Suppression of mucosal Th17 memory responses by acellular pertussis vaccines enhances nasal Bordetella pertussis carriage

Pertussis has made a spectacular rebound in countries that have switched from whole-cell (wPV) to acellular pertussis vaccines (aPV). Here, we show that, unlike wPV, aPV, while protective against lung colonization by Bordetella pertussis (Bp), did not protect BALB/c mice from nasal colonization, but instead substantially prolonged nasal carriage. aPV prevented the natural induction of nasal interleukin-17 (IL-17)-producing and interferon-γ (IFN-γ)-producing CD103+ CD44+ CD69+ CD4+-resident memory T (TRM) cells. IL-17-deficient, but not IFN-γ-deficient, mice failed to clear nasal Bp, indicating a key role of IL-17+ TRM cells in the control of nasal infection. These cells appeared essential for neutrophil recruitment, crucial for clearance of Bp tightly bound to the nasal epithelium. Transfer of IL-17+ TRM cells from Bp-infected mice to IL-17-deficient mice resulted in neutrophil recruitment and protection against nasal colonization. Thus, aPV may have augmented the Bp reservoir by inhibiting natural TRM cell induction and neutrophil recruitment, thereby contributing to the pertussis resurgence.


INTRODUCTION
Pertussis, also known as whooping cough, is a highly contagious respiratory disease, mainly caused by Bordetella pertussis (Bp). The most severe manifestations of pertussis occur in young children, but pertussis also affects adolescents, adults, and the elderly 1 . After the implementation of whole-cell pertussis vaccines (wPVs) combined with diphtheria-and tetanus-toxoid-based vaccines, referred to as DTP vaccines, in the 1950s and 1960s, the disease burden markedly lessened. However, concerns about the safety of wPV led to the replacement of DTP vaccines with diphtheria tetanus acellular pertussis vaccines (aPVs) in most industrialized countries in the 2000s. Less reactogenic than wPV, aPVs contain one to five purified Bp antigens and are highly effective in preventing pertussis disease 2 .
Despite high vaccination coverage, pertussis is now reemerging in several areas of the world 3,4 . Possible reasons that may contribute to this resurgence include improved diagnosis, the emergence of antigenic variants that escape vaccine-mediated immunity, suboptimal and fast waning immunity, especially after aPV vaccination [5][6][7] , and the inability of the current vaccines to prevent infection and transmission of Bp 8,9 . Improving our understanding of pertussis pathogenesis and how vaccines may modulate Bp interactions with its host is needed to enhance whooping cough control. Although mice do not develop the characteristic cough associated with the disease, they share many features of human whooping cough 10,11 and have been used in many studies to decipher immune mechanisms and various aspects of Bp infection 12 .
Bp clearance from the lungs of immunized mice has been shown to correlate with pertussis vaccine efficacy in children [13][14][15][16] . Similar to infection, immunization with wPV was associated with the induction of strong T helper type 1 (Th1)/Th17 immune responses, whereas aPV preferentially induces Th2 cells in mice 13,14,16,17 . Similar observations were reported in humans and in non-human primates 8,18 . These Th subsets are mainly identified by signature cytokines such as interferon-γ (IFN-γ) (Th1), interleukin-17 (IL-17) (Th17), and IL-4, IL-5, and IL-13 (Th2). Complementary roles for cellular and humoral immunity in protection have been identified in the mouse model, showing that both antibodies and T cell responses are important for protection. In prototypical Th1 C57BL/ 6 mice, lung infection by Bp is prevented by the accumulation of tissue-resident memory T (T RM ) cells in the lungs 19,20 , and induction of T RM cells in the nose negatively correlates with nasal bacterial burden 20 . While vaccination with aPV strongly reduces the Bp load in the lungs, it does not protect against nasal carriage of Bp in the nasal cavity of mice 21 , consistent with its inability to do so in a nonhuman primate model 8 .
In this study, we took advantage of Th2-prone BALB/c mice to model the impact of aPV immunization on immune responses to Bp infection. This model was chosen as it reflects the situation in newborns, who are biased toward a Th2-type differentiation 22 . We report here that aPV immunization, while providing strong protection in the lungs, not only fails to protect against nasal colonization by Bp but also prevents natural clearing of Bp infection in the noses of BALB/c mice by suppressing Th17 T RM cell responses and ensuing neutrophil recruitment.

RESULTS
aPV immunization prolongs nasal carriage of Bp in Th2-prone mice To evaluate the relative protection conferred by wPV and aPV in the upper and lower respiratory tracts of mice, 6-week-old C57BL/ 6 and BALB/c mice were immunized twice subcutaneously with 1/ 10 human dose of Infanrix (aPV) or Shan5 (wPV) vaccine at a 4week interval or left unvaccinated as controls (Fig. 1a) and challenged intranasally 4 weeks after the second immunization with B1917GR, a gentamycin-resistant derivative of a clinical Bp strain representative of currently circulating strains 23 . The bacterial burden in the noses and lungs of the mice was determined by counting colony-forming units (CFUs) on the homogenized organs at different times post challenge (p.c.) (3 hours p.c. (h.p.c.) (D0) and 7, 14, 28, 56 days p.c. (d.p.c.)) ( Fig. 1a).
Both aPV and wPV protected equally well against lung colonization after infection with a high-dose challenge (10 6 CFU) (Fig. 1b). Protection was also evident when the mice were infected with a lower dose (10 5 CFU) ( Supplementary Fig. 1a). However, neither vaccine protected against nasal infection of C57BL/6 mice (Fig. 1c, left panel), even when the mice were challenged with a moderate dose of Bp ( Supplementary Fig. 1b, left panel). In BALB/c mice, immunization with wPV led to a slight, but significant reduction of nasal colonization at the early time points after infection, whereas aPV provided no protection at all (Fig. 1c, right panel and Supplementary Fig. 1b, right panel). Instead, vaccination with aPV resulted in significantly prolonged nasal carriage. The amounts of CFU in the nasal tissues were as high at 56 d.p.c. as they were in the beginning of the infection, with no trend of a decrease in bacterial load (Fig. 1c, right panel and Supplementary Fig. 1b, right panel).
Both BALB/c and C57BL/6 mice develop resident memory Th17 cells in the nose after Bp infection IL-17-producing CD4 + T RM cells, expressing CD69, CD44, and/or CD103, were recently shown to correlate with protection against nasal colonization by Bp in C57BL/6 mice 20,24 . We, therefore, investigated whether the Th2-prone BALB/c mice might be less efficient than C57BL/6 mice in inducing these cells in the nasal tissues upon Bp infection. To identify T RM cells in the nasal tissues, mice were injected intravenously with an anti-CD45-PE antibody (CD45iv), enabling us to distinguish infiltrated resident cells from circulating immune cells. Following the intranasal challenge, up tõ 30% of immune cells in the noses of BALB/c mice were lymphocytes expressing CD3 (Fig. 2a). Among the CD3 + T cells, the proportion of CD69 + CD4 + T cells doubled 14 d.p.c. (Fig. 2a). These CD69 + CD4 + T cells up-regulated CD44 and CD103 expression over time after infection (Fig. 2a), indicating that BALB/c mice were able to induce T RM cells in the nasal tissue after Bp infection. However, compared to C57BL/6 mice, expression of these T RM markers was somewhat delayed in BALB/c mice (Fig.  2b). Nonetheless, at 14 d.p.c. C57BL/6 and BALB/c mice shared a similar expression pattern of both markers (Fig. 2b).
Measurement of IL-17-and/or IFN-γ-producing cells after stimulation with phorbol myristate acetate (PMA) and ionomycin in the presence of brefeldin A showed that similar to T RM cells from C57BL/6 mice, T RM cells from BALB/c mice were activated quickly after challenge and expressed mainly IL-17 (Fig. 2c). However, only 14% of T RM cells from BALB/c mice versus 32% of T RM cells from C57BL/6 mice expressed IFN-γ or both IL-17 and IFN-γ 14 d.p.c. (Fig.  2c). Thus, BALB/c mice are not intrinsically unable to induce nasal IL-17-producing CD4 + T RM cells upon infection with Bp.
aPV immunization impairs the induction of nasal Th1 and Th17 T RM cells in BALB/c mice upon Bp infection As Bp infection leads to the induction of Th1 and Th17 T RM cells in the nasal tissues of BALB/c mice, we examined the effect of aPV or wPV immunization on Bp infection-induced T RM cells in the noses of these mice. Whereas upon Bp infection an increase in both CD103 + CD44 + and CD103 − CD44 + T RM cells was detectable among CD69 + CD4 + T cells in the nasal tissues of BALB/c mice from 7 d.p.c. on and peaking at 14 d.p.c., prior aPV vaccination completely abolished the induction of these cells (Fig. 3a, upper panels). Representative examples of CD103 + CD44 + T cell enrichment among CD69 + CD4 + -resident T cells between 7 and 28 d.p.c. are shown in Fig. 3a (lower panels). In contrast, immunization with wPV did not inhibit the induction of CD4 + T RM cells upon Bp infection, but instead resulted in earlier induction of the CD103 + CD44 + T RM cells that were significantly enriched 7 d.p.c. compared to non-vaccinated mice (Fig. 3a).
In addition, the CD103 + CD4 + T RM cells from aPV-immunized mice were poorly activated, as <20% of CD103 + T RM cells from aPV-immunized mice expressed IL-17 or IFN-γ upon stimulation with PMA and ionomycin 7 or 28 d.p.c., whereas most of (>50%) the stimulated CD103 + CD4 + T RM cells from control and wPVimmunized mice produced IL-17 and/or IFN-γ 28 d.p.c. (Fig. 3b). IL-17 + IFN-γ + cells were also enriched in the CD103 + CD4 + T RM cell subset of control and wPV-immunized mice. No IL-17 + IFN-γ + cells were detected in aPV-immunized mice. Absolute counts of IL-17 + T RM and IFN-γ + T RM cells show that IL-17 + T RM cells predominated in wPV-immunized mice (Fig. 3c, d). IL-17 + CD103 + T RM cells were significantly enriched in wPV-immunized mice compared to control mice 7 d.p.c., while at later time points after infection, the numbers of IL-17 + CD103 + T RM cells tended to be higher in control mice than in wPV-immunized mice (Fig. 3c). A similar observation was made for the IFN-γ-producing CD103 + CD4 + T RM cells (Fig. 3d). Thus, early protection conferred by wPV against nasal colonization of BALB/c mice was associated with a rapid expansion of IL-17 + CD103 + CD4 + T RM and to a lesser extend IFNγ + CD103 + CD4 + T RM cells. In contrast to wPV, aPV immunization totally inhibited the induction of these cells in the nasal tissues of Bp-infected BALB/c mice. A similar trend was also observed for IL-17 + CD103 − CD4 + T RM and for IFN-γ + CD103 − CD4 + T RM cells.
To examine whether the difference between the aPV and wPV effect on the inhibition of T RM cell induction in the nose of BALB/c mice might be related to differences in serum antibody responses to the two vaccines, we measured serum IgG titers to whole B. pertussis cell extracts in aPV-and in wPV-immunized mice before and after B. pertussis challenge. Both before and after the challenge, wPV and aPV induced high levels of anti-B. pertussis serum antibodies ( Supplementary Fig. 2). These titers were slightly, but significantly higher in wPV-immunized than in aPV-immunized mice, making it unlikely that the difference in nasal T RM cell induction was due to a blunting effect by aPV-induced serum antibodies.
IL-17 is essential for the control of nasal but not of lung infection by Bp As aPV immunization appeared to prevent natural clearance of nasal Bp colonization and to inhibit the induction of IL-17-and/or IFN-γ-producing CD4 + T RM in the noses of Bp-infected BALB/c mice, we investigated the relative role of IL-17 and IFN-γ in controlling nasal carriage of Bp. IL-17 knockout (KO) mice and IFNγ KO mice were infected nasally with Bp, and lung and nasal colonization was followed over time. Whereas both KO mice cleared the lung infection as quickly as the control mice, regardless of the initial bacterial burden (  IFN-γ deficiency only slightly affected Bp clearance compared to control mice. We then assessed the effect of aPV and wPV immunization on lung and nasal clearance of Bp in IL-17 KO mice compared to control mice. With a moderate challenge dose, either vaccine provided protection against lung colonization in IL-17 KO mice similar to the control mice ( Supplementary Fig. 3b, c, left panels, respectively), while protection conferred by aPV and wPV was slightly impaired in IL-17 KO mice when mice were infected with a high inoculum (Fig. 4b, left panel and Fig. 4c, left panel, respectively). As expected, neither vaccine was able to provide protection against nasal carriage of the IL-17 KO mice, regardless of the challenge dose (Fig. 4b, c, right panels and Supplementary  Fig. 3b, c, right panels).
T RM cell-mediated protection against nasal colonization depends on IL-17 As IL-17 appeared to be essential for nasal clearance of Bp infection, we assessed whether transfer of IL-17 + CD103 + CD4 + T RM cells could rescue the ability of IL-17 KO mice to clear nasal infection by Bp. A total of 10 4 CD4 + T RM cells isolated from the noses of C57BL/6 mice 14 d.p.c. were transferred intravenously (i.v.) to IL-17 KO mice (KOt), 10 h.p.c. and 7 d.p.c. (Fig. 5a). Control mice were injected intravenously with phosphate-buffered saline (PBS). CFUs in the noses were counted 3 h.p.c. and 7, 14, and 28 d.p.c. The first transfer of CD4 + T RM cells did not affect nasal colonization compared to non-transferred mice, as at 7 d.p.c. the bacterial load in the noses of KOt and KO mice were similar (Fig.  5b). After the second transfer given 7 d.p.c., the bacterial load in the noses of KOt mice started to decrease at a rate similar to that of C57BL/6 control mice, indicating that the transfer of Bp-induced CD4 + T RM cells allowed IL-17 KO mice to reduce nasal Bp infection as efficiently as infected C57BL/6 mice (Fig. 5b).
Lack of CD4 + T RM cell induction results in lack of neutrophil recruitment In an attempt to identify the effector cells involved in nasal clearance, we examined the nasal influx of neutrophils upon Bp infection. CD45 + Ly-6G + neutrophils infiltrated in the noses of Bpinfected C57BL/6 mice 14 d.p.c. were compared to non-infected mice (Fig. 5c, d). No significant increase in neutrophils was observed in infected IL-17 KO mice, whereas transfer of nasal CD4 + T RM cells from Bp-infected C57BL/6 mice resulted in significant neutrophil recruitment in the noses of IL-17 KO mice   14 d.p.c. (Fig. 5c, d). The decrease in nasal bacterial loads in C57BL/6 and KOt mice over non-transferred IL-17 KO mice was thus associated with an increase in the amounts of recruited neutrophils (Fig. 5d). Overall, the data suggest that upon Bp infection nasal CD4 + T RM cells induce the recruitment of neutrophils by releasing IL-17 and thereby enhance bacterial clearance.
To determine whether the inability of aPV-immunized BALB/c mice to clear the nasal infection by Bp was also due to their failure in recruiting neutrophils in the noses, we immunized BALB/c mice with aPV or wPV and examined the recruitment of nasal neutrophils over time. While Bp infection of non-vaccinated BALB/c mice resulted in a significant increase in the nasal influx of neutrophils, this influx was totally abolished in aPV-immunized, but not in wPV-immunized mice (Fig. 5e, f). The recruitment of other nasal CD3 − CD11b + immune cells was not impaired in aPVimmunized mice (Supplementary Fig. 4).
aPV immunization allows extracellular persistence of Bp in the noses of BALB/c mice As in vitro data suggest that Bp can invade and survive within eukaryotic cells 25 and can also adopt a biofilm lifestyle allowing persistent colonization in mice 26 , we assessed the impact of aPV immunization on intracellular versus extracellular nasal Bp carriage. aPV-or wPV-immunized BALB/c mice were nasally infected with Bp B1917GR, and nasal washes (NWs) and homogenates were assessed for CFU counts at different time points after challenge (Fig. 6a). The nasal homogenates were treated with collagenase D and DNase I for tissue and biofilm disruption 26,27 . After treatment, the cells were harvested by centrifugation and extracellular bacteria were counted by CFU measurements of the supernatants. To measure the amounts of intracellular bacteria, the cells were washed and incubated with polymyxin B in order to kill potentially remaining extracellular bacteria. Cells were then lysed and intracellular bacteria were quantified by CFU measurements.
Approximately 10-to 100-fold more bacteria were found in the nasal homogenates compared to the NWs at each time point, suggesting that most bacteria were tightly associated with the nasal epithelium (Fig. 6b). In contrast to control mice and wPVimmunized mice, the bacteria remained associated with the nasal epithelium in aPV-immunized mice for at least up to 56 d.p.c., whereas both vaccines reduced the bacterial load in the NWs (Fig.  6b). Virtually no intracellular bacteria were detected 3 h.p.c. (day 0), whereas substantial amounts of intracellular bacteria were detectable at 7 d.p.c., especially in control mice and in aPVimmunized mice (Fig. 6c). Intracellular bacteria were eventually cleared in all three groups 28 d.p.c. (Fig. 6c).

DISCUSSION
Several reasons have been proposed to explain the resurgence of pertussis, especially in countries using aPV, although many studies have shown a good efficacy of aPV against pertussis disease. These reasons include the rapid waning of protective immunity conferred by aPV 6,28 compared to wPV 29,30 and the inability of aPV, as well as of wPV to prevent Bp infection and pertussis transmission. Using mouse models, we show here that in addition to the failure of immunization with aPV or wPV to protect against nasal colonization aPV immunization resulted in substantially prolonged nasal Bp carriage in Th2-prone BALB/c mice when compared to non-immunized mice, while it was protective against lung colonization. These findings are consistent with observations in a non-human primate model showing that aPV immunization, while protecting against disease, caused prolonged carriage of Bpchallenged baboons 8 . We found that aPV-induced prolonged nasal carriage in BALB/c mice was paralleled with vaccinemediated inhibition of nasal IL-17-and/or IFN-γ-producing CD103 + CD4 + T RM cells induced by Bp infection. Recently, Wilk et al. 20 showed that, in contrast to wPV immunization, aPV immunization fails to expand T RM cells in the lungs and noses and provided evidence that T RM cells promote bacterial clearance in mouse lungs by adoptive transfer of CD4 + T RM cells in irradiated mice or by treating mice with FTY720 during the immunization phase, which prevents migration of T and B cells from lymph nodes to the circulation 19,20 . We found here that aPV not only fails to induce IL-17 + and/or IFN-γ + CD4 + T RM cells, but even prevents the induction of these cells in the nasal tissue of Bp-infected BALB/ c mice, while they were strongly induced in non-vaccinated mice upon Bp challenge, peaking at 14 d.p.c., and were maintained for  ) and to evaluate the induction of neutrophils in infected C57BL/6 (*, d). Kruskal-Wallis tests were performed to compare aPV ( # ) and wPV (*) immunized BALB/c mice to control mice (f). * ,# P < 0.05, ** ,## p < 0.001, ***p < 0.001, and ****p < 0.0001. Only significant differences are indicated.
at least up to 28 d.p.c. In contrast, in wPV-immunized mice, these cells were induced earlier than in non-vaccinated mice, and peaked at 7 d.p.c., which was paralleled by a significant decrease in bacterial burden in the noses of wPV-immunized BALB/c mice 7 d.p.c. compared to non-vaccinated controls.
While Bp infection resulted in the expansion of both IFN-γproducing Th1 and IL-17-producing Th17 CD4 + T RM cells in the nasal tissue, the latter were the main cells involved in the control of nasal colonization by Bp, as IL-17 KO mice were unable to clear nasal Bp infection, while able to control lung infection as well as control mice. In contrast, IFN-γ KO mice cleared the infection both in lungs and noses, similar to control mice. Furthermore, adoptive cell transfer of nasal CD4 + T RM cells isolated from Bp-infected C57BL/6 mice to IL-17-deficient mice conferred protection against nasal colonization. Thus, aPV immunization prevents nasal Bp clearance in Th2-prone mice by preventing the natural induction of nasal IL-17 + CD103 + CD4 + T RM cells upon Bp infection.
IL-17 deficiency was associated with failure to recruit neutrophils upon Bp infection, and adoptive transfer of Th17 T RM cells to IL-17 KO mice was associated with accelerated neutrophil recruitment, suggesting that Th17 T RM cells control the outcome of the infection by enhancing neutrophil recruitment in the noses of Bp-infected mice. Through releasing of IL-17, Th17 T cells are known to attract neutrophils to the site of infection via CXC chemokines and to modulate neutrophil granulopoiesis via the induction of endogenous granulocyte colony-stimulating factor and stem cell factor 31,32 . Consistent with a protective role of neutrophils in immunized or convalescent mice, neutropenia severely impairs clearance 33 . Furthermore, Bp clearance from mouse lungs after the transfer of antibodies from convalescent mice was abrogated by neutrophil depletion or Fcγ receptor deletion 34 , suggesting that neutrophils mediate the killing of Bp in the lungs by opsonic phagocytosis. We recently showed that passive transfer of serum from aPV-immunized mice or mice immunized with the live attenuated BPZE1 vaccine conferred protection in the lungs of severe combined immunodeficiency mice but not in the nose, suggesting that serum immunoglobulins (Igs) do not efficiently protect in the nose 21 . Instead, by using knockout models and passive transfer of mucosal Igs, it was demonstrated that mucosal IgA induced by BPZE1 contributes to nasal protection as well as IL-17, possibly through the induction of Th17 T RM cells.
We found that Bp was present in three locations in the noses of BALB/c mice. A minority was loosely attached to the nasal epithelium and could easily be recovered in NWs. A small fraction was found in intracellular compartments and the vast majority of Bp organisms was tightly attached to the nasal epithelium. Both wPV and aPV immunization diminished the loosely attached bacteria, suggesting that they may have been cleared by transuded wPV-or aPV-induced serum antibodies. Immunization with aPV slightly prolonged the presence of intracellular bacteria in the noses of BALB/c mice. The strongest effect of aPV immunization was seen on the Bp organisms that were tightly attached to the nasal epithelium, suggesting a major role of IL-17 + T RM cells in clearing tightly attached bacteria. IL-17 plays a central role in maintaining the mucosal barrier integrity 35,36 , and Bp has been shown to alter the epithelial cell barrier 37 , which results in the exposure of cryptic receptors on the basement membrane allowing the pathogen to attach more efficiently 38 .
The localization of CD103 + CD4 + T RM cells in the nasal epithelium positions them well to act as the first line of defense at subsequent exposure to Bp. While CD103 + T RM cells interact with epithelial cells through E-cadherin 39,40 , CD44 + T cells may be retained in the underlying extracellular matrix 41 . The greatest differences in T RM cell numbers and activation between aPVimmunized BALB/c mice and control mice were observed in the CD103 + subset of T RM cells. In addition, early accumulation of CD103 + T RM cells by wP-primed mice compared to control mice was associated with a decrease in nasal bacterial burden at early In conclusion, we provide evidence here that aPV immunization may prolong nasal carriage of Bp and thereby increase the Bp reservoir in aPV-immunized populations, which may have contributed to the increase in pertussis incidence in countries in which aPV are in use. The effect of aPV on the prolonged nasal carriage was seen in Th2-prone BALB/c mice, which were unable to clear nasal Bp carriage, especially Bp organisms that tightly adhere to the nasal epithelium, for at least up to 56 d.p.c. These findings are relevant in the context of newborns in which a bias toward a Th2-type differentiation has been demonstrated 22 , but may also be relevant for older Th2-prone individuals. The aPV effect on nasal carriage is due to aPV-mediated inhibition of IL-17 + CD103 + CD4 + T RM cells expansion, which in turn prevents neutrophil recruitment important for bacterial clearance. This study provides thus a coherent explanation of how aPV enhances nasal carriage of Bp and supports the hypothesis of asymptomatic Bp carriage and transmission as a major driver of pertussis resurgence in countries that have switched from wPV to aPV immunization programs 9 . However, aPV induces antibodies that provide protection against lung colonization in mice and against pertussis disease in humans, which is the purpose of these vaccines.

Bacterial strain
Bp strain B1917, a pertactin-producing strain, came from the RIVM collection (Bilthoven, The Netherlands). For counter selection purposes, B1917 was electroporated with the pFUS2-BctA1 suicide plasmid to acquire gentamicin resistance as described 43 , thereby yielding B1917GR. After electroporation, gentamicin-resistant derivatives were checked by PCR to verify the site of insertion of the pFUS2-BctA1 suicide vector in the bacterial genome. Bacteria were cultured at 37°C on Bordet-Gengou (BG) agar (Difco Bordet-Gengou Agar Base), supplemented with 1% glycerol, 10% defibrinated sheep blood, and 10 μg/ml gentamycin, as described 44 . After 40 h of growth, the bacteria were harvested by scraping the plates and resuspended in PBS at the density of 5 × 10 7 or 5 × 10 6 CFU/ml. Whole B. pertussis cell lysates were prepared and used for antibody determination as described 45

Protection experiments
Mice were immunized subcutaneously with 1/10 human dose of aPV (Infanrix, GlaxoSmithKline) or wPV (Shan5, Shantha Biotechnics, India) and boosted 4 weeks later with the same dose. Control mice were injected with PBS. Eight weeks after the first immunization, mice were challenged intranasally with 10 6 or 10 5 CFU of virulent B1917GR. Mice were euthanized at different times post challenge, that is 3 h.p.c. and 7, 14, 28, and 56 d.p.c. NWs were performed with stainless steel feeding tubes (24 ga × 25 mm, Phymep). One milliliter of PBS was slowly injected through the trachea and the NW was collected. Lungs and noses were harvested and homogenized for CFU measurement by plating 10-fold serial dilutions onto BG agar plates containing 10 μg/ml gentamicin as described 21 .
Cell isolation for CFU tests Infected BALB/c mice were euthanized at different times post challenge (3 h.p.c. and 7, 28, and 56 d.p.c.). NWs were collected prior to nose harvest to assess extracellular CFU in the nasal fluid. Nasal tissue was scraped from the nasal cavity, cut into small pieces, and digested with collagenase D (0.6 mg/ml; Roche) and DNase I (20 U/ml; Sigma-Aldrich) for 40 min at 37°C. The nasal tissue was flushed several times and then passed through a cell strainer to obtain a single-cell suspension. Cells were pelleted and the supernatant was conserved to assess extracellular CFU. Cells were washed twice and then incubated for 1 h with 100 µg/ml polymyxin B sulfate (Sigma) to eliminate extracellular bacteria 46 . Cells were washed twice and then lysed with 0.1% saponin (Sigma). CFU counting was performed on cell lysates by plating 10-fold serial dilutions onto BG agar plates containing 10 μg/ml gentamicin to estimate the number of intracellular bacteria.
Adoptive transfer of CD4 + T RM cells from the nose CD4 + T RM cells were purified from noses of C57BL/6 mice infected with 10 6 CFU of B1917GR 14 d.p.c. Cells were treated as described above. CD45iv − CD45 + CD44 + CD69 + CD4 + T cells were sorted using BD FACS Aria II sorter (BD Biosciences). Purity of the isolated cells was checked prior to transfer (Supplementary Fig. 6). A total of 10 4 cells were transferred intravenously to IL-17 KO mice 10 h.p.c. and 7 d.p.c.

Statistical analyses
Statistical analyses were performed by non-parametric T tests, Kruskal-Wallis, or one-way analysis of variance tests and Mann-Whitney tests using the GraphPad Prism software. P values < 0.05 were considered significant.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
V. Dubois et al.