|Home | About | Journals | Submit | Contact Us | Français|
In this study, we developed a large-scale screening of bacterial strains in order to identify novel candidate probiotics with immunomodulatory properties. For this, 158 strains, including a majority of lactic acid bacteria (LAB), were screened by two different cellular models: tumor necrosis factor alpha (TNF-α)-activated HT-29 cells and peripheral blood mononuclear cells (PBMCs). Different strains responsive to both models (pro- and anti-inflammatory strains) were selected, and their protective effects were tested in vivo in a murine model of influenza virus infection. Daily intragastric administrations during 10 days before and 10 days after viral challenge (100 PFU of influenza virus H1N1 strain A Puerto Rico/8/1934 [A/PR8/34]/mouse) of Lactobacillus plantarum CNRZ1997, one potentially proinflammatory probiotic strain, led to a significant improvement in mouse health by reducing weight loss, alleviating clinical symptoms, and inhibiting significantly virus proliferation in lungs. In conclusion, in this study, we have combined two cellular models to allow the screening of a large number of LAB for their immunomodulatory properties. Moreover, we identified a novel candidate probiotic strain, L. plantarum CNRZ1997, active against influenza virus infection in mice.
Probiotics are defined by the Food and Agriculture Organization (FAO) of the United Nations World Health Organization (WHO) as “live microorganisms which, when administered in adequate amounts, confer health benefits on the host” (1). Their use is more and more popular for both prevention and treatment of a number of human diseases (2–5). Most probiotic microorganisms belong to lactic acid bacteria (LAB). However, not all probiotics display the same properties, and careful selection of specific strains based on their claimed beneficial effects is needed. Although several criteria have been proposed to identify novel probiotic strains, some studies have reported the selection of potential candidates at the preclinical level (i.e., animal trials), including evaluation of both safety and potential efficacy (6–8). The efficacy of probiotic bacteria is greatly influenced by their functional properties, such as antimicrobial activity, persistence in the gastrointestinal tract (GIT) for intestine-targeted probiotics, and immunomodulatory properties (9, 10).
Despite advances in medicine, common respiratory virus infections (RVI) such as the common cold or flu continue to cause a considerable economic burden; fortunately, some probiotic strains have been studied for their positive effects on these virus infections (11, 12). Indeed, products containing probiotics have been shown to have an immunomodulatory effect and a protective effect against RVI in both mice and humans (13–15). For example, oral daily administration of the LAB Lactobacillus plantarum L-137 (a strain selected for its proinflammatory properties in vitro) before and after influenza virus H1N1 challenge in mice enhanced survival and decreased virus titers in lungs of infected mice (15). Furthermore, several other LAB strains, such as Lactobacillus fermentum CECT5716 or Lactobacillus casei DN114-001, were described recently to enhance the effects of influenza virus vaccination and to improve antibody responses to influenza virus vaccination in humans, respectively (16, 17). Also, a mixture of Lactobacillus gasseri PA 16/8, Bifidobacterium longum SP 07/3, and Bifidobacterium bifidum MF 20/5 reduced the severity of symptoms related to common cold episodes in humans (14). In another clinical trial, the most commonly used probiotic strain, Lactobacillus rhamnosus GG, tested alone or in association with B. animalis subsp. lactis BB-12, reduced the incidence of RVI (18, 19). A clinical trial using Lactobacillus acidophilus strain NCFM alone or in association with Bifidobacterium animalis subsp. lactis BI-07 showed that these probiotics reduce influenza-like symptoms (fever, rhinorrhea, cough incidence, and duration of antibiotic prescription) (20). In addition, in another interesting study, Boge et al. (16) demonstrated that daily consumption of a probiotic fermented dairy drink improves antibody responses to influenza virus vaccination in the elderly in two randomized controlled trials. Altogether, these preclinical and clinical trials suggest that probiotics can be successfully used as preventive and therapeutic agents in RVI.
Over the last 5 years, the interest in the immunomodulatory properties of LAB strains has significantly increased. In this context, a wide variety of strain-dependent properties have been reported (8, 10, 13), and several in vitro cellular models have been developed in order to analyze and classify the immunomodulatory properties of these strains. The aim of this study is thus to determine the immunomodulatory properties of a large set of LAB (158 strains) using two different cellular models, tumor necrosis factor alpha (TNF-α)-activated HT-29 cells and peripheral blood mononuclear cells (PBMCs), in order to identify the most efficient strain and develop a new probiotic supplement able to alleviate symptoms of RVI and, more particularly, the common cold. After establishing the immunomodulatory profile (based on their cytokine production profile) of each strain of the collection, different strains responsive to both models (pro- and anti-inflammatory strains) were selected and tested in vivo in a murine model of RVI, an influenza virus infection, to determine their protective effects.
The collection of this study comprised 158 strains isolated from different biotopes and including a majority of LAB (90 Lactobacillus spp., 31 Lactococcus spp., 31 Bifidobacterium spp., 3 Streptococcus spp., 2 Pediococcus spp., and 1 Bacillus sp.; INRA Collection CIRM, Rennes, France) (Table 1). Lactococcus spp. and Streptococcus spp. were grown in M17 medium (Difco, KS) supplemented with 5% lactose at 30°C without shaking, Lactobacillus and Pediococcus spp. were grown in MRS medium (Difco, KS) at 37°C without shaking, Bifidobacterium spp. were grown in MRS medium supplemented with 10% cysteine (Sigma-Aldrich, France) at 37°C under anaerobic conditions (GENbox Microaer; bioMérieux, Marcy l'Etoile, France) without shaking, and Bacillus spp. were grown in Luria-Bertani medium (Difco, KS) at 37°C with shaking.
Resistance to bile salts was studied to mimic the passage of the strains in the GIT. A high-throughput method based on sterile microwell plates (384 wells; Greiner, Bio-one) was used. From a culture of each strain grown overnight, the optical density at 600 nm (OD600) was adjusted to 1 in peptone water. Five hundred microliters from this adjusted culture was added either to 500 μl of a cholic acid (Sigma-Aldrich, France) and deoxycholic acid (Sigma-Aldrich, France) solution in peptone water (0.05% [vol/wt] for each acid) or to 500 μl of peptone water. After 1 h at 37°C, stressed and nonstressed bacterial suspensions were centrifuged. Pellets were resuspended in 1 ml of appropriate culture medium and diluted 10 times. Sixty microliters from those suspensions was deposited into a well already containing 60 μl of medium. Four serial 2-fold dilutions were then performed. Growth of the stressed and nonstressed strains was monitored every 15 min for 19 h by measuring the OD600 using a thermoregulated plate reader (InfiniteM200 Pro; Tecan, France). The resistance of each strain to bile salts stress was determined by measuring the growth delay (i.e., delay for the time to reach mid-exponential phase) between stressed and nonstressed cultures. For each strain, 5 growth delays (in each experiment) corresponding to the five dilutions on the microplate were averaged. Experiments were performed in triplicates.
The human colon carcinoma cell line HT-29 was cultured in 24-well culture plates in Dulbecco's modified Eagle's medium (DMEM) (Lonza, Switzerland) supplemented with 10% heat-inactivated fetal calf serum (FCS)–1% glutamine at 37°C in a 10% CO2–air atmosphere. Medium was changed every day. Experiments were initiated on day 7 after seeding, when cells were at confluence (~1.83 × 106 cells/well). Twenty-four hours before bacterial coculture (day 6), the culture medium was changed for a medium with 5% heat-inactivated FCS and 1% glutamine. On the day of coculture, bacteria were added at a multiplicity of infection (MOI) of 1:40 in 50 μl DMEM in a total volume of 500 μl. Cells were stimulated simultaneously with recombinant human TNF-α (5 ng/ml; Peprotech, NJ) for 6 h at 37°C in 10% CO2. All samples were analyzed in triplicate. After coincubation, cell supernatants were collected and frozen at −80°C until further analysis of interleukin-8 (IL-8) concentrations by an enzyme-linked immunosorbent assay (ELISA) (Biolegend, San Diego, CA).
Commercial PBMCs (StemCell Technologies SARL, Grenoble, France) from healthy donors were used. For the first screening (i.e., 158 strains), we used PBMCs from one healthy donor (American man, Caucasian, aged <65 years, with a body mass index of <30), nonsmoking, with no drugs with anti-inflammatory effects taken during 15 days prior to sampling and negative for HIV and hepatitis A and B viruses. For the confirmation phase, we used PBMCs from 3 different healthy donors presenting similar health characteristics. After reception, cells were stored in liquid nitrogen until use. To prepare PBMCs for coculture experiments with bacteria, the vial cells were thawed at 37°C in a water bath and then transferred into a medium containing RPMI 1640 medium (Lonza, Switzerland) supplemented with 10% heat-inactivated FCS, 1% l-glutamine, and 0.1% penicillin-streptavidin. DNase (10 mg/ml) was added to this mix to avoid clumping. Cells were then centrifuged at 200 × g for 15 min, counted by using trypan blue, and spread onto 24-well plates at 1 × 106 cells/well. Bacteria were added in triplicate at an MOI of 1:10 in 20 μl of phosphate-buffered saline (PBS) in a total volume of 1 ml, and plates were incubated for 24 h at 37°C in 10% CO2. Samples were finally stored at −80°C until further analysis of IL-10 and IL-12p70 concentrations by ELISA (Mabtech, Sweden).
Specific-pathogen-free BALB/c mice (females, 6 weeks of age; Janvier, France) were maintained under normal husbandry conditions in the animal facilities of the National Institute of Agricultural Research (UEAR, INRA, Jouy-en-Josas, France). All animal experiments were started after the animals were allowed 2 weeks of acclimation and were performed according to European Community rules of animal care and with authorization 78-149 of the French Veterinary Services.
Influenza virus H1N1 strain A Puerto Rico/8/1934 (A/PR8/34; a mouse-adapted strain) was grown in allantoic cavities of 11-day-old fertile chicken eggs for 2 days at 35°C (21). The viral titer (i.e., PFU determination) was quantified by a standard plaque assay using Madin-Darby canine kidney (MDCK) cells, and the virus stock was stored at −80°C until use. For intranasal infection, mice were fully anesthetized by intraperitoneal injection of ketamine (Imalgene 1000, Merial, France) and xylazine (Rompun, Alcyon, France) (0.1% ketamine plus 0.06% xylazine; 150 μl for a mouse weighing 20 g) and then infected by intranasal application of 50 μl of virus suspension (25 μl into each nostril).
Candidate probiotic bacteria were grown as described above. Pellets from cultures grown overnight were then harvested by centrifugation at 3,000 × g at 4°C and washed with sterile PBS. The pellet was suspended in PBS to a final concentration of 5 × 109 CFU/ml. Plate counts were performed with all inocula to corroborate the CFU administered. As a positive control for our in vivo experiments, we used L. casei DN114-001 and L. rhamnosus GG, two probiotic strains having well-documented in vitro and in vivo immunomodulatory properties and protective effects in different models of influenza virus infection, as positive controls (13, 16, 22, 23).
Groups of mice (n = 8) were daily administered intragastrically 1 × 109 CFU of each strain suspended in 200 μl of PBS (with a feeding needle) 10 days before and 10 or 14 days after virus challenge. PBS was used as a negative control. Mice were monitored daily for mortality, weight loss, and visual score by scientists blinded to the study. For visual score, we used the following five-point scale, as reported previously (24): 5, healthy (no clinical symptoms); 4, mild (fur slightly ruffled); 3, moderate (fur moderately ruffled and lethargy); 2, severe (fur severely ruffled and thin); 1, very severe (no reaction to stimulation).
On days 10 and 14 postinfection, blood samples were obtained from the retro-orbital venous plexus and centrifuged, and sera were stored at −80°C until further analysis. Mice were then sacrificed by cervical dislocation, and bronchoalveolar lavage fluid (BALF) was collected. To recover BALF, a catheter was tied to the exposed trachea, and a hypodermic needle and syringe were attached and used to inject and withdraw the lungs with a total volume of 1 ml of PBS. BALF samples were stored at −80°C until further analysis. Furthermore, lungs were collected for virus titration and stored at −80°C until analysis.
Total RNA was isolated by using a Qiagen RNeasy minikit (Qiagen, France). Reverse transcription was carried out with Superscript II reverse transcriptase (Invitrogen, France) and the specific influenza A virus (IAV) M1 primer (21), 5′-TCT AAC CGA GGT CGA AAC GTA-3′, according to the supplier's recommendations. Virus titers were quantified by quantitative PCR. We used the specific IAV M1 primer 5′-TCT AAC CGA GGT CGA AAC GTA-3′ and the Mastercycler Realplex system (Eppendorf, France). The PCR conditions and cycles were as follows: initial DNA denaturation for 10 min at 95°C followed by 40 cycles (15 s at 95°C, 20 s at 64°C, and 30 s at 72°C), followed by 15 s at 95°C, 15 s at 60°C, 20 min of melting curve (to assess the purity of the PCR product), and 15 s at 95°C. To normalize gene expression, β-actin levels were determined for all samples (sense primer 5′-AGA AAA TCT GGC ACC ACA CC-3′ and antisense primer 5′-CTC CTTAAT GTC ACG CAC GA-3′) (21).
For animal experiments, the formula used to calculate sample size was n = 1 + 2 × C × (s/d)2 (where s is the standard deviation, d is the difference to be observed with a C value of 10.51 [for an alpha value of 0.05 and a beta value of 0.1], and s/d equals 0.5). With these values, n equals 6.255, which means a sample size of 7 animals (and we used an extra animal to take into account the fact that mice can sometimes escape virus infection). Data were analyzed by using Dunnett's test or t test to compare the differences between groups and controls using Prism software. A P value of <0.05 was considered significant.
Our bacterial strain collection was first analyzed in a stress resistance experiment. As shown in Fig. 1, 44% of the strains (70 strains) had a growth delay of less than 5 h and were qualified as resistant to bile salts. We then determined the immunomodulatory properties of the 158 strains in the two cellular models: TNF-α-stimulated HT-29 cells and PBMCs.
For the HT-29 model, we analyzed TNF-α-induced IL-8 secretion by HT-29 cells. Since IL-8 is considered a major inflammatory mediator, bacteria enhancing its secretion are considered to have proinflammatory properties, while those inhibiting its secretion are considered to have anti-inflammatory properties. As shown in Fig. 2, the 158 strains revealed a very distinct pattern of secretion of IL-8 in TNF-α-stimulated HT-29 cells and can be classified as highly to weakly proinflammatory, including some neutral strains. None of the bacteria tested alone (i.e., nonstimulated HT-29 cells) induced IL-8 secretion (data not shown).
Alongside the establishment of either the anti- or proinflammatory profile of the 158 strains using the HT-29 model, we then determined their ability to modulate IL-12p70 and IL-10 secretion by PBMCs. The IL-12p70/IL-10 ratio allows us to classify strains with a strong or a weak proinflammatory profile, with a high versus low IL-12p70/IL-10 ratio, respectively. As shown in Fig. 3, the 158 tested strains again displayed different immunomodulatory profiles after coincubation with PBMCs, confirming partially the results obtained with the HT-29 model.
In order to determine the potential probiotic effect of the most interesting strains, 3 different responsive strains (Fig. 2 and and3,3, arrows) were chosen for further in vivo experiments: one with a highly proinflammatory profile (L. plantarum CNRZ1997), one with a weakly proinflammatory profile (L. brevis VEL12208), and a third with a markedly anti-inflammatory profile (L. paracasei VEL12195). We first confirmed the immunomodulatory profile of the 3 selected strains by calculating the IL-12p70/IL-10 ratio after coculture of the candidate strains with PBMCs from 5 different donors. As expected, the results obtained with the 5 different donors confirmed the profile of each strain and clearly showed a significant difference between anti- and proinflammatory strains (Fig. 4).
The effects of the 3 selected strains in a murine model of influenza virus infection were then investigated (Fig. 5). Ten days after virus challenge (2,000 PFU/mouse), the area under the weight curve (AUWC), representative of the cumulative weight loss of mice over the time course of the assay, revealed severe injury to PBS-treated mice (Fig. 6A). Mice treated with the proinflammatory strain CNRZ1997 showed the highest AUWC value compared to nontreated mice or mice treated with either a weakly proinflammatory (VEL12208) or anti-inflammatory (VEL12195) strain. Surprisingly, no significant differences were observed between nontreated mice and mice treated with either the L. casei DN114-001 or L. rhamnosus GG positive-control strain (data not shown).
These promising results demonstrate a tendency of L. plantarum strain CNRZ1997 to decrease the body weight loss after infection with a high dose of IAV (i.e., 2,000 PFU/mouse). We then decided to test the probiotic effects of strain CNRZ1997 with a lower dose of IAV (100 PFU/mouse). This lower virus dose was used to induce moderate symptoms in mice, which can be significantly counterbalanced by our candidate probiotic bacterium.
PBS-treated and infected mice displayed a weight loss of ~30% (at day 10) compared to healthy control mice, while mice treated with proinflammatory L. plantarum strain CNRZ1997 displayed a reduction in weight loss of only ~10% of body weight (at day 10). In addition, weight loss was delayed in CNRZ1997-treated mice over the time course of the assay, and an earlier initiation of recovery was also observed than for PBS-treated mice (data not shown). These results were confirmed by AUWC analysis (Fig. 6B).
To better characterize the beneficial effects of strain CNRZ1997 in the murine model of IAV infection, we used a five-point scale of visual scores. At day 0, the 3 groups of mice showed a score of 5 (Fig. 7A). At day 10, PBS-treated and infected mice presented a score of 3 (moderate symptoms), while mice treated with strain CNRZ1997 showed a score of ~4.
Virus titration in the lungs of infected mice revealed that CNRZ1997-treated mice presented a lower virus titer than infected control mice at days 10 and 14 postinfection (Fig. 7B). However, at day 10, no significant difference was observed in virus loads between control mice and mice receiving strain CNRZ1997; this difference became significant at day 14.
Despite advances in medicine, RVI (such as the common cold or flu) continue to cause a considerable economic burden; fortunately, some probiotic strains have been studied for their positive effects on certain infectious diseases, and in the last 5 years, their use to prevent and treat RVI has significantly increased (11, 12). There is thus a clear interest in the identification and characterization of new candidate strains with well-demonstrated probiotic properties against RVI.
In this study, we determined the immunomodulatory properties of 158 strains of LAB in order to identify the most interesting candidate probiotic strain able to alleviate RVI symptoms. For this, we first performed a large-scale screening of the 158 strains for their resistance to bile salts. Indeed, bile salts may constitute a deleterious factor preventing a given strain from exerting its beneficial properties in vivo. A high-throughput method was used to rapidly characterize strain resistance. Among the tested strains, L. plantarum CNRZ1997 had a delay of 2.9 h. Eighty strains (62.5% of the collection), including VEL12208 and VEL12195, showed moderate resistance, with a growth delay of between 5 and 13 h. Eight strains were not resistant to bile salts. This simple yet robust protocol could also be used to test other properties, such as other stresses or the ability to grow under specific conditions.
The immunomodulatory effects of the bacterial collection were then assessed by using two cellular models: TNF-α-activated HT-29 cells and peripheral blood mononuclear cells (PBMCs). These in vitro cellular models are commonly used to validate the immunomodulatory properties (anti- or proinflammatory cytokine profile) of a given candidate bacterium before validation in animal models. Indeed, these tests have a predictive value and allow a reduction of the number of bacterial strains to be tested in animal models, which are expensive and time-consuming. One of these predictive in vitro models is PBMCs, in which the immunomodulation potential of probiotic strains is usually assessed by measuring IL-10 and IL-12 cytokine levels after stimulation with live bacterial strains. In this manner, Foligne et al. (9) successfully established a correlation between this PBMC model, the in vivo immunomodulation potential of probiotic strains, and the ability to prevent experimental colitis in mice. Bacteria inducing higher levels of the anti-inflammatory cytokine IL-10 and lower levels of the proinflammatory cytokine IL-12 in vitro displayed the best protection in the murine colitis model (15). The human intestinal epithelial cell line HT-29 has been also used successfully to evaluate the immunomodulatory properties of bacteria (25, 26). Once coincubation with bacterial strains was achieved, IL-8 quantification led to the selection of anti-inflammatory bacteria (reducing IL-8 production) and proinflammatory bacteria (enhancing IL-8 production) (25, 27). After we tested our bacterial collection with these two cellular models, we selected 3 different responsive strains according to their immunomodulatory profile, one with a highly proinflammatory profile (L. plantarum CNRZ1997), one with a weakly proinflammatory profile (L. brevis VEL12208), and one with a markedly anti-inflammatory profile (L. paracasei VEL12195), for in vivo experiments using a model of A/PR8/34 strain infection in mice. Based on the preliminary results observed with a high dose of IAV (2,000 PFU), we found that L. plantarum strain CNRZ1997 (highly proinflammatory profile) was the most interesting candidate among the three candidate strains tested, with beneficial effects against IAV. The results obtained with a less challenging model (lower dose of IAV of 100 PFU) as well as with the five-point scale of visual scores confirmed these observations. Indeed, this strain was effective in preventing body weight loss (Fig. 6B) and clinical condition alterations by alleviating significantly the symptoms of the infected mice (P < 0.05) (Fig. 7A). Kawase et al. (28) previously found similar results (i.e., improved clinical symptoms) when studying the oral administration of two LAB strains, L. rhamnosus GG and L. gasseri TMC0356, in a murine model of IAV infection.
In addition, when we analyzed virus titration in lungs of infected mice at 10 and 14 days postinfection, we found that strain CNRZ1997 was able to decrease virus titers in these infected mice. This result indicates a positive effect of strain CNRZ1997 in mice by perhaps accelerating IAV elimination from the lungs. These results are similar to those obtained previously by Kawashima et al. (29), where L. plantarum strain LpYU (with a proinflammatory profile observed in vitro) was tested in a model of virus infection with the H1N1 virus (A/NWS/33).
In this work, we decided to use strains GG and DN114-001 as positive controls in our model of IAV, since they have already been successfully studied for their anti-IAV properties in preclinical trials (23) and in a human clinical trial against RVI (13). Interestingly, the AUWC results observed for mice treated with either GG or DN114-001 showed that there are no protective effects of these two well-known probiotic strains in our IAV model. These unexpected results can be explained by the fact that strain GG was reported previously to be protective against IAV in mice after “intranasal” administration, in contrast with our study, where we used an “intragastric” route. In order to confirm this, we performed experiments with L. rhamnosus GG by the intranasal route in our IAV model, and we confirmed a protective effect; however, we cannot favor this route, since the aim of this study was to develop a probiotic oral dietary supplement. Second, most trials using strain DN114-001 against IAV have been done with the fermented final product (which means that it contained L. casei strain DN114-001 combined with two cultures commonly used in yogurt, Streptococcus thermophilus and Lactobacillus bulgaricus), while in our study, we used DN114-001 as a single isolated strain, so we cannot discard a synergistic effect of strain DN114-001 and the fermented final product.
The effects of probiotic bacteria against infectious diseases, such as IAV, have been demonstrated, but the mechanisms of action are not yet fully understood; however, some hypotheses have been advanced. For example, oral administration of acidic exopolysaccharide extracts of L. delbrueckii OLL1073R-1 to mice infected with IAV provided protection (30). In another study, the protective properties of L. plantarum strain LpYU against IAV were dependent on Toll-like receptor 2 (TLR2), which recognizes the peptidoglycan (PG) and the lipoteichoic acids (LTA) of Gram-positive bacteria (29). These two results seem interesting to explore in order to identify the mechanisms of action of our proinflammatory strain CNRZ1997.
Most of the studies using LAB to prevent and treat IAV infection have been performed with intranasal administration of either live or heat-killed bacteria (15, 31, 32). In this study, we chose to test live bacteria administered orally, in order to develop a potential probiotic supplement for use in humans. Of note, parallel research demonstrated that the proinflammatory L. plantarum strain CNRZ1997 is compatible with all manufacturing and formulation technological processes (data not shown), and currently, a double-blind, randomized, controlled trial is in progress in order to investigate whether the consumption of this strain influences the severity of symptoms and the incidence and duration of common cold infections. In conclusion, our results suggest the feasibility of a large in vitro screening of bacterial strains using two cellular models (TNF-α-activated HT-29 cells and PBMCs), in order to determine their immunomodulatory properties based on a cytokine profile. Oral administration of the most proinflammatory strain (L. plantarum CNRZ1997) conferred protection against IAV infection. Further research is being conducted in our laboratory to identify the mechanism of action of this new candidate probiotic bacterium.
Noura Kechaou is the recipient of a grant from the Fonds Unique Interministériel (FUI; France). We thank the Fonds de Compétitivité des Entreprises (FCE; France) from the Direction Générale des Entreprises (contract no. 08 2 90 6473) for financial support.
We thank all the UEAR staff (INRA, Jouy-en-Josas, France) who helped us during in vivo experiments, in particular Laetitia Guedeville and Alice Belleivaire, as well as Edwige Bouguyon for virus preparation and quantification.
Published ahead of print 21 December 2012