Introduction

Antimicrobial resistance (AMR) comprises a rapidly emerging growing global threat that exerts enormous economic burdens. Based on recent estimates, AMR’s economic costs varied from £3–11 billion to US $100 trillion1. Furthermore, the emergence, widespread occurrence, and spread of antimicrobial resistance among animals, particularly within domestic pets, has raised high levels of global interest. The growing interest among AMR among animals is due to the high possibility of animal-human transmission of highly pathogenic bacteria possessing unique resistance mechanisms such as ESBL, carbapenemases, and mcr genes2,3,4,5. For example, human-specific pathogens have recently been isolated in dogs4. Furthermore, pathogenic strains possessing the same extended-spectrum and AmpC beta-lactamases (ESBL/AmpC) have been detected within the same households of humans and dogs5. More recently, our research group has identified highly resistant carbapenemase-producing and mcr-9 coharbouring bacteria associated with pets that had an active respiratory infection that could be transmitted between animals and humans3. Within this context, the bacteria associated with respiratory diseases in cats and dogs are of particular importance due to the widespread nature of these contagious diseases, and ease of transmission, in addition to the close proximity of pets and humans.

β-Lactam antibiotics comprise an important group of antibiotics that are widely utilized in human and animal fields, exhibiting strong therapeutic potential. Hence, WHO has identified several β-lactam classes, namely 3rd, 4th, and 5th generation cephalosporins, and carbapenems in addition, as being “critically important” for humans use6. Furthermore, several β-lactams are exclusively utilized in the field of veterinary medicine, including ceftiofur and cefquinome, which comprise 3rd and 4th generation cephalosporins7. Moreover, in the nation of Egypt, pets are typically treated with antibiotics manufactured for human use, including those forbidden within the veterinary practice, in order to reduce resistance rates. Hence, the heavy use of β-lactam across the veterinary and human practices has acted as a selective pressure facilitating the development of β-lactam resistance. Among the various mechanisms of β-lactam resistance, particular attention had been paid to ESBLs/AmpCs, as these especial enzymes are capable of hydrolyzing penicillins, cephalosporins, and monobactams thus facilitating resistance to multiple antimicrobial agents2,5,8. The faulty therapeutic outcomes and high mortality rates associated with ESBL/AmpC producers8, have bolstered the scientific interest to elucidate the molecular mechanism of β-lactam resistance, in particular in pets, which act as a potential consignor for human infection4,5.

Respiratory diseases in cats and dogs are descriptively known as cat flu and kennel cough. These conditions are complex, with infections caused by various viruses and bacteria, and typically initiated by stress-induced immunosuppression within infected pets9. Various Gram-negative bacteria have facilitated these particular challenging diseases and have been reported in several studies to be zoonotic10. Such as members of the Enterobacterales, alongside Pseudomonas aeruginosa that capable of acting as primary or secondary complicating pathogens within kennel cough and cat flu. Other Gram-negative pathogens could possibly also be involved and necessitate transtracheal wash or deep sampling to address their isolation, including Bordetella bronchiseptica, which is a commonly occurring bacterium associated with kennel cough and cat flu11. At present, a paucity of information is known and available regarding the bacterial resistance mechanisms associated with the respiratory diseases in pets. Hence, the present study was designed to determine the molecular mechanisms of resistance within Gram-negative bacteria isolated from cat flu and kennel cough cases within Egypt and intended to describe the relevant pathogen-specific signs. In our team’s present knowledge, this is the first report investigating the antimicrobial resistance mechanisms associated with respiratory diseases in pets.

Results

Clinical features associated with the isolates

Out of the 875 studied cats, 47 specimens (44 nasal swabs and 3 conjunctival swabs) were collected from 44 animals with symptoms of respiratory disease in a sample fraction equal to 5% (95% CI 3.7–6.7%). Thirty-one Gram-negative bacteria were isolated from 24 cats in a percentage comprising 54.5% (95% CI 38.9–69.7%). Various species of Gram-negative bacteria were isolated from the sampled diseased cats, either as singleton or mixed infections. Enterobacter cloacae (n = 15) occurred the most commonly in isolates from infected cats. followed by Escherichia coli (n = 10), and E. fergusonii, Klebsiella oxytoca, Leclercia adecarboxylate, Pantoea septica, P. aeruginosa, and Raouletella ornithinolytica, each occurring as a singleton infection from a single animal. Symptoms and their relation to the infecting bacteria are displayed in Table 1.

Table 1 Symptoms present in diseased cats.
Full size table

Out of the 503 dogs examined in the present study, 37 specimens (32 nasal and 5 conjunctival swabs) were harvested from 32 animals in a sampling fraction of 6.4%. Gram-negative bacteria were isolated from 27 animals (84.4%; 95% CI 67.2%–94.7%). Akin to cats, E. cloacae (n = 15) was the most commonly occurring Gram-negative bacteria isolated from dogs, followed by E. coli (n = 12), K. pneumoniae (n = 4), Citrobacter braakii, and R. ornithinolytica (n = 3 each), C. freundii (n = 2), and E. hormaechei, C. farmeri, K. oxytoca, and Serratia marcescens (n = 1 each).

Symptoms and their relation to the infecting bacteria are displayed in Table 2. The results of multinomial logistic regression showed that there is a significant association between the older ages and longer course of the disease (P < 0.05) (see the “Statistical analysis” section in the supplemental material). Furthermore, cats were almost significantly associated with shorter course than dogs (P < 0.08). On the other hand, there was no significant association found between the course of illness and isolate type (P > 0.05). There was no significant association either between animal species and specific type of isolates (P > 0.05) or between animal age and isolate type (P > 0.05). There was no significant association between the type of isolate and associated clinical signs except for E. cloacae which was found to be significantly associated with sneezing and ocular discharge (P < 0.05).

Table 2 Symptoms present in diseased dogs.
Full size table

Phenotypic characterization of the isolates

A multi-drug resistance profile, based on resistance to at least three antimicrobial classes was observed in 27% of the isolates (n = 20) (95% CI 17.4–38.6%). With the exception of carbapenems, the isolates demonstrated high resistance rates to almost all the β-lactam tested (Fig. 1, Supplementary Table S1). In regards, the isolates demonstrated resistance rates by 87.8%, 74.3%, 68.9%, 47.3%, and 19% for ampicillin (AMP), ceftriaxone (CRO), amoxicillin–clavulanic acid (AMC), cefoxitin (FOX), and cefoperazone (CFP), respectively (Fig. 1). Furthermore, the isolates also displayed a high resistance rate to tetracycline (TET) by 68.9%. However, the isolates exhibited low resistance rates to chloramphenicol (CHL, 16.2%), nalidixic acid (NAL, 10.8%), gentamicin (GEN, 8.1%), amikacin (AMK, 5.4%), ciprofloxacin (CIP, 4.1%), imipenem (IPM, 2.7%), and all isolates were susceptible to meropenem (MEM) (Fig. 1).

Figure 1
figure 1

Level of antimicrobial resistance to different antibiotics and phenotypic characters of the isolates. The isolates showed high resistance rates to tetracycline and almost all tested β-lactam antibiotics, except carbapenems with low resistance rates to other tested antibiotics. Phenotypic ESBLs and AmpCs production were detected within 40.5% (95% CI 29.3–52.6%), and 24.3% (95% CI 15.1–25.7%) of the isolates, respectively. On the other hand, all the isolates were negative for phenotypic carbapenemases production according to modified carbapenem inactivation assay (mCIM). Antibiotics: Ampicillin (AMP), ceftriaxone (CRO), tetracycline (TET), amoxicillin–clavulanic acid (AMC), cefoxitin (FOX), cefoperazone (CFP), chloramphenicol (CHL), nalidixic acid (NAL), gentamicin (GEN), amikacin (AMK), ciprofloxacin (CIP), imipenem (IPM), and meropenem (MEM).

Full size image

Of note, phenotypic ESBL production was detected within 40.5% (n = 30) (95% CI 29.3–52.6%) of the isolates, and 24.3% (n = 18) (95% CI 15.1–25.7%) of the isolates were phenotypically positive to AmpC production (Fig. 1, Supplementary Table S1). However, all the isolates were negative for carbapenemase production via modified carbapenem inactivation (mCIM) assay (Fig. 1, Supplementary Table S1).

Prevalence of ESBL-encoding genes

Varying ESBL-encoding genes were identified via PCR screening and DNA sequencing within 33.8% (25/74) (95% CI 23.2–45.7%) of the isolates (Fig. 2A, Supplementary Table S1). The ESBL with the highest abundance was blaSHV, which was identified within 25.7% (95% CI 16.2–37.2%) of the isolates; 18 isolates contained blaSHV-12; a single isolate possessed blaSHV-11. The second most prevalent ESBL-encoding gene was blaCTX-M, which was present within 17.6% (95% CI 9.7–28.2%) of the isolates, while five isolates contained blaCTX-M-15; four isolates contained blaCTX-M-14; 2 isolates contained blaCTX-M-37; 2 isolates possessed blaCTX-M-156. Nine isolates (12.2%; 95% CI 5.7–21.8%) contained blaTEM, while all isolates were confirmed to be blaTEM-1 which is not an ESBL-encoding gene. Five isolates registered as phenotypically positive for ESBL production, while genotypic screening confirmed that the isolates were either negative (n = 4) or contained blaTEM-1 (n = 1). It follows, the isolates may also possess other unchecked or rare ESBL-encoding genes.

Figure 2
figure 2

Molecular analysis of ESBL- and AmpC-encoding genes (A) and plasmid-mediated quinolone resistance genes (B). (A): blaSHV was the most prevalent ESBL-encoding gene (25.7%; 95% CI 16.2–37.2%) followed by blaCTX-M (17.6%; 95% CI 9.7–28.2%). Furthermore, 10.8% (95% CI 4.8–20.2%) of the isolates contained varying AmpC-encoding genes with blaCMY and blaACT identified ones within three isolates for each and two isolates were confirmed to possess blaDHA-1. (B) The plasmid-mediated quinolone resistance genes were identified within 64.9% (95% CI 52.9–75.6%) of the isolates, with qnrS was the most prevalent (44.6%; 95% CI 33.02–56.6%) followed by qnrA and qnrB detected within 10.8% and 9.5% of the isolates, respectively.

Full size image

Prevalence of AmpC-encoding genes

Multiplex PCR followed by single PCR and sequencing confirmed that 10.8% (8/74) (95% CI 4.8–20.2%) of the isolates contained varying AmpC-encoding genes. blaCMY and blaACT comprised the most prevalent identified ones within three isolates for each; two isolates contained blaCMY-169 (G492A); two isolates contained blaACT25/31; a single isolate possessed blaCMY-2; a single isolate contained blaACT23. Two isolates were confirmed to possess blaDHA-1 (Fig. 2A, Supplementary Table S1). Ten isolates registered as positive for phenotypic detection of AmpC production, while they were negative via PCR screening. These particular isolates may possess other rare or unchecked AmpC-encoding genes.

Prevalence of other resistance genes

The plasmid-mediated quinolone resistance genes (PMQR) were characterized within 64.9% (48/74) (95% CI 52.9–75.6%) of the isolates (Fig. 2b, Supplementary Table S1). The most prevalent qnr gene detected within this study comprises qnrS, found in 44.6% (33/74) (95% CI 33.02–56.6%) of the tested isolates. The second most prevalent qnr is qnrA detected within 10.8% of isolates, followed by qnrB, found within 9.5% of isolates. All the isolates were negative for the qepA gene and 16S rRNA methylases.

Discussion

Over the course of the past decade, antimicrobial resistance within the veterinary field has gained worldwide traction due to its important impact upon human health and the growing number of reports of animal-human transmission of antimicrobial resistance2,3,4,5,10. Within the nation of Egypt, our previous reports regarding AMR within clinical, veterinary, and food settings underscores high levels of resistance, and, in addition, the identification of unique resistance mechanisms within these settings3,8,12,13,14,15,16,17. Furthermore, the unregulated use of antimicrobial agents within the veterinary field, coupled with a lack of data regarding antimicrobial resistance within animals15, have resurrected interest driving investigation into their resistance mechanisms, with particular regards to pets. due to the growing concern of resistant bacteria or possible gene transmission between pets and humans.

The results of the present study indicate a widespread prevalence of multidrug-resistant bacteria within pets inflicted with kennel cough and cat flu by 27%, but lower than that previously identified within enteric E. coli from domestic pets in Portugal (49.7%)18. Furthermore, our team found it surprising that the isolates demonstrated extremely high resistance levels for different β-lactams, in particular AMP (87.7%), CRO (74.3%), AMC (68.9%), and FOX (47.3%,). The ampicillin resistance level is much higher than that previously reported from fecal E. coli from Portugal (51.3%)18 and Denmark (40%)19. Furthermore, β-lactams resistance was also comparable or higher than previously reported within healthy dogs in other European countries including the Netherlands, the United Kingdom, and France20,21,22. The isolates also demonstrated a high resistant rate to tetracycline (69%), which is also much higher than rates previously recorded in Portugal by 45.2%18. The widespread prevalence of β-lactam and tetracycline resistance among Gram-negative bacteria associated with respiratory diseases of pets may possibly be attributed to dependence on these drugs as a cheap and readily available remedy for the treatment of pets’ infections8. On the other hand, our results demonstrated low levels of resistance to other drugs, in particular for carbapenems, which are strictly utilized only for humans in Egypt13, and aminoglycosides and chloramphenicol, which may be due to their limited use, owing to their costs or their complications8.

ESBL production is regarded as the major mechanism responsible for β-lactam resistance2,5,8. Our results showed that 40.5% of them were phenotypically positive for ESBL production, with 33.8% of the isolates containing at least one ESBL-encoding gene (Supplementary Table S1). Our results are much higher than levels previously identified within Enterobacterales from fecal swabs gathered in Spain/France and Switzerland, where 5.3% and 2–2.9% of pets possessed ESBL-encoding genes, respectively2,23. Furthermore, our results are comparable to those gathered in China, where phenotypic ESBL production ranged from 24.5% in healthy animals to 54.5% in infected animals23. The ESBL genes most widely identified were blaSHV and blaCTX-M, which is in agreement with previous reports from various countries, such as Spain, France, the Netherlands, the United Kingdom, China, and Tunisia2,20,21,22,24,25. Of note, the majority of the identified blaSHV variants is comprised of blaSHV-12, which are considered the most widespread among the clinical nosocomial isolates of Enterobacterales26. Furthermore, blaSHV and blaCTX-M were also the most widespread within clinical settings and food production in Egypt8,14, indicating that they are the major circulating ESBL within Egypt. Although blaCTX-M-15, blaCTX-M-14, and blaSHV-12 were identified from pets in different countries2,19,21,23, this is the first report of the identification of blaCTX-M-37, blaCTX-M-156, and blaSHV-11 from pets. Furthermore, this is the first report of blaCTX-M-37, and blaCTX-M-156 identification within Egypt and the entirety of Africa.

Of importance, in the present study, we report ESBL-producing R. ornithinolytica and L. adecarboxylate, and to our present knowledge, this is the first report of their identification stemming from pets in Africa. Both are rare human pathogens and are both associated, but recently are being increasingly reported within clinical cases, which may be associated with severe, life-threatening infections, especially cancer and immunocompromised patients26,27. The identification of both pathogens within pets in Egypt poses a considerable public health hazard, with the high possibility of human transmission, facilitating the development of severe, untreatable infections.

The results of the present study demonstrated that 24.3% of the isolates registered as phenotypically positive for AmpC production, with 10.8% carrying AmpC-encoding genes. These results are higher than previously reported levels from pets in the UK, where 16% of the isolates were phenotypically positive, and only 4% harbored blaAmpC genes21. Furthermore, the prevalence of blaAmpC genes identified within this study is much higher than those identified within other European countries, including France/Spain, Denmark, and France, which ranged from 2.1 to 5.4%2,19,22. Interestingly, the majority of the identified blaAmpC genes from pets were blaCMY-2, which was detected in pets in France/Spain2, Denmark19, the Netherland20, France22, Tunisia25, and Italy28. On the other hand, blaDHA and blaACT are rare blaAmpC genes within the field of veterinary medicine, and to the best of our team’s present knowledge, this is the first report of their identification alongside blaCMY-169 (G492A) from pets in Egypt, the Middle East, and even the entirety of Africa. blaDHA-1 has been recently identified within pets in Italy28, and blaACT gene was identified in Germany29.

Quinolones comprise a major group of antibiotics that are commonly used in both human and veterinary medicine and have been identified as one of the most critically important drugs for humans6. Due to this fact, in recent years PMQR genes have gained special attention due to their high transmission rates between bacteria via plasmid8. Unsurprisingly, our results highlight a high prevalence of qnr genes within pets in Egypt by 64.9%, with qnrS occurring as the most prevalent (44.6%), supporting our previous reports from clinical settings and food settings in Egypt8,14. Our results are much higher than those previously reported within other countries, where qnr genes could not be identified in pets from France/Spain and the UK2,21. In a recent large-scale study within Europe, qnr genes were detected in 20% (32/160) of enrofloxacin non-wild type (resistant) Enterobacterales, which represent 3.8% of the total isolates (32/843)30. However, qnr genes induce low levels of quinolone resistance, but mediate the selection mutations leading to high quinolone resistance31. Our results are inconsistent with the previous findings, as 14.3% and 16.3% of qnr containing isolates were resistant, and intermediate to nalidixic acid, respectively, and 6.1% and 49.9% of the isolates were resistant and intermediate to ciprofloxacin (data not shown). Our findings confirm our previous studies regarding the lower role of the qepA gene and 16S rRNA methylases in the mediation of quinolones and aminoglycosides resistance, respectively6,8.

Conclusion

In summary, the present study reveals wide variation within clinical signs, and a high prevalence of ESBL, AmpC, and PMQR genes across Gram-negative bacteria associated with cat flu and kennel cough within Egypt. Furthermore, unique resistance mechanisms and bacterial species were isolated and identified for the first time within Egypt and the entirety of Africa. Our results are of great clinical concern, as pets could possibly act as a potential reservoir for these antimicrobial resistance determinants that could further disperse them into the human environment. Taking into consideration that the majority of the resistant determinants were previously identified from clinical settings and food in Egypt, the query of animal-human transmission is still needed to be confirmed through future studies.

Materials and methods

Animals and sampling

In the present study, a total of 1,378 pet animals, including cats (n = 875) and dogs (n = 503) hailing from five shelters located in Giza which collect stray pets, as well as those released by their owners in lower and middle Egypt, were selected based upon information regarding outbreaks of respiratory illness. A total of 76 diseased animals (44 cats and 32 dogs) were subjected to precise clinical examination between March and April 2017, and 84 swabs—including 76 nasal swabs and 8 conjunctival swabs—were aseptically collected. All the swabs were transferred to the lab at 4 °C under aseptic conditions for further analysis. All experimental protocols were carried out according to the Kafrelsheikh University Animal Experimentation Regulations, and approved by the Committee on Animal Experiments, Faculty of Veterinary Medicine, Kafrelsheikh University.

Bacterial isolates

A total of 74 Gram-negative bacteria were recovered from the nasal cavity (n = 65 isolates) and the conjunctiva (n = 9 isolates) of 51 pets, including dogs (n = 43 isolates) and cats (n = 31 isolates), that were tested in this study. All the isolates were identified using MALDI-TOF MS (Matrix Assisted Laser Desorption Ionization-Time of Flight Mass Spectrometry). The isolates were detected to comprise Enterobacter spp. (n = 31), Escherichia spp. (n = 23), Klebsiella spp. (n = 6), Citrobacter spp. (n = 6), R. ornithinolytica (n = 4), Pseudomonas aeruginosa (n = 1), S. marcescens (n = 1), L. adecarboxylate (n = 1), and P. septica (n = 1) (Fig. 3).

Figure 3
figure 3

Bacterial isolates identified in this study. Enterobacter species include E. cloacae (n = 30) and E. hormaechei (n = 1). Escherichia species include E. coli (n = 22) and E. fergusonii (n = 1). Klebsiella species include K. pneumoniae (n = 4) and K. oxytoca (n = 2). Citrobacter species include C. braakii (n = 3), C. freundii (n = 2), and C. farmeri (n = 1). Others include S. marcescens (n = 1), P. septica (n = 1), L. adecarboxylate (n = 1), and P. aeruginosa (n = 1).

Full size image

Antimicrobial sensitivity testing

The Kirby–Bauer disc diffusion method was utilized in order to evaluate all of the bacterial isolates for the antimicrobial susceptibility phenotypes to different classes of antibiotics, according to the Clinical and Laboratory Standards Institute (CLSI)32. Various antibiotic discs representing the majority of antibiotic families were tested, including AMC (20–10 µg), AMK (30 µg), AMP (10 µg), CHL (30 µg), CRO (30 µg), CFP (75 µg), FOX (30 µg), CIP (5 µg), GEN (10 µg), IPM (10 µg), MEM (10 µg), NAL (30 µg), and TET (30 µg). All of the discs were purchased from Becton, Dickinson and Company Sparks (MD 21,152, USA).

Phenotypic detection of carbapenemases production

Phenotypic carbapenemase detection was tested via the mCIM method as previously described33. Briefly, the bacterial isolates were cultured overnight at 37 °C and a 1-μl inoculation loop was added to 2 ml of tryptic soy broth within a tube. Following 15 s of vortexing, a 10-μg MEM disc was added to the bacterial suspension. The bacterial submission containing the MEM disc was incubated at 37 °C for 4 h. Before the end of the incubation time, the indicator organism (0.5 McFarland standard E. coli ATCC 25922) was inoculated at the Mueller Hinton agar plate (MHA). After 4 h incubation of bacterial submission containing the MEM disk, the MEM disk was placed on the inoculated MHA, followed by incubation at 37 °C for 18–24 h. After the incubation period, the result was evaluated by measuring the inhibition zone around the MEM disc.

Phenotypic detection of ESBL and AmpC production

Phenotypic ESBL and AmpC detection were evaluated utilizing the D68C AmpC and ESBL detection kits (Mast Diagnostics, Mast Group Ltd., Merseyside, U.K.) following the instruction of the producing company and as previously reported34. Briefly, the bacterial isolates were cultured overnight at 37 °C on MHA plates, followed by their inoculation at a concentration of 0.5 McFarland standard on MHA plates utilizing a sterile swab. The D68C AmpC and ESBL detection kits contain four different discs, Disc A contains 10 μg of cefpodoxime, disc B 10 μg of cefpodoxime and ESBL inhibitor, disc C 10 μg of cefpodoxime and AmpC inhibitor, and disc D 10 μg of cefpodoxime and both AmpC and ESBL inhibitors. The four discs were placed on the inoculated plates, followed by incubation at 37 °C for 24 h. The final results were evaluated by measuring and contrasting the inhibition zones’ diameters around the four discs. In the phenotypic assays, E. coli ATCC 25922 was employed as a negative control and an indicator, while ESBL-, AmpC-, and carbapenemases-producers from our previous reports were employed as positive controls8,12,13.

DNA preparation for PCR experiments

The DNA was prepped by boiling lysates following the method as previously indicated14. Briefly, a full 1-μl inoculation loop of bacteria culture incubated overnight on MHA was mixed with 100 µl of sterilized, distilled water, followed by boiling for 8 min. The bacterial suspension was thoroughly shaken then centrifuged, and the supernatant was utilized as a template for PCR experiments.

Molecular characterization of the isolates

Genetic characterization of the isolates was conducted by PCR experiment and sequencing. The isolates were tested for a wide range of antimicrobial resistance mechanisms via single and/or multiplex PCR as previously described35,36,37,38,39,40 (Supplementary Table S2). The ESBL-phenotypic positive isolates were examined for ESBL-encoding genes (blaCTX-M, blaSHV, and blaTEM) and the AmpC-phenotypic positive isolates were tested for plasmid-mediated AmpC β-lactamase genes (blaACC, blaLAT, blaCMY, blaBIL, blaMOX, blaDHA, blaMIR, blaACT, and blaFOX), and all the isolates were checked for plasmid-mediated quinolone resistance genes (qnrA, qnrB, and qnrS), qepA gene which is quinolone efflux pump determinant, and 16S rRNA methylases (rmtA, rmtB, rmtC, rmtD, armA, and npmA).

Sequencing and data analysis

Target genes were confirmed via single PCR, and PCR fragments were then purified from PCR product or agarose gels using CICA GENEUS PCR & Gel Prep Kit (KANTO CHEMICAL CO., INC., Tokyo, Japan). Following sequencing, the similarity search was performed with the sequenced data using BLAST analysis (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Statistical analysis

The association between the course the illness and animal species, isolate type, and the animal age was examined using a multinomial logistic regression model. The association between animal species and the isolate type, age of infected animals and the isolate type, and clinical signs and isolate type were examined using chi-square test. All analyses were carried out using IBM SPSS Statistics for Windows version 21.0. (IBM SPSS Inc, Armonk, NY) and SAS 9.2 (Institute Inc 2008). A P value < 0.05 was considered statistically significant. (see the “Statistical analysis” section in the supplemental material).