Post-translational modification of the deubiquitinating enzyme otubain 1 modulates active RhoA levels and susceptibility to Yersinia invasion


B. M. Kessler, Henry Wellcome Building for Molecular Physiology, Nuffield Department of Clinical Medicine, University of Oxford, Roosevelt Drive, Oxford OX3 7BN, UK
Fax: +44 1865 287 787
Tel: +44 1865 287 799


Microbial pathogens exploit the ubiquitin system to facilitate infection and manipulate the immune responses of the host. In this study, susceptibility to Yersinia enterocolitica and Yersinia pseudotuberculosis invasion was found to be increased upon overexpression of the deubiquitinating enzyme otubain 1 (OTUB1), a member of the ovarian tumour domain-containing protein family. Conversely, OTUB1 knockdown interfered with Yersinia invasion in HEK293T cells as well as in primary monocytes. This effect was attributed to a modulation of bacterial uptake. We demonstrate that the Yersinia-encoded virulence factor YpkA (YopO) kinase interacts with a post-translationally modified form of OTUB1 that contains multiple phosphorylation sites. OTUB1, YpkA and the small GTPase ras homologue gene family member A (RhoA) were found to be part of the same protein complex, suggesting that RhoA levels are modulated by OTUB1. Our results show that OTUB1 is able to stabilize active RhoA prior to invasion, which is concomitant with an increase in bacterial uptake. This effect is modulated by post-translational modifications of OTUB1, suggesting a new entry point for manipulating Yersinia interactions with the host.

Structured digital abstract


hemagglutinin-tagged ubiquitin-bromide


multiplicity of infection


otubain 1


ras-related C3 botulinum toxin substrate 1


ras homolog gene family member A


ubiquitin-specific protease


Yersinia outer protein


Yersinia serine/threonine kinase


The genus Yersinia consists of three pathogenic species that are agents of a variety of diseases, one of which was historically the cause of major pandemics. These include the bubonic plague caused by Yersinia pestis, mesenteric adenitis and septicaemia caused by Yersinia pseudotuberculosis and gastroenteritis caused by Yersinia enterocolitica [1]. Even though the plague is not a major health concern today, cases are reported annually. Moreover, Y. pestis was weaponized in the former Soviet Union [2] and there are reports of emerging multidrug resistant strains [3]. Pathogenic Yersiniae are typically taken up through ingestion and first reach the intestine. The Yersinia surface protein invasin binds to β1 integrins on the apical surface of M cells, which facilitates translocation across the epithelium [4,5]. The pathogenicity and virulence of Yersiniae is mainly based on the plasmid-encoded type III secretion system that encodes for six effector proteins, which are injected into the host cell (primarily monocytes) to modulate the physiology of the infected cell and to prevent uptake and killing (reviewed in [6]). An additional chromosomally encoded Ysa type III secretion system has been described in Y. enterocolitica [7,8]. The injection of effector proteins promotes Yersinia growth and survival in lymphoid follicles (Peyer’s patches) underlying the intestinal epithelium and controls antibacterial activities of immune cells located at these sites. Four of these Yersinia outer proteins (Yops) are engaged in modifying the cytoskeleton: YopE, YopH, YopT and YpkA [9–11]. YpkA, an essential virulence factor, is a serine/threonine kinase that phosphorylates actin [12], binds the deubiquitinating enzyme otubain 1 (OTUB1) [13,14], the small G protein subunit Gαq [15] and interacts with members of the Rho family of small GTPases, ras homologue gene family member A (RhoA) and ras-related C3 botulinum toxin substrate 1 (Rac1) [16]. Although the interaction with actin, in particular G-actin, has been shown to be crucial for YpkA serine/threonine kinase activity, the functional relevance of the interaction with OTUB1 remains to be determined [12,13,17]. YpkA-mediated phosphorylation of Gαq impairs guanine nucleotide binding and subsequently inhibits Gαq-mediated signalling pathways including RhoA activation and cytoskeletal rearrangements in the host cell [15]. In addition, a crystallography-based study revealed that YpkA mimics host guanine nucleotide dissociation inhibitors (GDIs), thereby blocking nucleotide exchange in RhoA and Rac1, a process that is crucial for virulence in Yersinia [18]. YpkA therefore uses several ways to interfere with the function of small GTPases, which appears to be essential for Yersinia pathogenesis [19].

The Rho family of small G proteins represents a large group of the Ras superfamily of GTPases. More than 20 proteins of this class have been described to date, among which RhoA, Rac1 and Cdc42 are well characterized, particularly their role in cytoskeletal regulation. Specifically, RhoA is involved in the formation of stress fibres and focal adhesion complexes [20–23]. Yersinia is not the only pathogen that affects the function of small GTPases such as RhoA [24], indicating that interference with the function of small GTPases is of prime importance in bacterial pathogenesis because microbes have evolved a number of virulence factors that modulate the function of these proteins.

In this study, we show for the first time that susceptibility to bacterial invasion by Yersinia can be altered by changing expression of otubain 1 (OTUB1), a host cell-encoded deubiquitinating enzyme that belongs to the ovarian tumour domain-containing protein family. This effect is dependent on the catalytic activity of OTUB1 and its ability to stabilize the active form of RhoA prior to invasion. YpkA and OTUB1 modulate the stability of RhoA in opposing ways, therefore leading to cytoskeletal rearrangements that may be involved in bacterial uptake. During this process, OTUB1 was found to be phosphorylated, a post-translational modification that modulates its ability to stabilize RhoA. These findings provide a novel entry point for the manipulation of host cell interactions with Yersinia and perhaps other enterobacteria by deubiquitination.


OTUB1 controls cell susceptibility to Yersinia invasion

Yersinia virulence factors are injected into target host cell molecules to manipulate signalling pathways during invasion in order to prevent uptake and killing. In addition to actin, other host cell proteins have been shown to bind to the virulence factor YpkA, including OTUB1 [13]. In order to investigate the role of OTUB1 in Yersinia invasion of HEK293T cells, we established a cell culture invasion assay, in which the effects of overexpression and knockdown of OTUB1 could be monitored. Bacterial uptake into HEK293T cells was measured using a gentamicin-based invasion assay (Fig. 1). Cells transfected with wild-type OTUB1 were infected with Y. enterocolitica and the number of intracellular bacteria compared with the quantity observed in cells overexpressing either a catalytically inactive mutant C91S or an empty vector. We observed that susceptibility to Y. enterocolitica invasion was significantly increased upon overexpression of wild-type OTUB1 in HEK293T cells, an effect that was not seen when the catalytically inactive OTUB1 mutant (C91S) was expressed (Fig. 1A). A marked increase in susceptibility was also observed upon overexpression of OTUB1 and invasion with Y. pseudotuberculosis. Conversely, OTUB1 knockdown significantly attenuates Yersinia invasion (Fig. 1B). We repeated the OTUB1 knockdown experiment in primary human monocytes, which are among the first cells targeted for Yersinia invasion in vivo, and this also resulted in decreased invasion efficiency (Fig. 1C). These differences could not be accounted for by changes in cell viability or cell growth given the controls and time frame of the experiment. To confirm the initial observation by an alternate method, we used a double fluorescence staining technique that enables visualization of extracellular and intracellular bacteria in the same cell [25]. The results concurred with the data from the gentamicin-based invasion assay. The ratio of intracellular to extracellular bacteria was much higher in the case of cells overexpressing OTUB1 compared with control cells or cells overexpressing a catalytically inactive mutant of OTUB1 (CS91S, Fig. 2A). Increased susceptibility to Yersinia in the presence of overexpressed OTUB1 was observed as early as 15 min after invasion, and decreased over time, probably because of intracellular elimination. Taken together, our results indicated that it was the efficiency of bacterial uptake, not the proliferation of bacteria within the host cell that is modulated by OTUB1 (Fig. 2B).

Figure 1.

 OTUB1 controls susceptibility to invasion by Yersinia enterocolitica. (A) HEK293T cells were transfected with empty vector (EV), wild-type OTUB1-HA or the C91S mutant, followed by invasion with Yersinia enterocolitica (MOI 60 : 1). Gentamicin was added after 1 h to kill extracellular bacteria. After 2 h, cells were lysed and dilutions plated and cultured for 2 days at 27 °C. Susceptibility to invasion was measured as the ratio between the numbers of colonies for OTUB1 (black bar), C91S mutant (grey bar) relative to the number obtained in the control (white bar, set to as 1.0). Ten independent experiments were performed and the P-values are displayed as calculated using the Student’s t-test. The mean and standard deviations of the absolute numbers of observed colonies are indicated. (B) Number of colonies obtained relative to control when HEK293T cells were either transfected with empty vector (EV, white bar), transfected with negative scrambled control (sc, grey bar) or OTUB1 shRNA (black bar) for 24 h prior to Yersinia invasion. Six independent experiments were performed and the P-values are displayed, as calculated using the Student’s t-test. The mean and standard deviations of absolute numbers of the observed colonies are indicated. (C) Number of colonies obtained relative to control from primary monocytes that were previously isolated from human peripheral blood mononuclear cells and were either transfected with negative scrambled control or transfected with OTUB1 shRNA for 24 h (black bar) prior to invasion with Yersinia. Four independent experiments were performed and the P-values are displayed, as calculated using the Student’s t-test. The mean and standard deviations of the absolute numbers of observed colonies are indicated.

Figure 2.

 OTUB1 expression levels affect bacterial uptake but not intracellular proliferation. (A) HEK293T cells were transfected either with empty vector (EV), wild-type OTUB1-HA or the C91S mutant and after 24 h infected with Yersinia pseudotuberculosis (MOI 60 : 1) for 15, 30 and 60 min, followed by fixing and staining for extracellular bacteria using fluorescein isothiocyanate (FITC)-labelled Yersinia antibodies (green). Cells were then permeabilized and stained with tetramethyl rhodamine iso-thiocyanate (TRITC)-labelled Yersinia antibodies to label intracellular bacteria (red), followed by analysis using confocal microscopy. Pictures of the 30-min time point are shown. Control cells (upper, EV) and cells overexpressing OTUB1-HA (lower, OTUB1) have different ratios of intracellular (tetramethyl rhodamine iso-thiocyanate-stained, lower left compartment) versus extracellular bacteria (fluorescein isothiocyanate-stained, upper right compartment). The nuclei were visualized using 4′,6-diamidino-2-phenylindole staining (blue). (B) OTUB1-HA-overexpressing cells are characterized by a higher ratio of intracellular/extracellular bacteria in comparison with OTUB1-HA C91S mutant or control cells. This difference occurred as early as 15 min after invasion with Yersinia. Three independent experiments were performed for the statistical analysis, and relative ratios between intracellular (red) versus total/extracellular (green) bacteria are shown as well as the P-values calculated using the Student’s t-test.

Post-translationally modified OTUB1 interacts with the virulence factor YpkA

Previous evidence suggested that the Yersinia-encoded virulence factor YpkA interacts with OTUB1 in vitro [13], providing a potential molecular entry point to explain this effect. We therefore aimed to validate this result and examine whether this interaction also occurs during bacterial invasion in living cells. To test whether YpkA interacts with OTUB1, wild-type YpkA and an inactive kinase mutant D267A were overexpressed in HEK293T cells, followed by YpkA immunoprecipitation and separation by SDS/PAGE. This was compared with a control immunoprecipitate from cells transfected with empty vector, and the presence of OTUB1 was assessed by immunoblotting. We observed that endogenous OTUB1 and YpkA are part of the same protein complex (Fig. 3A). Inactivation of the YpkA kinase activity by a D267A mutation did not abolish this interaction. Moreover, this interaction was also observed with endogenous YpkA present in host cells during bacterial invasion (Fig. 3B). We noted that multiple forms of OTUB1 can be detected, as described previously [26,27], and that the form of OTUB1 that co-immunoprecipitated with YpkA has an apparent molecular mass of 37 kDa, corroborating the findings of a previous study [13]. However, the majority of endogenous OTUB1 protein is detected at its expected molecular mass, 31 kDa (Fig. 3A, left). We also observed increased levels of this higher molecular mass form of OTUB1 in infected HEK293T cells compared with control (Fig. 3C). Nevertheless, the appearance of this form did not depend on YpkA kinase activity (Fig. 3D). We therefore examined whether this corresponds to the previously identified alternative spliced form of OTUB1 referred to as ARF-1, which has an apparent molecular mass of 35 kDa [26]. Overexpression of HSV-tagged ARF-1 was detected by anti-HSV, but not by OTUB1 immunoblotting, indicating that our antibody does not recognize ARF-1 (Fig. S1). We therefore hypothesized that this form of OTUB1 may be post-translationally modified, leading to a change in apparent molecular mass and enhancing interaction with YpkA. Consistent with this, treatment with protein phosphatase suggested that the 37 kDa form of OTUB1 may contain multiple phosphorylation sites, based on the observed differential migration pattern (Fig. 3E). To further shed light on the role of these OTUB1 modifications in the invasion process, we embarked on identification using a tandem mass spectrometry approach (LC-MS/MS). Endogenous OTUB1 was isolated from HEK293T cells, separated by SDS/PAGE and the stained material subjected to in-gel trypsin digestion and analysis by LC-MS/MS (Fig. 4A). An OTUB1-derived N-terminal peptide containing three phosphorylation sites, Ser16, Ser18 And Tyr26 was identified. In addition, OTUB1 which was overexpressed in HEK293T cells was isolated and analysed in a similar manner, revealing a different N-terminal peptide that contained the same phosphorylated residues (Fig. 4B). Based on these results, OTUB1 mutants were generated in which Ser16, Ser18 and Tyr26 were replaced with glutamic acid in order to mimic the negative charge caused by phosphorylation (S16E, S18E and Y26E). This approach was successfully used to imitate phospho-serine and -threonine residues, but is to some extent less ideal for phospho-tyrosines [28]. Interestingly, we observed that the OTUB1 Y26E and S18E mutants exerted increased affinity to YpkA in co-immunoprecipitation experiments, thereby resembling the increased binding of the 37 kDa form of OTUB1 to YpkA (Fig. 5A). This is consistent with the notion that phosphorylation of OTUB1 affects the interaction with YpkA, although the regulation might be more complex, because the OTUB1 S16E/S18E/Y26E triple mutant did not show any increased binding to YpkA.

Figure 3.

 Interaction between OTUB1 and YpkA in living cells and during Yersinia invasion. (A) Empty vector (EV), wild-type YpkA-FLAG, or the YpkA-FLAG inactive kinase mutant D267A were transfected into HEK293T cells. After 24 h, cell extracts were prepared and YpkA material immunoprecipitated using anti-FLAG Ig. Association with endogenous OTUB1 was demonstrated by immunoblotting using OTUB1 antibodies in the presence of YpkA wild-type and D267A inactive kinase mutant. (B) HEK293T cells were infected with Y. pseudotuberculosis for 2 h. Cell extracts were prepared and YpkA immunoprecipitated using YpkA antibodies. In infected cells, association with endogenous OTUB1 was demonstrated by anti-OTUB1 immunoblotting (hc, heavy chain; lc, light chain; *, a smaller form of OTUB1 was also detected). (C) Modification of OTUB1 during Yersinia invasion. HEK293T were infected with Y. enterocolitica at an MOI of 10 : 1 for 2 h, followed by cell lysis, separation by SDS/PAGE and anti-OTUB1 immunoblotting. (D) Modification of OTUB1 does not depend on YpkA kinase activity. HEK293T cells were left untreated or infected with Y. pseudotuberculosis wild-type (wt) and YpkA kinase inactive mutant at an MOI of 10 : 1 for 2 h. Cell extracts were prepared and two forms of OTUB1 (31 kDa unmodified form and 37 kDa modified form) were visualized using OTUB1 antibodies. (E) OTUB1 37 kDa form is phosphorylated. HEK293T cells were lysed and incubated with calf intestinal phosphatase (CIP) for 1 h, resulting in the appearance of multiple forms between 27 and 37 kDa, which indicates the presence of several phosphorylation sites. OTUB1 was visualized using anti-OTUB1 immunoblotting. Two independent experiments are shown.

Figure 4.

 Detection of OTUB1 phosphorylation using MS. (A) Detection of endogenous phosphorylated OTUB1. HEK293T cells were lysed, followed by immunoprecipitation of OTUB1. As a control, lysate was incubated with agarose without the antibody. Immunoprecipitated material was analysed by SDS/PAGE and silver staining, and the large band corresponding to the expected molecular mass of OTUB1 as well as the area above (rectangle) was excised and digested with trypsin. Digested material was analysed by a nano-LC Ion Trap mass spectrometer. For the peptide 14–36 ([M + 2H]2+, 934.1 Da) containing the phosphorylated tyrosine and two serines, the b- and y-fragment ion series are shown. (B) Detection of phosphorylated OTUB1 in an overexpression model. Control or HEK293T cells overexpressing OTUB1-HA wild-type were lysed, followed by immunoprecipitation of OTUB1. Eluted material was analysed by SDS/PAGE gel and Coomassie Blue staining, and the band corresponding to a modified OTUB1 (rectangle) was excised and digested with trypsin. The peptide mixture was analysed by a nano-UPLC-QTOF tandem mass spectrometer. For the peptide 13-36 ([M + 2H]2+, 966.6 Da) containing the phosphorylated tyrosine and two serines, the b- and y-fragment ion series detected are shown.

Figure 5.

 OTUB1 modification controls its function and its effect on Yersinia invasion. (A) Binding of YpkA to OTUB1 depends on OTUB1 modification. Empty vector (EV control), OTUB1-HA wild-type, catalytically inactive mutant (C91S) or mutants mimicking phosphorylated OTUB1-HA (S16E, S18E, Y26E) were co-expressed with YpkA-FLAG in HEK293T cells. Cells were lysed and YpkA-FLAG immunoprecipitated with anti-FLAG Ig. Binding of OTUB1 mutants to YpkA was measured by immunoblotting using HA antibodies. OTUB1 expression levels as well as the loading control (PDI) were shown in the input, whereas YpkA-FLAG was visualized in immmunoprecipitated material. One representative out of three experiments is shown. (B) OTUB1 modification affects bacterial invasion. HEK293T cells were transfected either with empty vector (EV control), wild-type OTUB1-HA or mutants mimicking phosphorylated OTUB1 listed in Fig. 5A followed by invasion with Y. enterocolitica (MOI 60 : 1). Gentamicin was added after 1 h to kill extracellular bacteria. After 2 h, cells were lysed and dilutions plated and cultured for 2 days at 27 °C. The number of colonies for OTUB1 and the OTUB1 mutants were counted and presented relative to the number obtained for the control (EV). The P-values were calculated using a Student’s t-test. Expression of OTUB1 in infected cells is shown using anti-OTUB1 western blotting and the loading control using anti-PDI western blotting. (C) Mimicry of phosphorylation on Tyr 26 interferes with OTUB1 function. HEK293T cells were transfected either with empty vector (EV), HA-tagged wild-type OTUB1, catalytically inactive OTUB1 C91S, mutants mimicking phosphorylated OTUB1 (see above) or the Y26F mutant. Cells were lysed and extracts incubated with an HA-tagged ubiquitin Br2 probe to measure OTUB1 activity as described previously [30]. As a control, cells were treated the same way but without addition of the probe. OTUB1 and OTUB1–probe adduct were visualized by immunoblotting using HA antibodies and quantified for two experiments (black and grey bars). The intensities of the corresponding bands were measured and the ratio between them is shown (labelling ratio), reflecting reactivity towards the probe. Two independent experiments are shown.

Mimicry of OTUB1 phosphorylation modulates susceptibility to Yersinia invasion

If the interaction between OTUB1 and YpkA were relevant for increased susceptibility to invasion, one would expect that modification of OTUB1 may have an effect on this process. To examine this, we repeated the gentamicin-based invasion assay with cells overexpressing the OTUB1 mutants that mimic phosphorylation. Overexpression of the OTUB1 mutants S16E, S18E, Y26E, S16E/S18E and S16E/S18E/Y26E abolished the observed increase in susceptibility to invasion seen with wild-type OTUB1 (Fig. 5B) or the S16A and Y26F control mutants (data not shown), thereby confirming that modification of OTUB1 has an impact on the magnitude of Yersinia invasion. Because no effect on invasion was seen with the catalytically inactive mutant C91S OTUB1 (Fig. 1), we set out to test whether the constructed proteins mimicking phosphorylated OTUB1 were functional by monitoring their reaction with the deubiquitinating enzyme-specific probe, hemagglutinin-tagged ubiquitin-bromide (HA-Ub-Br2), which was previously shown to covalently bind active OTUB1 [29,30]. Interestingly, the OTUB1 Y26E mutant did not react with the HA-Ub-Br2 active-site probe, whereas all other mutants were able to do so (Fig. 5C). We conclude that phosphorylation of OTUB1, in particular at Tyr26, modulates OTUB1 function by interfering with its enzymatic activity, ubiquitin binding or substrate recognition. Next, we examined whether OTUB1 phosphorylation may be attributed to the Ser/Thr kinase activity of YpkA directly. Recombinant OTUB1 and immunoprecipitated YpkA expressed in HEK293T cells were incubated in a radioactive in vitro kinase assay. Recombinant OTUB1 was weakly phosphorylated by YpkA, consistent with previous findings, but to a much lesser degree than the control protein myelic basic protein (Fig. S2A). By contrast, OTUB1 isolated from cell lysates was not phosphorylated by YpkA at a detectable level, although wild-type YpkA was readily autophosphorylated and therefore active (Fig. S2B). These results indicate that modification of OTUB1 by phosphorylation has an effect on OTUB1-mediated Yersinia bacterial uptake, but did not resolve the relevance of YpkA’s Ser/Thr kinase activity in this process.

OTUB1-mediated susceptibility to invasion is modulated by the YpkA GTPase-binding domain

YpkA consists of several domains including a serine/threonine kinase and a GTPase-binding domain, both of which contribute to virulence [6] (Fig. 6A). In order to dissect which of these functionalities contribute to OTUB1-mediated susceptibility to invasion, we used Yersinia strains that either had mutations in the kinase (ypkAD270A) or GTPase-binding domain (Yersinia contact A mutant strain) [18]. OTUB1-mediated susceptibility to invasion with the Yersinia ypkAD270A strain was unaltered, but was compromised with the contact A mutant strain (Fig. 6B). These results show that the YpkA GTPase-binding domain, but not the Ser/Thr kinase activity, interferes with susceptibility to Yersinia invasion provoked by overexpression of OTUB1 in host cells.

Figure 6.

 OTUB1-mediated susceptibility to invasion requires YpkA and its GTPase-binding domain, but not its serine/threonine kinase activity. (A) Scheme of the domains present in YpkA. Mutated amino acid positions in the mutant strains used in this study are indicated. (B) Increased susceptibility to Yersinia invasion is not dependent on YpkA-mediated phosphorylation. Control HEK293T cells, HEK293T cells overexpressing OTUB1-HA or OTUB1-HA C91S were infected with either wild-type Y. pseudotuberculosis or Y. pseudotuberculosis mutants containing an inactive kinase domain (D270A) or the YpkA contact A mutant (unable to bind GTPases). Cells were collected after 3 h, lysed and cell extracts plated on agar plates. Colonies were counted after 2 days of incubation at 27 °C and the colony numbers were displayed as ratios relative to the control. Experiments were performed at least three times and the P-values were calculated using a Student’s t-test. For each strain, susceptibility to invasion was measured as the ratio between the numbers of colonies for OTUB1 (black bar), C91S mutant (grey bar) relative to the number obtained in untransfected cells (EV, white bar, set to as 1.0). (C) YpkA is in a complex with RhoA and OTUB1. Cells were co-transfected with OTUB1-HA, RhoA-myc and YpkA-FLAG, followed by immunoprecipitation with HA or FLAG antibodies. OTUB1-HA, RhoA-myc and YpkA-FLAG were visualized by immunoblotting using OTUB1, RhoA or FLAG antibodies, respectively.

Previous experiments have demonstrated an interaction between YpkA and the small GTPases RhoA or Rac1 [16,18]. Our data suggest that the ability of YpkA to bind GTPases may be critical for the OTUB1-mediated increased Yersinia uptake. We therefore tested whether YpkA and RhoA interact in vitro and whether this protein complex includes OTUB1. YpkA was immunoprecipitated and the presence of OTUB1 and RhoA examined by immunoblotting (Fig. 6C). YpkA, OTUB1 and RhoA were found to be part of the same complex. Moreover, OTUB1 is associated with RhoA in the absence of YpkA, as demonstrated by co-immunoprecipitation of OTUB1 and RhoA (Fig. 6C, lane 3).

OTUB1 stabilizes active RhoA

The existence of all three components in the same complex and the association between OTUB1 and RhoA suggested that OTUB1 might play a role in modulating the ubiquitination status and stability of RhoA. In order to investigate this, we expressed both proteins in HEK293T cells and examined the polyubiquitination status and the stability of RhoA by immunoprecipitation/western blotting experiments (Fig. 7A–C). When OTUB1 was overexpressed, the total amount of RhoA increased marginally. The same observation was made for endogenous RhoA levels which were elevated upon overexpression of OTUB1 (Fig. 7A). However, levels of endogenous active (GTP-bound) RhoA isolated from noninfected cells using a rhotekin-based pulldown were stabilized considerably by OTUB1, but not by a catalytically inactive OTUB1 C91S mutant (Fig. 7B). This was not accounted for by an increase in RhoA activation through its guanine nucleotide exchange factor LARG, for which a marginal increase was noted in the presence of wild-type and catalytically inactive OTUB1 (Fig. 7B, lower). A more striking effect was observed when immunoprecipitated RhoA was incubated with recombinant OTUB1 in vitro (Fig. 7C). The experiment was performed by expressing a constitutively active RhoA (Q63L mutant) to enrich for polyubiquitinated material. The quantity of ubiquitinated RhoA was significantly decreased in presence of wild-type OTUB1, whereas levels of unmodified RhoA increased with time. The catalytically inactive mutant OTUB1 C91S was unable to deubiquitinate RhoA (Fig. 7C right). These results clearly indicate that OTUB1 is responsible for stabilization of active RhoA and that it is dependent on the deubiquitinating activity of the enzyme (Fig. 7B).

Figure 7.

 OTUB1 stabilizes active RhoA. (A) Protein lysates from HEK293T cells co-transfected with RhoA wild-type and OTUB1-HA wild-type or control plasmid (EV) were subjected to RhoA detection by immunoblotting. Levels of RhoA (transfected RhoA-myc, left; endogenous RhoA, right) were increased if cells were co-transfected with OTUB1-HA but not in the presence of empty vector (EV). Loading control is shown using anti-PDI western blotting. (B) OTUB1 stabilizes active RhoA. Endogenous active RhoA was isolated using Rhotekin-coupled beads from HEK293T cells overexpressing either empty vector (EV), OTUB1-HA wild-type or C91S mutant. LARG, PDI (loading control) and OTUB1-HA wild-type were visualized using western blotting of the input material. (C) OTUB1 deubiquitinates RhoA in vitro. Purified ubiquitylated RhoA isolated from HEK293T cells previously transfected with the constitutively active mutant RhoA-myc QL63 was incubated with recombinant wild-type OTUB1 (both panels) the catalytically inactive mutant C91S (right) for 0, 15 and 30 min at 37 °C. RhoA deubiquitination was visualized by anti-ubiquitin and anti-RhoA immunoblotting. (D) OTUB1-mediated stabilization of active RhoA is impaired during invasion and if co-expressed with YpkA. Active RhoA was enriched using recombinant Rhotekin from HEK293T cells overexpressing empty vector (EV), OTUB1-HA wild-type, infected or not with Y. pseudotuberculosis for 3 h (left) or from HEK293T cells overexpressing YpkA-FLAG alone or together with OTUB1-HA or the control plasmids (HA and FLAG plasmids, right). The loading control (PDI), OTUB1-HA and YpkA-FLAG were visualized by western blotting of the input material. (E) Mimicry of OTUB1 phosphorylation impairs its ability to stabilize active RhoA. Active RhoA was enriched using recombinant Rhotekin from HEK293T cells overexpressing empty vector (EV), OTUB1-HA wild-type or the mutants S16E, S18E, Y26E, S16E/S18E and S16E/S18E/Y26E mimicking OTUB1 phosphorylation. The loading control (PDI) and OTUB1-HA were visualized by western blotting of the input material. Moreover, RhoA in the flow through material (ft) was also visualized. One out of two experiments is shown.

Correlation between RhoA stabilization and enhanced susceptibility to Yersinia invasion

Because RhoA has been shown previously to be implicated in modulating host–pathogen interactions by regulating cell morphology and uptake [31,32], our results raised the question of whether OTUB1-mediated enhanced susceptibility to invasion may involve RhoA. To examine this in further detail, we first tested whether levels of the GDP- or GTP-bound form of RhoA are affected during invasion. A rhotekin-based pulldown assay was used to isolate the active form of RhoA from infected and noninfected cells. The amount of active RhoA is substantially increased when OTUB1 was overexpressed, but not during invasion (Fig. 7D). Therefore, overexpression of OTUB1 does stabilize active RhoA prior to, but not after, invasion. Co-transfection experiments revealed that YpkA alone counteracts OTUB1-mediated stabilization of RhoA (Fig. 7D), therefore identifying two factors that have an opposing effect on RhoA function and stability. Finally, to underscore the relevance of OTUB1-mediated stabilization of RhoA in enhanced susceptibility to invasion, we tested whether OTUB1 mutants mimicking phosphorylation were able to stabilize active RhoA (Fig. 7E). Overexpression of the OTUB1 mutants S16E, S18E, Y26E, S16E/S18E and S16E/S18E/Y26E did not rescue active RhoA levels to the same extent as observed with wild-type OTUB1, thereby corroborating their effect on enhanced susceptibility to invasion (Fig. 5B).


This study describes a role for the host cell-encoded deubiquitinating enzyme, OTUB1, in modulating cell susceptibility to bacterial invasion. OTUB1 has been shown to disassemble lys48-linked polyubiquitin chains [14,29,33], and is involved in anergy induction in CD4+ lymphocytes through its interaction with the E3 ligase gene related to anergy induction in lymphocytes (GRAIL) and ubiquitin-specific protease (USP)8 [26]. However, OTUB1 is expressed ubiquitously in most tissues, which suggests its involvement in other cell biological processes not restricted to lymphoid tissues. Indeed, OTUB1 was suggested to stabilize estrogen receptor alpha levels in breast and endometrial cancer cells [27]. In addition, previous evidence suggested that OTUB1 may be linked to Yersinia invasion based on its reported interaction with the Yersinia-encoded virulence factor YpkA, and it was also proposed that OTUB1 may be a substrate for its serine/threonine kinase activity, at least in vitro [13]. Consistent with this, we found YpkA to be present in the same protein complex as OTUB1 in living cells and during bacterial invasion, as assessed by co-immunoprecipitation. Moreover, we confirmed that YpkA can phosphorylate OTUB1 in vitro (Fig. S1A). As observed previously, we also noted that the form of OTUB1 interacting with YpkA has an approximate molecular mass of 37 kDa, which is different from its expected molecular mass of 31 kDa (Fig. 3). Our data demonstrate that endogenous OTUB1 is modified by phosphorylation in living cells (Fig. 4). However, our results question the in vivo relevance of OTUB1 phosphorylation by YpkA which has been observed in vitro. First, a 37 kDa form of OTUB1 can be detected in addition to its normally expected size at 31 kDa in HEK293T cells independently of bacterial invasion (Fig. 3A, input material) or YpkA expression (Fig. 3C,E). Second, OTUB1, as a 37 kDa polypeptide, was also found in a complex with the inactive kinase mutant YpkA D267A (Fig. 3A). Third, invasion with the Yersinia mutant strain expressing an inactive YpkA kinase (D270A) did not affect OTUB1-mediated susceptibility to invasion (Fig. 6B). Fourth, we did not observe YpkA-mediated phosphorylation of OTUB1 that was isolated from HEK293T cells, but noted a slight increase in YpkA autophosphorylation in the presence of OTUB1 (Fig. 1B, lower). Our results indicate that OTUB1 phosphorylation is an YpkA-independent event that is, however, crucial for their interaction. Further investigation using MS confirmed the presence of three phosphorylated residues in endogenous and overexpressed OTUB1 isolated from HEK293T cells (Fig. 4) consistent with the fact that multiple endogenous forms of OTUB1 were observed (Fig. 3E). Sequence analysis did not reveal any typical kinase consensus sites, so it is currently unknown what physiological process and which kinases are involved in OTUB1 phosphorylation. These modifications alone do not fully account for the apparent molecular mass shift observed with the 37 kDa form of endogenous OTUB1 (see Fig. 4, left), indicating that this form may harbour additional post-translational modifications that escaped our detection. However, OTUB1 mutants mimicking phosphorylation appear to have similar biochemical properties as the naturally occurring 37 kDa form of OTUB1, in particular the S18E and Y26E variants, both of which exert increased affinity to YpkA. The effect of these modifications on OTUB1 binding to YpkA did not fully account for loss of increased susceptibility to bacterial invasion. Alternatively, these modifications may also change OTUB1 deubiquitination activity, affinity to or recognition of substrates. Consistent with this, we observed that only wild-type OTUB1 was able to stabilize active RhoA, whereas mutations mimicking phosphorylation abrogated this effect (Fig. 7E). Modification of Y26 in OTUB1 interfered with active site labelling by the Ub-Br2 probe, suggesting impaired deubiquitination function, whereas mutations at positions S16 and S18 may alter substrate binding. The lack of stabilizing active RhoA by OTUB1 mutants correlated with their inability to sustain susceptibility to invasion (compare Figs 5B and 7E), indicating that controlling active RhoA levels is important for the magnitude of invasion, in line with previous observations [32]. Phosphorylation of deubiquitinating enzymes may be common and has been observed previously for USP7 [34], USP8 [35] and CYLD, the latter of which is also functionally altered through this modification [36].

YpkA is a multifunctional protein that interferes with host cell functions at several levels during Yersinia invasion [19]. In addition to its serine/threonine kinase activity [13,15,37] and binding to actin [12], YpkA has been shown to interact with small GTPases and inhibit nucleotide exchange in Rac1 and RhoA, mimicking the guanidine nucleotide dissociation inhibitors of the host [16,18]. Full virulence of Yersinia depends on all of these properties mediated by YpkA, because mutations or deletions in either the kinase or GTP-binding domains reduce the pathogenicity of these strains [18,37,38]. By contrast, null mutations in ypkA in Y. pseudotuberculosis appear to be similar to wild-type in their virulence, a trait that is thought to result from a possible compensatory mechanism evolving in this strain [18,39,40]. Our findings are consistent with the former observation in that the wild-type strain had the highest invasion efficiency (data not shown), and our results suggest a link between the GTPase-binding capacity of YpkA and OTUB1-mediated increase in susceptibility to infection (Fig. 6). The binding of small GTPases has been reported to be independent of the kinase activity of YpkA [16]. In line with this, we detected RhoA in immunopurified YpkA and OTUB1 complexes. We also noted an interaction between OTUB1 and RhoA in the absence of YpkA (Fig. 6C) and therefore hypothesized that RhoA may be a substrate for deubiquitination by OTUB1, a process that may be modulated by YpkA during invasion. Indeed, we showed that RhoA is stabilized by OTUB1 (Fig. 7A–D). This can be achieved either by induction of RhoA mRNA expression, deubiquitination of RhoA itself or promoting activation of RhoA, possibly through the manipulation of RhoA-specific guanine nucleotide exchange factors. Real-time PCR analysis showed that OTUB1 overexpression did not alter RhoA mRNA levels (Fig. S3). Distinguishing between the two latter possibilities proved to be more challenging. Our results indicate that RhoA can be deubiquitinated by OTUB1 directly, but not by the catalytically inactive mutant (Fig. 7C). This is supported by the fact that the catalytically inactive OTUB1 mutant also fails to stabilize RhoA (Fig. 7B). However, we cannot exclude any other mechanisms, such as deubiquitination and stabilization of RhoA-specific guanine nucleotide exchange factors, although LARG does not appear to be significantly affected by OTUB1 (Fig. 7B) [41]. YpkA interferes with this process by reducing active RhoA levels (Fig. 7D). In general, YpkA does not appear to have any effect on OTUB1 deubiquitinating activity (data not shown), but it may bind and sequester post-translationally modified OTUB1 and GDP bound RhoA to interfere with active RhoA formation, and perhaps provoke premature degradation of RhoA during bacterial invasion.

Our results indicate that in absence of bacterial invasion OTUB1 prolongs the lifetime of the active (GTP-bound) form of RhoA, because this form is rapidly ubiquitinated and turned over [42,43]. Increased susceptibility to invasion provoked by OTUB1 overexpression seems to be dependent on stabilization of active RhoA by OTUB1 prior to bacterial invasion. The accumulated pool of active RhoA contributes to an enhanced uptake in the early phase of invasion, consistent with the involvement of the microtubule system and GTPases in this process [32]. Once YpkA is present in the infected host cell, further active RhoA formation is blocked (Fig. 7D) [6], which may prevent further bacterial uptake. Interestingly, we observed high levels of OTUB1-mediated bacterial uptake that decreased after prolonged invasion times (Fig. 2B). This may reflect a decrease in the efficiency of bacterial uptake once intracellular bacteria are present possibly combined with intracellular elimination. RhoA as well as Rac1 and Cdc42 are involved in modulating cytoskeletal rearrangements and endocytosis [44], further confirming that OTUB1-mediated stabilization of RhoA could affect bacterial entry into the host. In line with this, OTUB1 overexpression appears to also affect the stability of Rac1 and Cdc42 (unpublished data). YpkA and other Yersinia-encoded virulence factors target small GTPases to limit bacterial uptake in order to prevent internalization and killing. In addition, a different GTPase targeted by YpkA-mediated phosphorylation, Gαq, may also be implicated in limiting bacterial uptake [15,45]. YpkA may therefore use both its kinase and guanine nucleotide dissociation inhibitor domains to interfere with RhoA activation more effectively [19].

In summary, our findings reveal a new aspect of the complex interplay in host–pathogen interactions and demonstrate a physiological role of the deubiquitinating enzyme OTUB1 in Yersinia invasion. OTUB1 as a potential key player in regulating RhoA stability may represent a novel pharmacological target for yersiniosis, but may also be linked to the biology of RhoA-mediated regulation of cell morphology, adhesion and migration in general.

Materials and methods

Cell lines and reagents

The HEK293T cell line was maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum, 1% glutamine and 100 μg·mL−1 streptomycin and penicillin in a humidified atmosphere of 5% CO2 at 37 °C. Primary peripheral blood mononuclear cells were prepared from buffy coats (National Blood Centre, London, UK) using a standard lymphoprep-based method (Axis-Shield PoC AS, Dundee, UK) and subsequent isolation of monocytes using CD14 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) was carried out according to the manufacturer’s instructions. Chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA), unless indicated otherwise. The antibodies used in this study are described in the Supporting information.

DNA constructs

The cDNA for human OTUB1 and OTUB1-HA C91S was obtained as described previously [29]. The OTUB1-HA S16E, S16A S18E, Y26E, Y26F, S16E/S18E and S16E/S18E/Y26E mutants were created using the QuikChange II site-directed mutagenesis kit by Stratagene (La Jolla, CA, USA). The initial OTUB1 mutants S16E, S16A, S18E, Y26E and Y26F were generated using the OTUB1 pcDNA 3.1 construct containing a C-terminal HA-TEV-SBP tag and the primers described in Table S1. The S16E/S18E double mutant was generated using the OTUB1 S16E mutant construct as a template. The S16E/S18E/Y26E triple mutant was generated using the OTUB1 S16E/S18E construct as a template. All OTUB1 mutant constructs were verified by sequencing.

The siRNAs specific for OTUB1 and a negative control (scrambled, SI 03650318, All Stars negative control) were purchased from Qiagen (Crawley, UK) and tested for their ability to knockdown endogenous OTUB1 (data not shown). The best siRNA (OTUB1_3 SI00676053) that has no reported off-target effects (information by the manufacturer) was used for this study. The RhoA-specific constructs RhoA-myc L63 (Q63L) and wild-type RhoA (both in pEXV Amp-R) were generously provided by M. Olson (Glasgow, UK). The YpkA-FLAG wild-type and D267A constructs were a kind gift from L. Navarro and J.E. Dixon (UCLA, Los Angeles, CA, USA). The OTUB1 ARF-1 construct was a gift from C.G. Fathman (Stanford University, Palo Alto, CA, USA).

Yersinia strains

The Yersinia strains used in this study were Yersinia pseudotuberculosis YPIII (pIB102) wt (Km-R), YPIII (pIB44), YPIII (pIB47) YpkA D270A (Tc-R), contact A mutant strain of Yersinia pseudotuberculosis (IP2777) containing Y591A, N595A and E599A mutations and Yersinia enterocolitica (Ye 8081). Strains were cultured in Lysogeny broth (Sigma-Aldrich, St Louis, MO, USA) at 27 °C overnight at 200 rpm.

Transfection assays

HEK293T cells were grown to confluence in DMEM containing 10% fetal bovine serum, 1% glutamine and 1% penicillin/streptomycin. Cells were then transferred to 150, 100, 50 mm or six-well tissue culture dishes at a concentration of 0.4 × 106 mL−1 and grown overnight at 37 °C. The cells were then washed with NaCl/Pi and the transfection was performed using SuperFect reagent (Qiagen), according to the manufacturer’s protocol, followed by an overnight incubation at 37 °C. For experiments in which overexpressed proteins were subsequently deubiquitylated in vitro, cells were treated with 10 μm proteasome inhibitor MG132 (Sigma-Aldrich) for 6 h in order to interfere with proteasomal degradation and accumulate polyubiquitylated proteins. For siRNA studies, cells were prepared in a similar way as for gene overexpression, but transfections were performed with HiperFect (Qiagen), and the cells were grown for 24 h at 37 °C.

Primary monocytes isolated from a buffy coat (National Blood Centre) were transfected using the Amaxa Nucleofector system (Amaxa/Lonza, Cologne, Germany) with the Human Monocyte Nucleofector Kit (Amaxa) according to the instructions provided by the manufacturer, and grown for 10 h post-transfection in Human Monocyte Nucleofector Medium (Amaxa).

Bacterial invasion assay

HEK293T cells (5 × 106 per sample) were incubated for 12–16 h after transfection, washed with NaCl/Pi and incubated in DMEM without fetal bovine serum and antibiotics during invasion. A Y. pseudotuberculosis or Y. enterocolitica overnight culture at 27 °C was diluted 1 : 20 and incubated further for 2 h at 27 °C, and then for 1 h at 37 °C. Bacteria were washed in NaCl/Pi and used to infect cells at an multiplicity of infection (MOI) of 60 : 1 for 1 h at 37 °C. In order to obtain comparable MOIs for the Y. pseudotuberculosis wild-type and different mutant strains, dilutions series of bacterial cultures were used to infect cells. Thereafter, cells were washed three times with NaCl/Pi and further incubated for 2 h at 37 °C in DMEM supplemented with 10% fetal bovine serum and 100 μg·mL−1 gentamicin in order to kill extracellularly located bacteria. Invasion of primary monocytes and U937 cells were conducted in the same way using RPMI-1640 medium. In order to measure cell susceptibility to invasion, cells were lysed in 0.5% NP-40, 150 mm NaCl, 5 mm CaCl2, 50 mm Tris pH 7.4, and the dilutions were plated Yersinia selective agar base (Sigma-Aldrich) containing Yersinia selective supplement (Sigma-Aldrich) and cultured for 2 days at 27 °C. Colony numbers were counted and a statistical analysis (Student’s t-test) was performed using sigmaplot software (Systat Software Inc, Salisbury, UK).

Immunoblotting and immunoprecipitation

Cells (5 × 106 per sample) were lysed in 0.1% NP-40, 150 mm NaCl, 20 mm CaCl2, 50 mm Tris pH 7.4 containing a protease inhibitor cocktail (Roche Applied Science, Basel, Switzerland). Samples were separated by SDS/PAGE and subjected to immunoblotting as described in the Supporting information. For immunoprecipitation protein lysates (5 mg per sample) were first diluted in NET buffer (50 mm Tris, 5 mm EDTA, 150 mm NaCl, 0.5% NP-40, pH 7.4) to a protein concentration of 1 mg·mL−1, and pre-cleared with agarose-coupled Protein A beads (Sigma-Aldrich) for 1 h at 4 °C. Immunoprecipitation was then carried out either for 2 h or overnight at 4 °C.

Analysis of OTUB1 by tandem MS

For analysis of the endogenous OTUB1, HEK293T cells were grown to confluence in DMEM in 175 cm2 tissue culture flasks. The cells were then washed in NaCl/Pi, lysed in 0.1% NP-40, 150 mm NaCl, 20 mm CaCl2, 50 mm Tris pH 7.4 containing protease inhibitor cocktail (Roche Applied Science) and 100 μm sodium orthovanadate (Sigma-Aldrich). For immunoprecipitation, protein lysates (100 mg per sample) were first diluted in NET buffer to a protein concentration of 1 mg·mL−1, and pre-cleared with Protein A agarose (Sigma-Aldrich) beads for 1 h at 4 °C. Mouse OTUB1 mAb was added in dilution 1 : 1000 and the immunoprecipitation was then carried out overnight at 4 °C, followed by incubation with Protein A agarose for 2 h at 4 °C to couple the beads. Material was eluted using 100 mm glycine pH 2.5, precipitated using chloroform and methanol, separated by SDS/PAGE and visualized using silver staining, as described in Edelmann et al. [29]. Gel bands that were unique to lanes containing OTUB1 as well as the corresponding areas in the control lane were excised and subjected to in-gel digestion with trypsin and analysis by nano-LC-MS/MS as described in [46]. The analysis of overexpressed OTUB1 by LC-MS/MS is described in the supplementary material.

RhoA activation assay

Cells (5 × 106 per sample) overexpressing either wild-type OTUB1, the C91S mutant or an empty control vector (pEF-IRES P) were lysed in 100 mm plates with cold lysis buffer (25 mm Tris/HCl pH 7.4, 5 mm MgCL2, 1% NP-40, 1 mm dithiothreitol, 5% glycerol) containing a protease inhibitor cocktail (Pierce, Rockford, IL, USA). The protein concentration was determined by a Lowry protein concentration assay (BCA; Bio-Rad, Hemel Hempstead, UK) and equal amounts of protein lysate were used for each reaction. This was followed by immunoprecipitation of activated RhoA using Rhotekin and immobilized glutathione discs according to the manufacturer’s instructions of the EZ-Detect Rho Activation Kit (Pierce). Samples enriched with active RhoA, as well as the flow-through, were separated by SDS/PAGE and RhoA was detected via immunoblotting.

Fluorescence microscopy

Double staining of intra- and extracellular Yersinia was performed essentially as described previously [25]. HEK293T cells were seeded in a 12-well plate (2 × 105 well−1) on coverslips and grown for 12 h in DMEM supplemented with 10% fetal bovine serum, 1% glutamine and 1% penicillin/streptomycin. Cells were then transfected using SuperFect (Qiagen) and infected on the following day. Prior to invasion cells were washed with NaCl/Pi and incubated in DMEM without fetal bovine serum and antibiotics for 30 min. A Y. pseudotuberculosis overnight culture was diluted 1 : 20 and incubated for another 2 h at 27 °C, and then for 1 h at 37 °C. Bacteria were washed in NaCl/Pi and cells were infected with Yersinia at an MOI of 60 : 1 for the indicated time course at 37 °C. Thereafter cells were washed three times with NaCl/Pi and fixed at room temperature with 3% paraformaldehyde for 10 min. Fixed cells were washed with cold NaCl/Pi, blocked for 30 min with 1% BSA, washed three times with NaCl/Pi and incubated with primary Y. pseudotuberculosis antibodies for 45 min. Cells were washed and incubated with fluorescein isothiocyanate conjugated rabbit secondary antibodies for 45 min in the dark, followed by washing and permeabilization in 2% Triton X-100 for 4 min. A second antibody-based staining was performed using tetramethyl rhodamine iso-thiocyanate rabbit secondary antibodies. Cells were washed and briefly dried and mounted on coverslips in Vectashield HardSet Mounting Medium with 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA, USA), and sealed with nail polish. The images were taken and analysed using a confocal microscope (Zeiss LSM510 Meta Confocal Imaging System, Jena, Germany). For the statistical analysis to determine the ratio between intracellular and extracellular bacteria the experiment was repeated three times and the numbers of bacteria associated with at least 300 cells were counted.


We would like to thank Dr B.W. Wren (UCL, London) for providing us with the Yersinia enterocolitica and Yersinia pseudotuberculosis strains, Dr K. Trülzsch (Max von Pettenkofer Institute, Germany) for the anti-YpkA serum, Dr Roland Nordfelth (Umeå University, Sweden) for providing us with the Yersinia YpkA D270A mutant strain, Prof. James R. Bliska (University of California-Berkeley, USA) for the Yersinia pseudotuberculosis contact A mutant strain, Dr C. Garrison Fathman (Stanford University, USA) for the OTUB1 ARF-1 construct, Dr M. Olson (Glasgow, UK) for the generous gift of RhoA DNA constructs and Dr L. Navarro and Dr J.E. Dixon (UCLA, USA) for sending us YpkA wild-type and D267A expression plasmids. We also thank Dr C. Wright (University of Oxford, UK) for assistance with the isolation of primary monocytes and Dr A. Simmons and J. Baker (University of Oxford, UK) for providing buffy coats from the National Blood Centre (UK). BMK was supported by a MRC New Investigation Award and is now supported by the Biomedical Research Centre (NIHR), Oxford, UK. MA is supported by the Swedish Research Council, Lars Hiertas Minne, the Loo and Hans Ostermans Foundation for Geriatric Research and the Foundation for Geriatric Diseases at the Karolinska Institutet, Stockholm, Sweden.