with an increase in stress levels. However, a notable decrease in the initiation rate follows if stress levels continue to rise. This behavior arises because eIF2 phosphorylation can influence initiation rate in two different ways. First, the phosphorylation of eIF2 influences the 43S bypassing and second it also influences 80S ribosome formation of weak and strong Kozak differently (13,14,61). Therefore, under stress conditions, the uORF (with a relatively weaker Kozak strength) bypasses more 43S ribosome subunits to the mORF. This increases the total number of 43S ribosomes reaching the start codon of the mORF. Thus, increasing the initiation rate of mORF, because of its strong Kozak context, allows minimal bypassing, even under stress conditions (see result section “strong Kozak and low ribosome scanning rate makes an mRNA resilient against stress conditions” for details). However, increasing eIF2 phosphorylation also decreases the number of PICs available for binding at the 5′ terminus, decreasing the initiation rate. Thus, these two competing effects give rise to a nonmonotonic variation in the initiation rate as a function of eIF2 phosphorylation.

We next assessed the influence of the Kozak strength of mORF on its initiation rate. To this end, we simulated protein synthesis and computed the initiation rate as a function of eIF2 phosphorylation at mORF Kozak strengths varying from 0.1 to 0.9. We find that the mRNAs with stronger Kozak strength has a relatively higher initiation rate (Fig. 3 C) because the 43S ribosome bypassing at mORF decreases as the Kozak strength of the mORF increases (Fig. 2 B). We also find a nonmonotonic behavior of the mORF initiation rate as a function of eIF2 phosphorylation. The reason for the nonmonotonic trend is already explained in the context of Fig. 3 B. However, we also note that this nonmonotonic trend gradually diminishes as we reduce the Kozak strength of the mORF. This is because, under the stress conditions, the mORF receives more 43S subunits due to bypassing; however, it fails to convert them to 80S ribosomes due to its weak Kozak strength. We then understand how the scanning speed of the 43S subunit shapes the initiation rate of the mORF. For this, we simulate protein synthesis and compute the initiation rate by varying the phosphorylation levels of eIF2 proteins at different scanning speeds. We set the KSF of the uORF and mORF at 0.9 and 0.2, respectively. We find a higher initiation rate at a faster scanning speed as the 43S subunit can quickly reach the start codon of the mORF (Fig. 3 D). This again highlights the role of 43S ribosome scanning speed on governing the translation initiation rate of an mRNA molecule.

In summary, we find that introducing a uORF in the 5′ leader region significantly impacts protein synthesis. We also show that weak Kozak context of uORF is necessary for upregulating protein production under stress condition.

Optimal regime of Kozak strengths leads to a significant increase in protein production under stress conditions

In the previous section, we show that not every Kozak strength can withstand the stress-induced challenges. For example, an mRNA molecule with randomly assigned Kozak strengths at uORF and mORF does not necessarily enhance protein production under stress conditions. Therefore, it becomes important to understand if there is any optimal regime of Kozak strengths that enables mRNA molecule to significantly elevate the protein production under stress conditions. To understand this, we systematically varied the KSF of uORF and mORF of the mRNA molecule in our simulation model and computed their initiation rates. Then, a heatmap of those initiation rates under normal and stress conditions was plotted as a function of f_uORF and f_mORF (Fig. 4). We find that, under normal condition, increasing f_uORF enhances the initiation rate of uORF (α_uORF) but remains uninfluenced by variations in f_mORF (Fig. 4 A). On the other hand, the initiation rate of the mORF rises with an increase in the Kozak strength of the mORF. However, an elevation in the Kozak strength of the uORF decreases this initiation rate (Fig. 4 B). The reason for this is that weak Kozak context of the uORF results in bypassing more 43S subunits, leading to lower initiation rates of the uORF. However, this increase in bypassing increases the 43S flux reaching the start codon of mORF, thus positively contributing to the mORF initiation rate.

Figure 4.

Heatmap showing the variation of the initiation rate of uORF and mORF as a function of f_uORF and f_mORF. The initiation rate heatmaps for uORF and mORF are plotted under normal conditions in (A) and (B), respectively, while those under stress conditions are presented in (C) and (D).

We also compute the initiation rate of uORF and mORF under stress conditions (60% phosphorylation of eIF2) as a function of f_uORF and f_mORF. We find a similar qualitative behavior as in the case of normal conditions. However, the regime of high uORF initiation rate under stress conditions (i.e., yellow region) has shrunk and remains limited to only the high Kozak strength region of uORF. Similarly, the regime of high mORF initiation rate under stress conditions (i.e., yellow region) has also shrunk and requires low Kozak strength of the uORF and high Kozak strength at mORF. This is because, under stress conditions, the recognition is low at uORF; therefore, to have high initiation at uORF, Kozak strength should be higher. Similarly, achieving a high initiation rate at the mORF requires a greater proportion of bypassing at the uORF and increased recognition at the mORF. These objectives can be accomplished by maintaining low Kozak strength at the uORF and high Kozak strength at the mORF, respectively.

The efficiency of both uORF and ORF is distinctly impacted by varying levels of eIF2_p

We have shown that protein production from the uORF with a weak Kozak strength is high under normal conditions but decreases significantly under stress. Similarly, the protein production from the mORF increases under the stress conditions but minimal amount of protein is produced from this ORF under normal situations. This switching of protein production between uORF to mORF allows efficient use of the cellular resources under both conditions. To have a better understanding of this switching phenomenon, we define the parameters χ_uORF and χ_mORF for the uORF and mORF, respectively, as follows:

$$X_x=\left(\frac{{\alpha }_x\left(S\right)/j_S}{{\alpha }_x\left(N\right)/j_N}\right)$$ (1)

Here, x ϵ {uORF, mORF}, j_N and j_S are the total flux of 43S ribosomes entering into the mRNA. In addition, symbols N and S are used to indicate normal and stress conditions. This parameter describes the proportion of 43S flux converted into 80S ribosomes during stress conditions compared with normal conditions for both uORFs and mORFs.

For an efficient use of resources under stress conditions χ_uORF should be minimum, whereas χ_mORF should be high. Similarly, under normal conditions, χ_mORF should be minimum because high protein production from mORF under normal conditions may lead to additional metabolic burden to the cell. Thus, together, both of these parameters capture the switching efficiency of the mRNA molecule. We computed both of these parameters as a function of eIF2 phosphorylation for KSF varying from 0.2 to 0.9 (Fig. 5, A and B). χ_uORF is higher for uORFs with a strong Kozak sequence in comparison with the weaker ones but is always less than one. On increasing the eIF2_p levels, the 43S ribosome bypassing is higher, therefore χ_uORF decreases monotonically. We also note that, for the uORFs with a strong Kozak sequence, the χ_uORF remains almost constant as eIF2_p levels are increased. This happens due to the high residence time of 43S ribosome on the initiation site of the uORF, thus giving an advantage for the 43S subunit to recognize the start codon of the uORF (see Fig. 2 B). Similarly, we also calculated χ_mORF as a function of eIF2 phosphorylation at varying Kozak strengths of the mORF. In these simulations, the Kozak strength of the uORF remains weak and fixed at f_uORF = 0.9. We find that χ_mORF is greater than 1 at each Kozak strength unless the phosphorylation is very high (Fig. 5 B). We also find a nonmonotonic trend in χ_mORF as a function of eIF2 phosphorylation. The initial increase in χ_mORF is due to a higher bypassing of 43S subunits at the uORF, thus giving more chance for the 80S ribosome formation at mORF. However, when the phosphorylation of eIF2 becomes high, the 43S subunits even fail to recognize the start codon in a strong Kozak context of mORF. Therefore, we see a nonmonotonic trend in χ_mORF. We also note that high Kozak strength leads to a higher value of χ_mORF and can withstand the stress conditions, even at high phosphorylation level, i.e., the peak shifts toward a higher eIF2_p level, because a stronger Kozak strength at the mORF allows minimal 43S bypassing, even at higher stress levels (see Fig. 2 B). Taken together, these results show that the efficient switching requires weak and strong Kozak strength at the uORF and mORF, respectively. Also, higher protein production by mORF under stress conditions requires a greater sacrifice of ribosome resources under normal conditions.

Figure 5.

The efficiency parameter χ_uORF and χ_mORF are plotted against the eIF2_p levels at different KSFs. In (A) f_uORF is varied from 0.1 to 0.9, whereas f_mORF is fixed at 0.2. In (B) f_mORF is varied from 0.1 to 0.9, whereas f_uORF is fixed at 0.9.

Nonlinear relationship between initiation rate and intercistronic distance

The model we have discussed so far is based on the experimental studies of mouse CHOP mRNA, where reinitiation has not been observed (14). In contrast, reinitiation does occur in mRNAs such as GCN4 and ATF4 (14,21,51,52,53,62). After ribosomes complete translation of a uORF, the 60S subunit and other molecular factors dissociate. However, the 40S ribosome subunits can remain attached to the mRNA and continue scanning the nucleotide sequence (51,52,53). Then, depending on the availability of TCs, the 40S subunit can acquire TCs during scanning and initiate protein synthesis at a downstream uORF or the mORF (21,52,53). To understand how ribosome reinitiation influences protein synthesis in a single uORF model, we introduced reinitiation by allowing 10, 20, 30, 40, and 50% of the 40S subunits of translating ribosomes at the uORF to remain attached to the mRNA and reinitiate at the mORF. We set the scanning speed at 8 s⁻¹ and maintained the KSF at 0.9 for the uORF and 0.2 for the mORF. Using these parameters, we simulated protein synthesis from this mRNA and computed the protein synthesis rate at varying levels of eIF2 phosphorylation. Our results show that protein production from the mORF is minimal under normal conditions and increases under stress conditions (Figs. 6 A and S3). This behavior is qualitatively similar to the result observed when reinitiation was absent (Fig. 3 A), again highlighting the dual role of the uORF in allowing minimal protein production under normal conditions while increasing it under stress conditions. We also note that the difference between the protein production of mORF in normal and stress conditions reduces as reinitiation is increased (Figs. 6 A and S3).

Figure 6.

(A) The protein synthesis rate of the mORF at a 10% ribosome reinitiation rate is plotted against different eIF2_p levels. (B) The translation initiation of the mORF at a 40% ribosome reinitiation rate is plotted against the intercistronic distance between the uORF and mORF at varying eIF2_p levels. In (A) and (B) the KSF of the uORF and mORF is fixed at 0.9 and 0.2, respectively.

This occurs because, due to ribosome reinitiation, a fraction of 40S subunits remains attached to the mRNA, initiating protein synthesis at the mORF. This fraction of reinitiating ribosomes adds to the bypassing of 43S subunits, thereby increasing the protein production under normal condition. Under stress conditions, the 43S ribosomes predominantly bypass the uORF, so ribosome reinitiation has little influence on increasing protein production under stress conditions. Thus, reinitiation reduces the difference in protein synthesis between normal and stress conditions (Fig. 6 A).

We also varied the intercistronic distance between the uORF and mORF from 13 to 105 nucleotides and calculated the initiation rate at different eIF2 phosphorylation levels (Fig. 6 B). We observe that, at each intercistronic distance, the initiation rate is higher when eIF2_p is higher. However, further increasing eIF2_p levels beyond a certain level (eIF2_p > 45%) results in a decrease in initiation rate due to lower availability of PICs (Fig. 6 B). This is consistent with the result shown in Fig. 3. We also found that increasing the nucleotide sequence length between the uORF and mORF initially increases the translation initiation rate and then saturates after a certain nucleotide length. This behavior has also been observed in previous studies (51,52), indicating that the relationship between nucleotide sequence length and protein synthesis rate scales up nonlinearly. This is because longer intercistronic distances increase the time it takes for the 40S subunit to reach the mORF start codon, thereby increasing the probability of the 40S subunit acquiring the TC on the way to the mORF start codon. Together, these findings elucidate how reinitiation modulates protein synthesis, highlighting its significant impact on translational regulation in both normal and stress conditions.

Discussion

Translation initiation stands as an important step in gene translation, exerting significant influence over the rate at which proteins are synthesized within a cell (3,7,8,9). It is a multistep process and is regulated by a multitude of factors including the binding of PIC, scanning of 43S ribosome, recognition of the start codon and Kozak strength, etc. (13,14,21,23,24). However, how these factors influence initiation rates is not fully understood (14,18,23). Specifically, the reason why certain mRNAs remains resistant against the stress condition while others do not (13). Moreover, why protein production from a few transcripts is enhanced under stress conditions, whereas it is decreased in most of the transcripts (14,18,61). To understand this, we develop a simulation model of protein synthesis that includes all relevant substeps of the process of translation initiation. Our analysis focused on two types of mRNAs: those lacking uORFs in their 5′ UTR, and those containing a single uORF within the UTR. Using our simulations, we find that mRNAs with strong Kozak context are more resilient against the stress condition. This is because the residence time of the 43S ribosome on a strong Kozak sequence is significantly higher than the timescale of 80S formation, thus minimizing the impact of ribosomal bypassing due to stress conditions (Fig. 2, A and B). We also show that the presence of a single uORF in an mRNA molecule can elevate the protein synthesis from the mORF under stress conditions. Because, under stress conditions, a larger fraction of 43S ribosomes bypasses the uORF due to its weak Kozak context (Fig. 3). This, in turn, increases the number of scanning ribosomes reaching the mORF, whose strong Kozak context can convert them into 80S ribosomes even under stress conditions (Fig. 3 C). We further demonstrate that Kozak strengths of both uORF and mORF must fall within an optimal range to resist the stress-induced challenges. For example, a significant increase in protein production from the mORF under stress conditions requires weak and strong Kozak strengths at uORF and mORF, respectively (Fig. 4 D).

Based on this experimental observed bypass phenomena in CHOP mRNA (14), we assume that the phosphorylation of eIF2 directly influences the recognition of the start site. By incorporating this into our model, we were able to reproduce the following three experimental observations in previously published studies (13,14). First, the strong Kozak context leads to a higher initiation rate (13). Second, mRNAs with strong Kozak context exhibit minimal changes in the initiation rate on the onset of stress conditions (13). Third, a uORF acts as a barrier for the downstream translation of mORF under normal conditions; however, it increases the protein production under stress conditions (14). The consistency with experimental data not only confirms the accuracy of our simulation model but also demonstrates that the bypassing phenomenon alone explains the upregulation of initiation rates under stress conditions.

In addition to the bypassing phenomena, we introduced ribosome reinitiation in our single uORF model and observed a similar qualitative behavior as in the case of no reinitiation (Fig. 3). However, it is important to note that allowing ribosome reinitiation beyond a certain level will negate the dual role played by the uORF under normal and stress conditions. This is because the 43S ribosomes that were translating the uORF and then dissociated at its termination site will now travel to the mORF start codon to initiate protein synthesis. This results in increased protein production from the mORF under normal conditions, thereby reducing the difference in protein production between normal and stress conditions (Fig. S3). This may explain why GCN4 and ATF4, which rely on ribosome reinitiation, employ a different mechanism to enhance protein production under stress conditions that involves multiple uORFs (21,51,52,62). We also varied the intercistronic distance between the uORF and mORF and found that the initiation rate increases nonlinearly as the intercistronic distance is increased. This is because a longer intercistronic distance increases the time of arrival of the 40S subunit to the mORF start codon, thus enhancing the probability of the 40S subunit acquiring TC to a saturation level. This result is consistent with the experimentally observed phenomena in GCN4, further validating the accuracy of our model (51).

Previous studies have modeled GCN4 mRNA and explained the increase in protein synthesis under stress conditions (52,53). For example, Coghill and co-workers developed a model to explain the roles of ribosome reinitiation, uORFs, and the intercistronic distance between uORF and mORF in regulating protein synthesis under both normal and stress conditions (52). McCarthy and co-workers then expanded this model to investigate cell-to-cell protein expression variability in GCN4 (53). However, our study fundamentally differs from these previous works for three reasons. First, we modeled uORF-less mRNA to examine how the Kozak context, scanning speed, and eIF2 phosphorylation affect translation initiation under both normal and stress conditions. This model explains why certain mRNAs are resilient to stress while others are not. Second, based on experimental findings of CHOP mRNA, we demonstrate that increased protein production under stress is possible without multiple uORFs due to the ribosome bypassing mechanism, unlike GCN4 and ATF4, which require multiple uORFs to elevate protein production under stress (21,52,62,71). Third, we highlight the role of the Kozak context sequence in upregulating protein synthesis under stress conditions, showing that a weak Kozak context in uORFs is necessary for enhancing protein production during stress.

A previous study has explained the upregulation of protein synthesis under stress conditions through high 43S ribosome density in the UTR (60). In that study, it was argued that, due to a high 43S ribosome density under normal conditions, a scanning ribosome takes longer time to reach the start codon of mORF, resulting in a low initiation rate. However, under stress conditions, 43S ribosome traffic is relieved due to limited availability of PIC, leading to a faster initiation at mORF. We also note that ribosome traffic on the coding sequence normally occurs in the low-density regime because it minimizes any traffic jams and thus allows an efficient use of all resources (47,55,72,73). If we consider the 43S ribosome movement as an extended rod moving on a one-dimensional lattice, then the scanning process becomes similar to the protein synthesis in the coding regime. This also means that high 43S ribosomes traffic makes the scanning process inefficient as it leads to the consumption of a significantly high number of ribosomes, which are present in a limited amount, even under normal conditions (7,74). Therefore, the reason for the increase in initiation rate under stress conditions is less likely to be linked with the traffic jam mechanism. In this study, we also show that upregulation of initiation rate under stress conditions is possible without relying on the inefficient high density phase of scanning ribosomes.

In our uORF-less mRNA simulation model, we find the initiation rate of the mRNA with reduced scanning speed remained unaffected, even at higher stress levels (Fig. 2 C). The scanning speed of a 43S ribosome can be influenced by specific sequence motifs and mRNA structures (49,50). Therefore, evolutionary selection pressure can possibly leverage these factors to fine-tune the scanning speed for making certain genes stress resistant. This observation also highlights a trade-off between protein production and stress resilience.

Our single uORF model is based on CHOP mRNA, which encodes a protein regulating apoptosis, gene expression, protein synthesis, and metabolic pathways in response to endoplasmic reticulum stress-induced eIF2 phosphorylation (19,21,45,62,75). The translation of CHOP mRNA is regulated by a uORF in its 5′ leader region, inhibiting mORF translation under normal conditions. Upon cellular stress, eIF2 phosphorylation alters this control, increasing protein production and contributing to the integrated stress response (14). Experimental evidence shows that stress-induced upregulation of protein synthesis involves bypassing 43S subunits at the uORF start codon, although the exact mechanism remains unclear. It is plausible that eIF2 phosphorylation activates additional translation factors facilitating this bypass. Unlike mRNA molecules, such as GCN4 and ATF4, CHOP mRNA does not support ribosome reinitiation or internal ribosome entry site activity; instead, regulation occurs through bypassing phenomena (14). Thus, our results also highlight that protein synthesis can be elevated under stress conditions without ribosome reinitiation. Moreover, it would be interesting if future experimental studies investigate the mechanism of bypass and the molecular players involved in this process.

In summary, this study elucidates the influence of stress-induced eIF2 phosphorylation on mRNA translation initiation. It highlights the critical roles played by Kozak sequence strength, uORF presence, and scanning speed of the 43S ribosomal subunit in modulating initiation rates under varying conditions. The findings offer insights into cellular adaptation to stress, showcasing how cells optimize resource utilization for efficient protein synthesis. Moreover, the study’s alignment with experimental data substantiates the model’s accuracy, indicating potential avenues for understanding and manipulating translation regulation in response to cellular stress in biological systems.

Data and code availability

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

Acknowledgments

I.U.I. acknowledges the Ministry of Education, Government of India, for the Prime Minister’s Research Fellowship (3001317). A.K.S. acknowledges support from the Department of Biotechnology, Government of India (BT/PR34367/BID/7/987/2020), and the SERB Core Research Grant (CRG/2022/001127).

Author contributions

I.U.I. developed the simulation model and performed the analysis. I.U.I. and A.K.S. designed the study and wrote the manuscript.

Declaration of interests

The authors declare no competing interests.

Editor: Ramon Grima.