Length control of the injectisome needle requires only one molecule of Yop secretion protein P (YscP)

Rationale of the Approach and Selection of the Appropriate Alleles.

Our approach consists in measuring the needles made by partial diploid bacteria that express simultaneously two yscP alleles—a short one and a long one—and compare the distribution of needle length to the predictions made by a stochastic molecular model describing different needle length regulation scenarios. As a short YscP protein, we chose YscP₃₈₈, a derivative of YscP in which the spacer between the two export signals and one copy of the repeats in the central part have been removed (Fig. 1A) (16). To generate a long YscP, we engineered a restriction cleavage site between codons 250 and 251 in the central part of the yscP gene. We then generated yscP₆₈₆ by inserting a copy of codons 214–374, encoding the repeated region, into the restriction site (Fig. 1A). The two recombinant genes, cloned on the same expression vector were then introduced in Y. enterocolitica LJ4036, a strain in which the WT copy of yscP has been deleted from the virulence plasmid, and the phenotypes were analyzed (Fig. 1B). To test whether the two sorts of needles could be discriminated, equal numbers of Y. enterocolitica LJ4036 bacteria producing either YscP₃₈₈ or YscP₆₈₆ were mixed and 200 needles were measured. As shown in Fig. 1C, the histogram of needle length distribution (nld) revealed two distinct peaks, at the expected sizes of 48 ± 6 nm and 100 ± 12 nm. Surprisingly, even though equal amounts of bacteria were mixed, ~60% of the needles measured were of the small type (Table S1). The reason for this unequal representation was not investigated. We can nevertheless exclude that long needles break down because no small needles appear in cultures producing exclusively long needles (Fig. 1C).

Predicted Needle Length Distribution for Merodiploid Bacteria According to the Static and Dynamic Models.

In a first step, the nld of haploid bacteria was analyzed and found consistent with a Gaussian distribution (Fig. S1 and Table S2). On the basis of this result, a stochastic model was developed to describe the features of both the static and the dynamic needle length control mechanisms (see SI Text for details).

The one-ruler-per-needle model (scenario 1, Fig. 2A, Scheme S1) predicts that merodiploid bacteria will produce a mixture of long and short needles similar to a mixed culture of two different haploid strains. A statistical analysis of such distribution shows that it is fully consistent with a bimodal Gaussian distribution (Fig. 2A, Table S3). Moreover, within this scenario all needles are predicted to belong either to the short or to the long needle population, and no deregulated needles (i.e., needles growing to various lengths up to 1 μm) (16) are predicted by the model.

In the case of the dynamic ruler model, two scenarios can be envisioned, depending on whether the needle growth is interrupted whenever the length of the YscP traveling molecule is shorter than the needle (scenario 2) or only when the length of YscP exactly matches (within a tolerance) the needle length (scenario 3). For scenarios 2 and 3 the nld was computed (Scheme S2) using as order parameter the ratio between [YscF] and [YscP], which represent the effective concentration of needle subunits and ruler proteins, respectively. Their ratio represents the relative probability of accessing the basal body (Fig. 2 and Fig. S2). The stochastic model gives the number of YscP proteins exported through each needle during the growing phase: thus, the average “ruler per needle” (rpn) value along with the SD can be estimated as a function of [YscF]/[YscP] (Table S3 and Fig. S3). The average rpn value is a nonlinear function of [YscF]/[YscP]; both scenarios feature an average of ~75 rulers exported per needle when the concentration of available YscF and YscP proteins is equal at the basal body and ~4 rpn when [YscF]/[YscP] = 50 (Table S3 and Fig. S3). Values of [YscF]/[YscP] < 1 were considered unrealistic and not included. Full results are given in Fig. 2 and Fig. S2, where the number of needles is shown as a heat map (Fig. 2 B and E). Needle length distribution for scenarios 2 and 3 for values of [YscF]/[YscP] that are significantly consistent with experimental results are reported in Fig. 2 C and E, respectively. The percentage of rulers belonging to either the long- or the short-needle population is also reported along with the percentage of deregulated needles predicted by each model (Fig. 2 B and D).

For scenario 2, when the effective concentration of rulers is similar to that of needle subunits ([YscF]/[YscP] ~ 1), the system is overregulated by the short rulers, which meet the stopping criterion (i.e., needle length greater than the ruler length) before the longer ones. This is evident in Fig. 2B and Fig. S2 A–C, where a unique peak (corresponding to the short needles) is present for low [YscF]/[YscP] values. As [YscP] decreases ([YscF]/[YscP] up to 50), the initial peak broadens. Nonetheless, two distinct peaks (namely a bimodal Gaussian distribution) were never observed, as confirmed by statistical analysis of the length distributions performed using the expectation-maximization algorithm (EMA) (30) (Fig. 2 B and C, Figs. S2 and S3, and Table S3). Because this scenario imposes a stop whenever the needle length is greater than the passing-by ruler, even needles allowed to grow by a low measurement frequency (low [YscP]) will be stopped as soon as a ruler is allowed to enter the needle (Fig. 2 B and C and Fig. S3). Hence the scenario excludes the appearance of deregulated needles (Fig. 2B and Fig. S3).

For scenario 3, overregulation by the short ruler is observed only when [YscF]/[YscP] ~ 1 (Fig. 2D and Fig. S2 D–F). As computed for scenario 2 the rulers-per-needle value decreases with decreasing [YscP], dropping from high values (~12 rulers per needle) when [YscF]/[YscP] ~ 10 and reaching a value of ~4 for low [YscP] when [YscF]/[YscP] ~ 50 (Table S3, Fig. S3). As the measurement frequency decreases with lower values of [YscP], two distinct peaks do appear in correspondence with short and long needle distributions, respectively. For values of [YscF]/[YscP] > 25 (i.e., <6 YscP per needle) the computed nlds start to be compatible with a bimodal Gaussian distribution as demonstrated by the EMA analysis (Fig. 2 D and E, Figs. S2 and S3, and Table S3). Below this threshold the computed distributions are unlikely to be either Gaussian or bimodal (see SI Text for details, Fig. S3). Moreover, for [YscF]/[YscP] > 25 the ratio between short and long populations is balanced, whereas for lower values the short needle population P₁ is largely dominant (Fig. 2D and Fig. S2). The stopping rule is in fact more restrictive than in scenario 2, allowing for growth interruption only when the needle length is equal (with a tolerance) to the passing-by ruler length. As [YscP] decreases, the number of deregulated needles rapidly increases (Fig. 2D and Fig. S3). Thus, scenario 3, in contrast to scenario 2, features an increasing number of deregulated needles: When ~ ≤12 YscP proteins are allowed, the number of deregulated needles steadily increases (Fig. 2E and Fig. S3). As an example, scenario 3 produces 30% deregulated needles when the ratio of [YscF]/[YscP] is ~40, and the average number of rulers passing through the needle is ~5 (Fig. 2 D and E and Fig. S3). Importantly, as soon as the nlds showed a bimodal behavior and a balanced population of long and short needles (i.e., for [YscF]/[YscP] > 25), the deregulated needle fraction became significant (i.e., >20%, Table S3).

Merodiploid Strains for Short and Long yscP Alleles Produce Two Distinct Populations of Regulated Needles and No Deregulated Needles.

Three different types of genetic constructs were engineered. First, the yscP₃₈₈ allele was inserted on the 70-kb virulence plasmid by allelic replacement of the WT gene and the yscP₆₈₆ allele was introduced in trans on an expression vector, downstream from the strong pBAD promoter (Fig. 3A). In Y. enterocolitica E40, under the conditions used, this vector leads to a detectable expression in every bacterial cell (Fig. S4A). Both YscP proteins were produced and secreted by the merodiploid bacteria (Fig. 4A). The needle length histogram revealed two peaks at 54 ± 6 nm and 98 ± 15 nm, similar to those obtained with haploid bacteria expressing either yscP₃₈₈ from the pYV plasmid or yscP₆₈₆ from the pBAD vector (50 ± 5 nm and 103 ± 12 nm) (Fig. 4A).

In the second type of construct, the two alleles were placed on the same pDuet expression vector, each of them downstream from an individual T7 promoter (Fig. 3B). As a control, the two alleles were cloned in the same sites but individually, on the pDuet. As shown in Fig. S4B, egfp and mCherry cloned together on this vector were simultaneously expressed in every bacterium. The recombinant plasmid was introduced into the ΔyscP strain LJ4036 together with a plasmid encoding an inducible T7 polymerase. Synthesis of YscP was induced at the same time as type III secretion. The histogram of needle lengths revealed a clear peak of short needles (50 ± 8 nm) similar to the needles produced by haploids (48 ± 6 nm) beside a peak of longer needles (106 ± 12 nm), also similar to the needles produced by the corresponding haploids (110 ± 10 nm) (Fig. 4B).

Finally, both alleles were fused in a single operon, downstream from the pBAD promoter (Fig. 3C). Bacteria endowed with this construct showed again two populations of needles (50 ± 6 nm and 91 ± 13 nm) similar to those produced by bacteria endowed with similar constructs carrying only one allele (48 ± 5 nm and 103 ± 12 nm) (Fig. 4C and Table S4).

Deregulated needles were never observed in six different experiments with the three different constructs (Fig. 4). The results were then pooled (1114 measurements) and found consistent with a bimodal Gaussian distribution (based on EMA analysis in Table S1). As shown in Fig. 5, two distinct peaks account for short and long needle populations and are consistent with the predicted bimodal distributions found for scenarios 1 and 3 (Fig. 2 A and E), whereas scenario 2 prediction features single-peak distributions. The needle distribution computed according to scenario 3 significantly changes with the number of rulers per needle (i.e., [YscF]/[YscP]) passing from nld dominated by the short needle population (P₁) to nld where short and long needle populations are balanced (Fig. 2 D and E and Fig. S2). Nonetheless, a large number of deregulated needles (>20%, Fig. 2 D and E and Fig. S3) were always associated with the computed nlds, which are consistent with a bimodal Gaussian distribution and are similar to the experimental nld (i.e., for values of [YscF]/[YscP] > 25). Moreover, when the system is predicted to start producing deregulated needles ([YscF]/[YscP] ~ 10), the short-needle population is much more abundant than the long-needle population (P₁, ~80%; P₂, ~20%; Figs. S2 and S3, Table S3). This behavior does not agree with the experimental results, which indicate that the two populations are not so unbalanced (P₁, 62.5%; P₂, 37.5% of total merodiploid samples; Table S1).

Thus, scenario 3 appears to be also inconsistent with the experimental results because it predicts a significant percentage of deregulated needles (Fig. 2 D and E), whereas none was experimentally observed. A more quantitative comparison between the computed and experimental nlds based on EMA analysis (see SI Text for similarity estimates defined by Q1 and Q2, Fig. S3) strongly indicates the bimodal Gaussian distribution of scenario 1 as the best representation of the experimental needle distribution. Thus, the static one-ruler-per-needle model, which predicts no deregulated needles, can uniquely interpret the experimental data. It must be stressed here that deregulated needles are very easy to detect on an EM grid and that a percentage as low as 10% of such needles cannot be missed.

Molecular Clock Mechanism.

An additional length control mechanism was proposed for the flagellar hook, namely the molecular clock model (29). In this model, the cleavage of the inner-membrane protein FlhB is proposed to act as a molecular clock (29). When the cleavage has occurred, time for needle elongation is over. If this timing device would as well occur in the Yersinia injectisome, it could mask the presence of deregulated needles in our experiments. To rule this out, we introduced the long and short yscP alleles in a Y. enterocolitica strain, which produces a noncleavable version of YscU (yscUN263A), the homolog of FlhB (23). The histograms showed two populations of needles, although with a broader distribution, but no deregulated ones (Fig. S5). Hence, this experiment ruled out that the presence of deregulated needles in scenario 3 could have been masked by a molecular clock mechanism.