Shear-dependent ‘stick-and-roll’ adhesion of type 1 fimbriated Escherichia coli

Summary

It is generally assumed that bacteria are washed off surfaces as fluid flow increases because they adhere through ‘slip-bonds’ that weaken under mechanical force. However, we show here that the opposite is true for Escherichia coli attachment to monomannose-coated surfaces via the type 1 fimbrial adhesive subunit, FimH. Raising the shear stress (within the physiologically relevant range) increased accumulation of type 1 fimbriated bacteria on monomannose surfaces by up to two orders of magnitude, and reducing the shear stress caused them to detach. In contrast, bacterial binding to anti-FimH antibody-coated surfaces showed essentially the opposite behaviour, detaching when the shear stress was increased. These results can be explained if FimH is force-activated; that is, that FimH mediates ‘catch-bonds’ with mannose that are strengthened by tensile mechanical force. As a result, on monomannose-coated surfaces, bacteria displayed a complex ‘stick-and-roll’ adhesion in which they tended to roll over the surface at low shear but increasingly halted to stick firmly as the shear was increased. Mutations in FimH that were predicted earlier to increase or decrease force-induced conformational changes in FimH were furthermore shown here to increase or decrease the probability that bacteria exhibited the stationary versus the rolling mode of adhesion. This ‘stick-and-roll’ adhesion could allow type 1 fimbriated bacteria to move along mannosylated surfaces under relatively low flow conditions and to accumulate preferentially in high shear regions.

Introduction

Many bacteria bind to tissues or other surfaces under fluid flow, but shear stress has been shown both to prevent bacterial attachment and to wash off bacteria that are already bound (Wang et al., 1995; Dickinson et al., 1997; Shive et al., 1999). It was traditionally assumed that this occurs because receptor-ligand bonds are ‘slip-bonds’ (Dembo et al., 1988) that weaken exponentially in the presence of a tensile force (Bell, 1978; Evans, 2001). However, it has been proposed that some bonds could instead be ‘catch-bonds’ that become stronger under disrupting force (Dembo et al., 1988; Isberg and Barnes, 2002; Thomas et al., 2002; Konstantopoulos et al., 2003; Marshall et al., 2003; Sarangapani et al., 2004). In particular, we have shown that mannose-specific type 1 fimbria of Escherichia coli are able to mediate a shear-enhanced interaction between the bacteria and target cells (Thomas et al., 2002). Our structural simulations and mutations together predicted that a force-induced conformational change in the fimbrial adhesin is associated with stronger binding, which suggests that catch-bonds are involved in the shear-enhanced adhesion of type 1 fimbriated E. coli.

Type 1 fimbriae are expressed by over 90% of E. coli and other enteric bacteria, and mediate binding to N-linked oligosaccharides with terminally exposed mannosyl residues. The adhesive properties of the type 1 fimbriae are determined by the fimbrial tip-associated lectin-like protein FimH (30 kDa). The FimH adhesin exhibits minor structural variations between different E. coli strains ( Sokurenko et al., 1994; 1995). While all naturally occurring FimH variants are capable of mediating strong interactions with oligomannose, the ability to bind single monomannose (1 M) residues is a highly variable property of FimH (Sokurenko et al., 1997; 1998). Most FimH variants found in E. coli of intestinal origin are unable to bind strongly to 1 M-coated surfaces in traditional assays that use static binding conditions, while FimH variants of uropathogenic E. coli origin can adhere via 1 M relatively well. The enhanced ability to bind 1 M was shown to result from any of a number of point mutations in FimH, none of which are in the mannose-binding site.

Recently we showed (Thomas et al., 2002) that bacteria expressing high-1 M-binding FimH variants were able to mediate agglutination of guinea pig red blood cells (RBCs) both in static conditions and in the presence of shear stress, while the low-1 M-binding FimH mediated bacterial binding to RBC only under shear stress. It was proposed that the low-1 M-binders demonstrate a shear-dependent adhesion mode, and this was confirmed in parallel plate flow chamber experiments, where the strength of interaction between RBCs and a carpet of surface-immobilized bacteria was measured. These experimental studies were paralleled by steered molecular dynamics simulations to study how mechanical force caused by shear stress effects the structure of the FimH protein (Thomas et al., 2002). Mechanical force broke six backbone hydrogen bonds to extend the terminal strand of the lectin (mannose-binding) domain into a long linker chain connecting the lectin domain to the pilin (fimbria-anchoring) domain. The hypothesis that this linker chain extension was associated with a switch from a low to a high affinity state was then supported by predicting and testing point amino acid substitutions in FimH that should promote or inhibit the linker chain extension. This provided the first structural insights into a mechanism by which a protein can be switched from weak to strong binding by mechanical force, i.e. for how a catch-bond might work (Thomas et al., 2002). The ability of high-1 M-binding FimH variants to bind at low shear was suggested to result from the enhanced ability of these variants to extend the linker chain under shear stress and switch to the high affinity state.

Although the shear-enhanced binding between FimH and target cells was clearly established in the previous study, many important points remain to be resolved. It is not clear, for example, what kind of mannose structures are capable of mediating shear-dependent adhesion with the FimH adhesin. It also remains to be determined how bacteria expressing different structural variants of FimH would adhere under the physiologically relevant setting in which free-flowing bacteria bind to mannose-coated surfaces (instead of using the assay of RBC binding to the bacterial carpet used in the previous study). Furthermore, it has been proposed for other systems that increased flow may bring, or transport, the binding partners together more quickly, resulting in an increased rate of bond formation during cell-surface adhesion (Alon et al., 1997; Chang and Hammer, 1999; Chen and Springer, 1999; 2001). It has yet to be determined whether such a mechanism contributes to shear activation in the FimH system.

Thus, we measured here the binding of E. coli to 1 M-coated surfaces in flow chambers in order to understand how shear stress affects the manner in which free-flowing bacteria expressing different FimH variants adhere to purified mannose compounds. This allowed us to demonstrate that an increased duration of bacterial binding, not a faster rate of bacterial attachment, is responsible for the shear-enhanced surface binding. It also allowed the discovery that the shear-dependent FimH−1 M interaction creates a complex adhesive behaviour of surface-attached bacteria. Finally, we discuss how these results correspond to a catch-bond hypothesis and in particular to the force-induced conformational change described earlier (Thomas et al., 2002).

Results

Modes of bacterial attachment to 1 M-coated surfaces

Type 1 fimbriated E. coli bacteria expressing the low-1 M-binding most common structural variant of FimH, FimH-f18 (Thomas et al., 2002) were passed through parallel plate flow chambers over a surface coated with monomannosylated bovine serum albumin (1 M-BSA). To detect even short-lived interactions with the surface, we recorded phase-contrast time-lapse videos holding the shutter of a digital camera open just long enough at each shear level to see bound bacteria but to blur out free-floating bacteria moving at the hydrodynamic velocity near the surface (as verified with non-adhesive control surfaces). Several examples of these videos are included in the supplementary material.

In order to determine how bacterial adhesion could be affected by shear-mediated effects on transport, or movement of bacteria to the surface, we first studied the initial attachment rates of bacteria. The rate of attachment to the 1 M-BSA surface was determined by counting all bacteria that bound to the surface, including those bound only transiently (less than 1.0 s). At low shear (below 0.26 pN µm⁻²), over 100 bacteria per minute attached in the field of view. As shear increased to 2.2 pN µm⁻², the rate of attachment decreased by 95% (Fig. 1A). At low shear, however, almost 90% of bacteria bound transiently, detaching within the first second (Fig. 1B) with only 2% of the bacteria remaining bound for over 30 s. This rapid attachment and detachment at low shear is illustrated in the movie, Fig.S1 of the supplementary material. In contrast, by 2.2 pN µm⁻² the number of transient attachments decreased to less than 10% while the number of long-term (>30 s) adhesions increased to over 70%. The movies, Fig.S2 and S3 of the supplementary material, illustrate the attachment and detachment of bacteria at moderate and high shear stress. Thus, shear decreased the initial rate of binding to 1 M, suggesting that any shear-related phenomena that could be hypothesized to increase the rate of transport of bacteria to the 1 M surface cannot be responsible for any shear-enhancement of bacterial binding observed here.

Figure 1. The effect of shear on attachment and detachment of E. coli expressing FimH-f18 binding to surfaces coated with 1 M-BSA. 1 M-BSA-coated surfaces were placed in a parallel plate flow chamber through which E. coli were washed for 5 min at various levels of shear stress (1 pN µm⁻² = 1 Pascal, or 10 dynes cm⁻²). Time-lapse videos were taken at a rate of two frames per second for 5 min. Bacteria were tracked by hand to determine when they first attached and how long they remained attached. A. The initial attachment rate was measured by counting the rate at which newly bound bacteria appear in images taken every half-second. The error bars in (A) and (B) give 95% confidence intervals. B. The duration of binding was determined by tracking each newly bound bacteria identified in (A). Bacteria that bound for less than 1 s were classified as transiently binding (•), while those binding over 30 s were classified as long-term binding (▪); the remainder bound between 1 s and 30 s and were not shown. C. To illustrate that the bacteria attach in two modes – rolling and stationary – the distance travelled by three bacteria attaching at moderate shear (0.4 pN µm⁻²) is plotted as a function of time. Note that detached bacteria moved at least 360 µm s⁻¹ at this flow rate, or about 10-fold faster than the rolling bacteria.

However, shear also increased the length of time each bacterium remains bound, and two adhesive modes were observed to be involved in the ability of bacteria to remain bound to the surface; these are illustrated in Fig. 1C. First, bacteria at times stuck firmly to the surface, binding in a ‘conventional’ stationary mode. In the stationary mode, surface-bound bacteria did not move significantly (over 1 µm) during at least 1 s of observation. Second, bacteria often rolled along the surface. This rolling mode of adhesion was characterized by movement in the direction of flow that is too fast to be defined as stationary but is at least twofold slower than observed for free-floating cells. While this definition is broad, rolling bacteria typically moved at roughly 30 µm s⁻¹ (illustrated in Fig. 1C in the top trace). Rolling bacteria sometimes halted abruptly to stick firmly in the stationary mode for times ranging from seconds to minutes (Fig. 1C, central trace). E. coli can reverse this process as well, reverting from sticking to rolling and even detaching (Fig. 1C, bottom trace). Thus, under flow conditions, bacteria exhibited a novel type of surface adhesion that we hereafter refer to as ‘stick-and-roll’ adhesion (illustrated in the movie, Fig.S4 of the supplementary material).

Accumulation of bacteria on the 1 M-coated surface under flow conditions

The ability of bacteria to colonize a surface depends on how many bacteria accumulate over time, which reflects the balance between the attachment rate and the duration of binding. In order to determine how shear affects accumulation, we passed E. coli expressing the FimH-f18 variant over 1 M-BSA surfaces for 5 min before counting accumulated bacteria in both the rolling and stationary modes. At low shear (0.01–0.05 pN µm⁻²), E. coli failed to accumulate in significant numbers (Fig. 2A) presumably because they detach so rapidly in these conditions as shown above. However, an increasing number of bacteria accumulated in both the rolling and stationary modes on 1 M-BSA surfaces when the shear stress was above 0.1 pN µm⁻², resulting in an over 100-fold higher number of bound bacteria at around 0.4 pN µm⁻² (Fig. 2A). At even higher shear stress, accumulation decreased again, so only 8% as many bound bacteria were observed by 4.0 pN µm⁻² as at a 10-fold lower level of shear stress. This presumably happened because of the failure of bacteria to attach initially under such high shear, as shown above. The fraction of bacteria that were stationary increased with shear stress (Fig. 2A, inset). Thus, a biphasic accumulation curve was observed whether both rolling and stationary or just the stationary bacteria were counted. This reflects a trade-off between the shear-inhibition of initial attachment and the shear-enhancement of binding duration that results from an increased conversion to the stationary mode of adhesion at higher shear.

Figure 2. Accumulation of bacteria expressing FimH-f18 on 1 M-BSA and control surfaces. Surfaces coated with the indicated proteins were placed in a flow chamber through which bacteria were washed at the indicated shear, and the number of E. coli that had accumulated was measured at the end of 5 min. A. Accumulation of bacteria on 1 M-BSA (○), and the number of accumulated bacteria in the stationary mode of adhesion (•). The number of rolling bacteria is indicated by the shaded difference between the two curves. Insert: the ratio of stationary to total attached bacteria after 5 min of accumulation. In this figure, the stationary bacteria were determined by comparing two images taken 1 s apart. B. Accumulation of E. coli on anti-FimH antibodies to measure binding on slip-bonds (□) and on galactose-BSA to measure non-specific binding (◊). No rolling bacteria were observed on either surface, so the stationary bacteria are the same as the total bacteria and are not shown. The error bars in all graphs give 95% confidence intervals.

To determine whether the biphasic accumulation curve results from a specific nature of the FimH-mannose bond, we tested E. coli accumulation on surfaces coated with anti-FimH antibodies. In contrast to FimH-mediated binding to 1 M, when anti-FimH antibodies were presented on the surface, the accumulation of bacteria was always inhibited by shear (Fig. 2B, squares). Most of the bacteria remained bound long term on anti-FimH surfaces at all shear levels (not shown), so that the accumulation curve reflects the initial rate of attachment. Furthermore, the ‘stick-and-roll’ mode of adhesion was not observed for bacteria binding to anti-FimH antibodies, because at all shear levels all bacteria bound exclusively in the stationary mode. The fact that shear has the opposite effect on binding to antibodies as on binding to 1 M-BSA demonstrates that a special property unique to the FimH-mannose bond, rather than other mechanical or transport properties of the system, is responsible for the ability of shear stress to enhance accumulation.

Together, the increase in accumulation up to 0.4 pN µm⁻² and the switch to stationary behaviour above 1 pN µm⁻² demonstrate that the ability of the low-1 M-binding FimH variant of E. coli to bind to a surface coated with a purified 1 M protein is enhanced by shear stress. This suggests that the shear-dependent adhesion of RBC to the carpet of bacteria expressing the FimH-f18 variant described in our previous publication (Thomas et al., 2002) results from the shear-enhanced interaction of FimH with 1 M-containing receptors on the RBC surface.

Effect of FimH structural variations on 1 M-binding under flow

Using the flow chamber assay with a 1 M-BSA-coated surface, we characterized the flow-dependent binding capability of bacteria expressing a different FimH variant that originated in the uropathogenic E. coli strain J96. FimH-j96 contains an A27V replacement that enhances its ability to bind 1 M under static conditions (Sokurenko et al., 1997; 1998). In the previous study, a carpet of bacteria expressing this FimH variant did not show shear activation in the binding of RBCs, as the RBCs did not move along the bacterial carpet even at low shear (Thomas et al., 2002).

In contrast to the previous RBC binding assay, the current 1 M-binding flow assay demonstrated similar patterns of shear-activation in E. coli expressing FimH-j96 as in those expressing FimH-f18. Both variants displayed a bimodal effect of shear on accumulation (Fig. 3A) and an increase in the stationary mode of adhesion with shear (Fig. 3B). Bacteria expressing FimH-j96 accumulated on the 1 M surface at a significantly higher rate than those expressing FimH-f18 throughout most of the shear stress range. However, because the former strain bound an order of magnitude better to 1 M under the lowest shear used, the relative effect of shear on accumulation at the peak was less dramatic for FimH-j96 (14-fold increase) than for FimH-f18 (100-fold increase). Moreover, the surface accumulation of FimH-j96 bacteria appeared to peak slightly earlier than that of the FimH-f18 strain. Finally, the reduced shear dependence of the FimH-j96 strain was demonstrated by its higher fraction of stationary bacteria at lower shear levels. Thus, the flow chamber 1 M-binding assay is clearly more sensitive than the RBC/bacterial carpet binding assay in detecting and comparing shear-activated phenotypes of FimH variants. It showed that the naturally occurring high-1 M-binding FimH variants, which are predominant among uropathogenic strains, are capable of a shear-activated interaction with 1 M-coated surface in a pattern somewhat related (but not identical) to the one demonstrated by low-1 M-binding variants common among intestinal E. coli isolates.

Figure 3. The effect of mutations in FimH on the shear dependence of binding. A and B. The effect of shear stress on the binding of E. coli expressing FimH-f18 (•) or FimH-j96 (▪). C and D. The effect of the S124A/Q32L mutations (▵) and of the V156P mutation (□) on the shear dependence of binding of FimH-j96. The S124A/Q32L mutations were predicted earlier to inhibit a force-activated state of FimH, while the V156P mutation was predicted to enhance this state (Thomas et al., 2002). E and F. The effect of the V156P mutation (◊) on the shear dependence of binding of FimH-f18 (•). A, C and E. One aspect of binding is the number of E. coli that accumulated on 1 M-BSA over 5 min. B, D and F. Another aspect of binding is the fraction of these bacteria that were stationary at the end of 5 min. The methods in this figure are the same as those used in Fig. 2 and the error bars in all graphs give 95% confidence intervals.

The A27V mutation in FimH-j96 is positioned in one of the loops of the FimH lectin (mannose-binding) domain that stabilizes the linker chain connecting it to the pilin (fimbriae-incorporating) domain. We hypothesized earlier that one of the structural effects of shear stress on a receptor-bound FimH protein is extension of the strand linking the lectin and pilin domains, and that this linker chain extension is associated with tighter binding. It was also shown that the V156P mutation within the linker chain, which was predicted to facilitate the linker extension by the tensile force, converts the shear-activated FimH-f18 variant into a phenotype in which shear dependence could not be demonstrated in the RBC binding assay (Thomas et al., 2002). To see whether or not this V156P mutation renders FimH shear independent, we tested the effect of this mutation in both the FimH-f18 and FimH-j96 variants in the more sensitive 1M-binding flow chamber assay. Similar to the A27V mutation, the V156P mutation increased the rate of accumulation at most shear levels (Fig. 3C and E). It also caused the shear stress required for peak accumulation to shift to lower levels and favoured a transition from rolling to stationary adhesion at lower shear levels (Fig. 3D and F). The high similarity between the functional effects of A27V and V156P mutations supports the hypothesis suggested previously (Thomas et al., 2002) that the naturally occurring A27V mutation also facilitates the linker chain extension under tensile force. Interestingly, the increased sensitivity of the 1 M-binding assay allowed the observation that the FimH-j96-V156P mutation shows even stronger low shear binding than did the FimH-f18-V156P mutation. This indicates a cumulative effect of A27V and V156P mutation on the linker chain extension rather than one mutation being functionally dominant over the other.

Finally, in the previous study, two other mutations, Q32L and S124A, were predicted to inhibit the linker extension by diverting force away from the hydrogen bonds that anchor the linker chain to the neighbouring loop regions of the lectin domain. When these mutations were introduced into FimH-j96, they decreased the accumulation rate, shifted the accumulation peak to higher shear stress (Fig. 3C) and greatly favoured the rolling over the stationary state (Fig. 3D) except at very high shear stress. Thus, the phenotype of the Q32L:S124A mutant of FimH-j96 variant was similar to that exhibited by the low-1 M-binding FimH-f18 with pronounced shear-activated properties. In contrast, other mutations did not have effects on the shear dependence; for example, a double mutation that is observed in many faecal isolates, S70N:N78S, had no effect on the shear dependence of binding of fimH-f18 to 1 M surfaces whether measured by accumulation or fraction stationary. Thus, taken together these observations validated further the hypothesis that a force-induced linker extension under shear stress is a functionally significant event leading to enhanced 1 M-binding of the FimH adhesin.

Changes in shear stress

In nature, fluid flow often changes over time in regular or more random fashion. In order to contrast how bacteria that are bound via FimH to 1 M versus to anti-FimH antibodies respond to changes in flow, we bound FimH-f18-expressing bacteria to surfaces coated with anti-FimH or 1 M-BSA for 5 min at the optimal shear for peak accumulation and then subjected them to changes in shear stress. Most of the bacteria bound to anti-FimH remained bound when the shear stress was reduced, but detached within 1 min after a sudden increase in shear stress (Fig. 4A, squares). This behaviour is typical of many previously studied bacterial adhesion systems (Dickinson et al., 1997). In contrast, bacteria bound to 1 M-BSA remained bound upon the switch to higher shear stress but, remarkably, most of them detached at the lower shear stress (Fig. 4A). That is, even after 5 min of accumulation at the optimal shear stress, bacteria were bound reversibly to the surface and detached when the shear stress was turned down. Thus, shear stress has the opposite effect on detachment of E. coli from 1 M-BSA as from anti-FimH surfaces.

Figure 4. Response of bound bacteria expressing FimH-f18 to changes in shear stress. A. The effect of shear stress on the fraction of accumulated bacteria that detached after 1 min from a surface coated with 1 M-BSA (○) or with anti-FimH antibody (▪). This experiment was performed on about 200 bacteria that were accumulated at 0.01 pN µm⁻² on anti-FimH and at 0.5 pN µm⁻² on 1 M-BSA as these were the optimal shears for accumulation. The error bars give 95% confidence intervals, and are not visible where they are smaller than the size of the markers. B. The effect of a reduction in shear stress from 0.5 pN µm⁻² to 0.01 pN µm⁻² on the number of bacteria still bound over the first 10 s after the change. The bacteria that were rolling just before the change in shear (○) were analysed separately from those that were stationary (▵) by tracking the movement of each bacterium over time. C. The effect of a reduction in shear stress from 0.5 pN µm⁻² to 0.01 pN µm⁻² (grey line), or of an increase from 0.5 pN µm⁻² to 3 pN µm⁻² (black line), on the number of bacteria adhering in the stationary mode for 5 min after the change. Here, the stationary mode was determined by leaving the shutter open long enough to blur out rolling and free-floating bacteria. D, E. The effect of a change from 0.4 pN µm⁻² to 2 pN µm⁻² and back to 0.4 pN µm⁻² on the fraction of rolling bacteria bound to 1 M-BSA was determined by comparing images taken 1 s apart. In (D) the shear stress was increased by raising the fluid flow rate fivefold while in (E) it was increased by raising the fluid viscosity fivefold with 10% polyethylene glycol, which took about 10 s to enter the field of view.

To determine how changes in shear stress affect the stationary and rolling modes of adhesion, we observed the behaviour of bacteria bound to 1 M-BSA in more detail when the shear stress was altered. After the shear stress was switched from the accumulation-optimal level (0.5 pN µm⁻²) to a lower level (0.01 pN µm⁻²), the bacteria that were initially rolling detached within 1 s, while those that were initially stationary detached more slowly (Fig. 4B), by converting briefly to the rolling mode (not shown). Nevertheless, over several minutes the number of stationary bound bacteria decreased dramatically (Fig. 4C). Thus, both modes of E. coli binding to 1 M-BSA surfaces are reversible at lower shear. When the shear stress was instead increased (3.0 pN µm⁻²), not only do all bacteria remain bound to the 1 M-BSA surface, but the number of stationary bacteria actually increased suddenly (Fig. 4C, black line). This jump in stationary bacteria reflected a shear-induced conversion of the rolling bacteria to firmly adherent bacteria, as illustrated in the movie Fig.S5 in the supplementary material. The percentage of rolling bacteria decreased instantly from ≈ 50% at the accumulation-optimal shear (0.4 pN µm⁻²) to essentially zero when the shear was increased to 2.0 pN µm⁻² (Fig. 4D). When bacteria were kept at high shear, none of them detached or even rolled long distances for at least 30 min, although individual bacteria occasionally jolted forward short distances before becoming stationary again.

This transition from rolling to stationary was also reversible, as demonstrated by the onset of some rolling behaviour upon a return to moderate shear stress (Fig. 4D). The higher fraction of bacteria rolling in the first versus the second period of moderate shear probably reflects their history, or behaviour before these periods, because bacteria transition slowly between the stationary and rolling modes. In the first period, many bacteria were newly bound, and initial binding was usually in the rolling mode, while in the second period, all bacteria were recently at high shear and thus stationary. When individual bacteria were tracked while the shear was switched to high shear and back to moderate, it was sometimes the same and sometimes different bacteria that rolled during the two periods of the original shear. This indicates that rolling and stationary adhesion do not reflect different subpopulations of bacteria, but rather different modes of adhesion that together characterize the FimH−1 M interaction. Thus E. coli can bind tightly to 1 M-coated surfaces to resist removal at high flow but then can revert to a rolling mode when the flow switches to a moderate one.

Finally, in order to verify that the shear-induced transition from rolling to stationary adhesion of bacteria results from mechanical drag-force acting on bacteria and not from transport effects caused by increased fluid velocity, we subjected the surface-bound bacteria to a fivefold more viscous solution instead of a fivefold faster flow rate (Fig. 4E). While both methods increased the drag force on the bacteria fivefold, the viscosity method had no effect on fluid velocity. The results for both methods were essentially identical except that the transition was more gradual and slightly delayed when the viscosity was increased (Fig. 4E), reflecting the flow of the more viscous fluid from the tubing junction into the field of view. Thus, the increased drag force resulting from higher shear stress, rather than increased transport from the higher flow rate, caused the transition from rolling to stationary adhesion.

Discussion

Here, we demonstrated that shear stress enhanced the accumulation of bacteria on the 1 M-coated surface by causing a switch from transient to long-term adhesion and that this switch was reversible. The behaviours described in this article are in many ways contrary to what has been expected for bacterial adhesion. While some former studies have suggested that shear may slightly enhance bacterial binding (Brooks and Trust, 1983; Li et al., 2000; Mohamed et al., 2000), it has not been reported previously that a lack of shear could prevent bacterial surface accumulation or cause bacterial detachment. As we observed here, a shear stress of 0.5 pN µm⁻² increased E. coli accumulation on 1 M surfaces by two orders of magnitude, and was required to maintain binding. This level of shear is well within the physiological range in many compartments in the human body, which has been estimated at up to 3 or 4 pN µm⁻² (30–40 dynes cm⁻²) in the blood, and 0.1 pN µm⁻² (1 dyne cm⁻²) in the saliva (Prakobphol et al., 1995). Our previous study (Thomas et al., 2002) utilized large RBCs that were already bound to bacteria. It could thus not have been predicted either the effect of shear on the rate of accumulation or the effect of shear on the behaviour of the much smaller bacteria, both of which are addressed by this study. The more detailed analysis allowed by this new assay showed that shear stress increased the duration of bacterial surface binding in two ways – by preventing rolling bacteria from detaching as they do at low shear and by favouring a transition from rolling to stationary adhesion. The rolling of bacteria along the adhesive surface is another important observation. To our knowledge, this rolling adhesion mode has not been described for bacteria previously.

The shear-dependent surface rolling appears to be a specific property of the FimH–mannose interaction, because it was not observed when the bacteria bound to anti-FimH-coated surfaces. Moreover, the rolling and stationary behaviour observed in this work suggests that FimH has two functional states that correlate with the two structural conformations predicted earlier in steered molecular dynamics simulations. Figure 5 shows the emerging model for how shear stress affects the adhesion of E. coli to 1 M surfaces. At low shear (Fig. 5A), most FimH molecules that interact with 1 M are in the unstressed equilibrium state with the interdomain linker chain tightly incorportated into the rest of the lectin domain as shown in Fig. 5D. This FimH state forms weak (short-lived) bonds with 1 M that often break before new ones can form, so that the bacteria detach from the 1 M-coated surface at a high rate. The fact that at moderate shear many bacteria roll forward instead of detaching (Fig. 5B) may reflect an increase in the rate of new bond formation with faster flow. Faster flow could rotate the bacterium forward more quickly, bringing new fimbriae into contact with the surface sooner. The bonds must still be relatively short-lived to mediate rapid rolling (Chang et al., 2000). We showed earlier that mechanical force favours an alternative conformational state of FimH, in which the interdomain linker chain is disassociated from the rest of the lectin domain as shown in Fig. 5E. We propose here that this conformational state forms a strong long-lasting bond with 1 M (in a manner yet to be determined) that causes the bacterium to adhere firmly for the duration of the bond (Fig. 5C). The observation that FimH-f18-expressing bacteria roll long distances at moderate shear before stopping suggests that FimH-f18 usually detaches from mannose faster than the linker chain extends at moderate shear, so that the strong state is rarely observed. By this model, the occasional abrupt switching back and forth between rolling and stationary adhesion at moderate shear thus reflects the stochastic manner in which individual bonds convert to the strong state and eventually break. However, as the shear increases, the increased drag force on the bacterium facilitates the linker chain extension (Fig. 5E) and favours the long-lasting bond state. Thus, virtually all bacteria convert immediately from rolling to stationary adhesion at high shear stress.

Figure 5. How shear stress changes mode of adhesion. This cartoon provides a model for how shear stress might affect the state of the FimH-1 M bonds and thus the modes of adhesion of E. coli bacteria to a monomannose surface. A. At low shear, the bacteria detach because short-lived bonds break more quickly than new bonds form; the broken bonds are indicated by a space between the hair-like fimbriae and the mannose circle. B. At moderate shear, bacteria are more likely to roll downstream as short-lived bonds break, because new bonds already exist. C. With low probability at moderate shear and high probability at high shear, a bacterium abruptly stops rolling and sticks firmly as one or more bonds become very long lived. D. In the absence of mechanical force, the structure of FimH was shown previously to be compact. This structure was shown to be associated with weak binding (Thomas et al., 2002) and may form the short-lived bonds observed at low and/or moderate shear. E. Mechanical force can stretch the FimH lectin, or binding, domain, causing the C-terminal linker chain that connects to the pilin, or anchoring, domain to extend. This was shown previously to be associated with strong binding (Thomas et al., 2002) and may form the long-lived bonds that cause the bacteria to abruptly halt and remain stationary for the lifetime of that bond.

In contrast, it is difficult to explain the switch from rolling to stationary adhesion through transport phenomena (i.e. changes in bond formation rates). First, the bacteria make this transition even when the fluid viscosity is changed but the flow rate is not. Second, when the bacteria stop abruptly, they reduce speed by over 1000-fold (from ≈ 30 µm s⁻¹ to < 1 µm min⁻¹), which cannot be explained by existing models for how bond number affects rolling velocity (Chen and Springer, 1999). Finally, this abrupt stopping and restarting occur when the shear stress is unchanged. This stochastic change cannot readily be explained by large changes in bond numbers. The switch to stationary adhesion also cannot be explained by hypothesizing the existence of active fimbrial retraction or additional adhesins, as the requirement of shear stress for stationary adhesion was recently replicated in cell-free assays in which polystyrene beads coated with 1M-BSA were bound to surfaces coated with type 1 fimbriae (Forero et al., in press). Thus, the data are more consistent with the proposition that FimH is a catch-bond that lasts longer at high tensile force than at moderate force because it is more likely to switch to a long-lived, or ‘high-affinity’ state.

The degree of shear activation varied dramatically between different structural variants of FimH, with both the engineered and naturally occurring mutations in FimH altering the probability that bacteria become stationary at a given shear stress. Mutations that were predicted to suppress the linker chain extension (Q32L and S124A) favoured the rolling over the stationary behaviour, as expected if they decrease the rate of the switch from the weak to strong bond state. On the other hand, mutations like A27V and V156P favour stationary adhesion over rolling as expected if they destabilize the linker chain and enhance its extension. The V156P mutation appears to destabilize the linker chain to the extent that the strong state is dominant even without significant force in static conditions. In these variants, however, the shear activation of the binding was still obvious, in particular for the naturally occurring A27V FimH variant. This was not observed in our earlier publication (Thomas et al., 2002), demonstrating that the 1 M-based flow chamber assay is more sensitive to showing shear activation. This may be because the small bacteria bind to 1 M-coated surface through a much smaller number of bonds than do the large RBCs binding to the bacterial carpet. Overall, these data and those in our previous report (Thomas et al., 2002) suggest that the ability of mechanical force to extend the linker chain and, thus, induce the strong-binding state of FimH can depend on the structural mutations in the adhesin.

To approximate the force on the bacterium in conditions that result in rolling or stationary adhesion, we can use Goldman's approximation that the drag force on a sphere tethered to a surface is 1.7·6πr²τ, where r is the radius of the sphere, and τ is the wall shear stress (Goldman et al., 1967). As fimbriated bacteria could be considered spheres with a radius of 1 µm, we calculated a drag force of 32 µm² times τ, so the drag force on each bacterium would be 0.3 pN at τ = 0.01 pN µm⁻² where the bacteria detach, 13 pN at τ = 0.4 pN µm⁻² where many of the bacteria are rolling and 64 pN at τ = 2.0 pN µm⁻² where the bacteria all become stationary.

Interestingly, surface rolling is a well-described phenomenon for leukocytes migrating over the surface of endothelial cells via adhesion proteins called selectins. The purpose of surface rolling for the leukocytes is presumably to maintain a near-wall position and velocity appropriate for patrolling blood vessel walls at low and high flow rates alike. Similar to what we have observed with bacteria expressing the FimH-f18 variant, leukocytes show a shear threshold for surface rolling, below which cells do not attach (Finger et al., 1996; Lawrence et al., 1997; Konstantopoulos et al., 2003). However, rolling leukocytes and microspheres that bind via selectins have been shown to speed up if exposed to higher shear, rather than to convert to a stationary mode of adhesion (Chen and Springer, 1999; Greenberg et al., 2000). To mediate firm adhesion of leukocytes, either a mutationally modified form of the selectin (Salas et al., 2002) or additional proteins (Campbell et al., 1998) were required. In contrast, we show that bacteria that are rolling on FimH-mannose bonds require only higher shear to stick firmly. Also, the shear threshold for leukocyte rolling was shown to involve transport phenomena in the form of an increase in the number of initial attachments (Alon et al., 1997; Chen and Springer, 2001) with the increase in the shear rate, but not the shear stress (Chen and Springer, 2001). In contrast, we report here that shear inhibits initial attachments for bacteria binding to mannose (Fig. 1A). Still, both the selectin and FimH systems display a shear-enhanced transition from transient binding to rolling adhesion (Alon et al., 1997; Chen and Springer, 1999; Figs 1 and 4). For the selectin system, this has been associated with a greater bond number with increased shear (Chen and Springer, 1999), reflecting either an increase bond on-rate or mechanical effects that decrease off-rates. The notion of decreased off-rates has just recently been supported by atomic force microscope experiments that demonstrate that P- and L-selectins can form catch-bonds with their PSGL-1 ligand that increase in lifetime as force increases (Marshall et al., 2003; Sarangapani et al., 2004). Regardless of the cause of the transition from transient binding to rolling adhesion for FimH-mediated bacterial adhesion, the model described here proposes that the ability of FimH to form catch-bonds gives the best explanation for the shear-induced transition from rolling to stationary adhesion. This transition of rolling to stationary adhesion is not observed at all with the selectin-mediated systems, although they have been studied under similar levels of physiological shear stress. Thus, while the selectin and FimH systems both display a shear threshold that can be attributed at least in part to catch-bonds, the existing data and models suggest substantial differences in these two systems.

The shear-enhanced interaction between the FimH adhesin and surface monomannose allows bacteria to display a far more complex adhesive behaviour than if it would be mediated by conventional slip-bonds. We suggested earlier that one function of the shear-enhanced property of FimH may be to bind in the presence of soluble inhibitors, because the low-1 M-binding shear-activated FimH variants are more resistant to inhibition by soluble mannose than the high-1 M-binding FimH variants that do not depend on the shear-activation for binding to the same extent (Thomas et al., 2002). In this study, we demonstrate that another advantage of the shear-enhanced character of FimH is to allow a ‘stick-and-roll’ adhesion in which bacteria switch from detaching to rolling to firm adhesion as shear stress increases. This ‘stick-and-roll’ ability may protect bacteria from detachment during high shear fluctuations caused by salivary washing, intestinal peristalsis, pumping of blood, emptying of the urinary bladder, etc. At the same time, E. coli could roll or even detach from the surface to spread into different body compartments if the flow decreases, or if they roll into a stagnant microenvironment. This ability could also allow bacteria to remain preferentially in high shear microenvironments that offer the most nutrient flow or, alternatively, to spread into novel compartments more efficiently at low flow. For the low-1 M-binding FimH variants (like FimH-f18) that are typical for the faecal isolates, the shear thresholds for a switch between detaching, rolling and stationary phenotypes might be well optimized for the intestinal ecology of E. coli. Possibly this adhesive phenotype is required for the type 1 fimbriated E. coli bacteria to move successfully from its transient niche in the oropharynx, through the stomach barrier into the intestine during the faecal–oral transmission. Another possibility is that stick-and-roll adhesion allows E. coli to move into high shear microenvironments of the convoluted intestinal surface and thus to maximize nutrient flow. In contrast, the naturally occurring high-1 M-binding variants that are common among uropathogens have a less pronounced dependence on the shear stress for binding and this phenotype might be optimized for the colonization of urinary tract. This may reflect the requirement of uropathogenic E. coli to bind to tissues displaying monomannose even during the long periods of low flow between brief times of urination. In addition, these bacteria may have little preference for high shear regions because they may ultimately get nutrition through invasion rather than by filtration from the fluid, as FimH has been shown to mediate invasion of bladder epithelial cells by uropathogenic E. coli (Martinez et al. 2000; Mulvey et al., 2000). The residual shear dependence among the urinary tract isolates such as FimH-j96 may reflect relatively recent evolution from intestinal isolates, or the possibility that stick-and-roll adhesion in a scaled-down version has advantages in the urinary tract, for example, allowing E. coli to migrate up the urethra between voiding while preventing detachment during voiding. However, the physiological importance of the stick-and-roll adhesion for the commensal and pathogenic E. coli remains to be investigated.

The FimH adhesin is the most common and well-studied type of bacterial adhesin, yet the phenomenon of shear-dependent ‘stick-and-roll’ adhesion mediated by FimH has just been revealed. Indeed, decades of work have demonstrated strong adhesion of FimH-expressing E. coli to eukaryotic cells in static conditions. This is presumably because these studies reported this behaviour in the uropathogenic strains (Hagberg et al., 1981) that commonly contain high-1 M-binding FimH variants, or used host cell lines that express proteins with oligomannose-like carbohydrate modifications. In both cases, high binding under static conditions is expected. Perhaps most importantly, the vast majority of studies on receptor-specific bacterial adhesion did not utilize variable flow conditions as done here. This suggests that studying other bacterial adhesins under the proper flow conditions may reveal shear-enhanced characteristics similar to ones found in FimH, with significant implications for how bacteria colonize tissue surfaces.

Experimental procedures

Preparation of protein-coated surfaces

Monomannosylated BSA (1 M-BSA) and galactosylated BSA were purchased from EY Laboratories, Inc (San Mateo, CA). Dishes coated with purified components were prepared by incubating 35 mm tissue culture dishes with 100 µl of either 200 µg ml⁻¹ 1 M-BSA, 200 µg ml⁻¹ galactosylated BSA or a 50 000-fold dilution of polyclonal anti-FimH antiserum in 0.02 M bicarbonate buffer for 75 min at 37°C. These were then washed three times with phosphate-buffered saline with 0.1% bovine serum albumin (PBS-BSA) at room temperature to remove unbound 1 M-BSA, galactosylated BSA or antibodies, and to block any remaining sites on the tissue culture plastic that could non-specifically bind bacteria later.

Preparation of bacterial solutions

Bacteria expressing the indicated strains of FimH were constructed and prepared through overnight growth as described previously (Thomas et al., 2002), and brought to 10⁸ cfu ml⁻¹ in PBS-BSA. In the studies of surfaces coated with anti-FimH antibodies, 2% alpha-monomannose was added to the PBS-BSA in order to inhibit any potential mannose-specific interaction between the bacteria and any carbohydrate modifications on the antibodies. An addition of 10% polyethylene glycol with a molecular weight of 8000 (PEG-8000, SIGMA) was used when indicated to increase the viscosity fivefold from 1 cP to 5 cP, as determined using a Brookfield digital viscometer model DV-II.

Parallel plate flow chamber experiments

All experiments were performed at room temperature in PBS-BSA buffer. The protein-coated dishes were inserted into a 2.5 cm × 0.25 cm × 250 µm parallel plate flow chamber (GlycoTech). The solution of bacteria was flown through the chamber using one or more Harvard syringe pumps at various flow rates to produce various wall shear rates. The bound bacteria were recorded using a Nikon TE200 inverted microscope with a 10× phase-contrast objective, a Roper Scientific high-resolution CCD camera and MetaMorph^® (Universal Imaging Corporation^TM) video acquisition software. The field of view was 500 µm by 370 µm with a resolution of 0.77 µm per pixel. Even when 400 bacteria bound per field of view, they covered only 1% of the surface area, so that any interactions between bacteria were minimal. The bound bacteria were recorded in time-lapse digital videos. Unless noted otherwise, the camera shutter was held open for a time that depended upon the shear rate so that free-floating bacteria were barely blurred out. At a shear stress of 0.0125 pN µm⁻², the shear rate was 12.5 s⁻¹, so fimbriated bacteria with a radius of around 1 µm would move about 12.5 µm s⁻¹ when floating freely at the surface. Leaving the shutter open for 750 ms caused these bacteria to appear as streaks of about 9 µm, or about 12 pixels long, that were faint enough to be ignored by the autocounting and autotracking software. The shutter speed was increased linearly with the shear rate, so that by 0.5 pN µm⁻² (a shear rate of 510 s⁻¹), the shutter was closed after 18 ms.

When we used two flow rates during an experiment, bacteria were bound to the 1 M surface and washed, i.e. free-floating bacteria were removed from the chamber by switching to a bacteria-free buffer at the initial flow rate.

Analysing the number of bacteria bound in flow chamber videos

The digital videos were analysed using MetaMorph^® imaging software. Bacteria appeared as dark spots of about two pixels diameter on a light background. To count bound bacteria, we manually determined an intensity threshold that correctly distinguished the bacteria from the background and used the automated cell counting package in MetaMorph^®.

We defined bacteria as rolling if they moved at least two pixels (one-half to one bacterial diameter, or about 1.5 µm) in 1 s. The number of rolling bacteria at a given time was identified by subtracting the intensity values in each pixel in the second image from the first; the stationary bacteria then blended in with the grey background while the rolling bacteria appeared as a pair of dark and light spots, which were counted using a threshold and the automated cell counting package. The rest of the bound bacteria were then assumed to be stationary.

The rolling bacteria could also be distinguished from the bound bacteria that detached from the surface during the observation. Because bacteria generally rolled far below the hydrodynamic velocity, they could easily be seen as round or oblong dark spots in the blended image, while those moving at the hydrodynamic velocity were nearly invisible grey streaks. Also, when noted, the tracking method was used instead, in which the positions of bacteria were tracked over many sequential images using either the manual point tracking package or the automated object tracking package in MetaMorph^®.

Statistical analysis

When the number of attachments or accumulated bacteria was measured, the exact 95% confidence intervals for a Poisson variable were used as long as the measured number, n, was less than 100. For higher numbers, the normal approximation (the square root of n) was used for the standard deviation. When instead a fraction was measured such as the fraction of bacteria that detached after 1 min, the exact 95% confidence intervals for a binomial variable were used as long as the total sample number, n, was less than 100. For n > 100, pn > 5 and qn > 5 (where p is the fraction chosen and q the fraction not chosen), the standard deviation was taken to be the square root of pq/n.

Acknowledgements

We would like to thank Dr Henry Hess and Mark Troll for their fruitful discussions and comments on the manuscript. The research was supported by the NIH Bioengineering Research Partnerships (BRP) Grant ♯ 1 R01 AI 50940 and NIH Grants ♯ R01 AI45820 and ♯ P01 DK53369 (E.V.S.), a Whitaker Foundation Graduate Fellowship (W.T.), an NSF fellowship ♯ DGE9616736AMO (L.N.), and a University of Washington ‘University Initiative Fund’ Center for Nanotechnology fellowship (M.F.).

Supplementary material

The following material is available from http://www.blackwellpublishing.com/products/journals/suppmat/mmi/mmi4226/mmi4226sm.htm

Five videos are presented here to illustrate qualitatively how E. coli expressing FimH-f18 bind and move across 1 M-BSA-coated surfaces in a flow chamber in several flow conditions. In all videos, the bacteria appear as dark spots on a light background in the phase contrast images, while the unbound cells are moving too fast to be seen (see Experimental procedures). Also in all videos, the field of view is greatly reduced (192 by 270 µm) from that used in the analysis.

Fig.S1. Attachment and detachment of bacteria at low shear stress.

Fig.S2. Attachment and detachment of bacteria at moderate shear stress.

Fig.S3. Attachment and detachment of bacteria at moderate shear stress.

Fig.S4. Movement of bacteria bound to 1 M-BSA surfaces at moderate shear to illustrate ‘stick-and-roll’ bacterial adhesion.

Fig.S5. Abrupt stopping of bacteria upon a switch to high shear stress.