 Research article
 Open Access
 Open Peer Review
Doseresponse model of murine typhus (Rickettsia typhi): time post inoculation and host age dependency analysis
 Sushil B Tamrakar^{1}Email author,
 Yin Huang^{1},
 Sondra S Teske^{1} and
 Charles N Haas^{1}
https://doi.org/10.1186/147123341277
© Tamrakar et al; licensee BioMed Central Ltd. 2012
 Received: 4 October 2010
 Accepted: 30 March 2012
 Published: 30 March 2012
Abstract
Background
Rickettsia typhi (R. mooseri) is the causative agent of murine typhus. It is one of the most widely distributed fleaborne diseases with a relatively mild febrile initial illness with six to 14 days of incubation period. The bacterium is gram negative and an obligate intracellular pathogen. The disease is transmitted to humans and vertebrate host through fleabites or via contact with infected feces. This paper develops doseresponse models of different routes of exposure for typhus in rodents.
Methods
Data from published articles were analyzed using parametric doseresponse relationship models. Doseresponse relationships were fit to data using the method of maximum likelihood estimation (MLE).
Results
Doseresponse models quantifying the effects of different ages of rats and time post inoculation in BALB/c mice were analyzed in the study. Both the adult rats (inoculated intradermally) and newborn rats (inoculated subcutaneously) were best fit by exponential models and both distributions could be described by a single doseresponse relationship. The BALB/C mice inoculated subcutaneously were best fit by BetaPoisson models. The time post inoculation analysis showed that there was a definite time and response relationship existed in this case.
Conclusions
Intradermally or subcutaneously inoculated rats (adult and newborn) models suggest that less than 1 plaqueforming unit (PFU) (1.33 to 0.38 in 95% confidence limits) of the pathogen is enough to seroconvert 50% of the exposed population on average. For the BALB/c mouse time post inoculation model, an average dose of 0.28 plaqueforming units (PFU) (0.75 to 0.11 in 95% confidence limits) will seroconvert 50% of the exposed mice.
Keywords
 Exponential Model
 Typhus
 Murine Typhus
 Maximum Likelihood Estimation Estimate
 Typhi Infection
Background
Murine typhus, also known as endemic typhus, is one of the most widely distributed flea borne diseases. The causative agent of murine typhus is Rickettsia typhi, previously known as R. mooseri. It is a relatively mild febrile illness with 6 to 14 days of incubation period [1–3]. It is considered less pathogenic than R. rickettsii and R. prowazekii (in terms of mortality rate), but R. typhi is virulent enough to cause severe infection in the elderly population [3]. The major reservoir of the pathogens is the rat (Rattus rattus and R. norvegicus) with the rat flea (Xenopsylla cheopis) as the main vector. Fleas are infected by transovarian transmission or acquire the contagion while feeding on an infected animal [4]. R. typhi is transmitted to the human body or vertebrate host by infected fleabites, or contamination of the broken skin, respiratory tract or conjunctivae of the host with infected feces or tissues during and after flea feeding [2, 3].
The flea once acquiring the infection remains infective for life. Interestingly, neither flea nor rat is harmed by the pathogens [2]. Although humans are infected mainly via rat fleas, murine typhus exists endemically in many places where rat and rat fleas are absent [3]. In the United States, the reported cases of murine typhus are focused in south and central Texas, Los Angeles and Orange County, California, where rats and rat fleas are rarely documented. The cat flea/opossum cycle may be one of the possibilities responsible for the disease[5].
The clinical symptoms of infection with R. typhi in humans are fever, headache, and myalgia. The fever lasts about 12 days in adults with temperature ranges between 102104F [6]. In severe cases the pathogen can cause meningoencephalitis, interstitial pneumonia and disseminated vascular lesions [7].
Many researchers have reported the response of animals to different doses of Rickettsia typhi in order to develop effective therapy and to study the pathology of infected animals. The purpose of this study is to develop doseresponse models and to compare the responses in term of age, route of infection and time post inoculation.
Methods
Since there have been no previously reported doseresponse relations, the aim of this study was to extract usable data from the literature and develop doseresponse curves. Criteria for data used in our analysis are described as:

Route of exposure is explicitly stated (such as inhalation, subcutaneous, intradermal, intravenous etc.)

Methods for dose estimation are described clearly

The number of subjects for each dose group is stated explicitly

The number of positive responses for each exposure route is explicitly stated

The criteria used to define a positive endpoint are stated

Pathogen is described in detail (source, strain)

The mode of preparation of pathogenic organisms is described
AringoJaramillo et al. (1984) carried out an experiment with R. typhi infection in adult and newborn laboratory rats. Nine different doses of R. typhi were transdermally and subcutaneously inoculated with seroconversion and death as the responses defined as endpoints [8]. Animals with an indirect fluorescent antibody titer of greater than or equal to 1:40 were considered to be seroconverted [9, 10]. However, no animals died.
Crist et al. (1984) experimented with R. typhi infection in normal and immune mice. Female BALB/c mice were subcutaneously inoculated with various doses of R. typhi and seroconversions on different days (after inoculation) were observed [10].
AringoJaramillo et al. (1988) conducted experimental inoculation of R. typhi in young rats of different age groups. Five different doses were inoculated orally in 3day, 7 day and 30 day old rats and seroconversion was recorded as the endpoint of response [9].
Analysis Method
Doseresponse analysis
Data Used
Pathogen/strain  Study/Reference  Mode of inoculation  Test animal/Response Organism/Reponses end point  Dose  Number of Test Animals  Positive Responses  Negative Responses 

R. typhi (Wilmington)  [10]  s.c.  BALB/c mice (seroconversion on day 9)  0.01(PFU)  10  0  10 
0.1  10  0  10  
1  10  0  10  
10  10  0  10  
100  10  0  10  
1000  10  0  10  
10000  10  10  0  
R. typhi (Wilmington)  [10]  s.c.  BALB/c mice (seroconversion on day 12)  0.01(PFU)  10  0  10 
0.1  10  0  10  
1  10  0  10  
10  10  0  10  
100  10  7  3  
1000  10  8  2  
10000  10  10  0  
R. typhi (Wilmington)  [10]  s.c.  BALB/c mice (seroconversion on day 15)  0.01(PFU)  10  0  10 
0.1  10  0  10  
1  10  1  9  
10  10  5  5  
100  10  9  1  
1000  10  9  1  
10000  10  10  0  
R. typhi (Wilmington)  [10]  s.c.  BALB/c mice (seroconversion on day 21)  0.01(PFU)  10  0  10 
0.1  10  0  10  
1  10  5  5  
10  10  8  2  
100  10  9  1  
1000  10  10  0  
10000  10  10  0  
R. typhi (Wilmington)  [10]  s.c.  BALB/c mice (seroconversion on day 28)  0.01(PFU)  10  0  10 
0.1  10  3  7  
1  10  8  2  
10  10  9  1  
100  10  10  0  
1000  10  10  0  
10000  10  10  0  
R. typhi (Ethiopian)  [8]  i.d.  Adult rat (seroconversion)  0.0435(PFU)  5  0  5 
0.435  5  1  4  
4.35  5  5  0  
43.5  5  5  0  
435  5  5  0  
4350  5  5  0  
43500  5  5  0  
R. typhi (Ethiopian)  [8]  i.d.  Newborn rat (seroconversion)  0.0435(PFU)  8  0  8 
0.435  8  2  6  
4.35  8  8  0  
43.5  8  8  0  
435  8  8  0  
4350  8  8  0  
R. typhi (Ethiopian)  [9]  Oral  Young rat 3 day old (seroconversion)  10(PFU)  3  1  2 
100  6  4  2  
1000  3  3  0  
10000  3  2  1  
100000  5  5  0  
R. typhi (Ethiopian)  [9]  Oral  Young rat 7 day old (seroconversion)  10(PFU)  3  1  2 
100  3  2  1  
1000  2  2  0  
10000  3  2  1  
100000  3  3  0  
R. typhi (Ethiopian)  [9]  Oral  Young rat 30 day old (seroconversion)  10(PFU)  3  0  3 
100  6  1  5  
1000  3  2  1  
10000  3  2  1  
100000  3  2  1 
The statistical programming language, "R" http://www.rproject.org was used for this computation. Two dose response models (exponential and BetaPoisson) were used [11]. Exponential and BetaPoisson MLE estimates were made using the BFGS algorithm. Confidence intervals to the bestfit models were determined via bootstrapping with 10,000 bootstrap iterations.
where P (d) is the probability of response at dose d and k is the probability that a single organism can survive and initiate infection.
where N_{50} is the median infective dose and α is the slope parameter for the BetaPoisson model. The equation (2) is derived from the exact BetaPoisson equation with certain assumptions [11, 12].
Goodness of fit for all models was determined by comparing the value of the optimized deviance to the critical χ^{2} value at degrees of freedom equal to the number of doses minus the number of fitted parameters at a 95% confidence level. Assessment of the statistical significance of improvement of fit that a two parameter model would provide over a single parameter model was made by comparing the reduction in minimized deviance with the critical χ^{2} value at 1 degree of freedom. Confidence intervals for the bestfit model were estimated via bootstrapping [11].
Pooling analysis was performed for the different animals and bacterial species to ascertain whether the data set had the same underlying distributions. A likelihood ratio test was used to determine if data could be pooled.
Time post inoculation analysis
In the experiment conducted by Crist and coinvestigators with BALB/c mice inoculated subcutaneously with R. typhi, the responses were also recorded at different post inoculation times [10]. Huang and Haas (2009) developed doseresponse models incorporating time post inoculation as an additional parameter [13]. In the exponential model, the k parameter is the probability that a single organism can survive and proliferate in order to initiate a response. It is well known that this is a timedependent process, and phenomenological responses of animals to bacteria vary not only with the initial dose of microorganisms, but also with the time post inoculation (TPI). In the BetaPoisson model, the N _{50} parameter is the dose required to produce a response in 50% of the exposed subjects. Directly related to the growth kinetics of a single organism, the initial dose to elicit response in 50% of the population (N _{50}) is also expected to vary with the time when the response is observed. To model these effects, Huang et al. [14, 15] set the parameter k and N _{50} equal to functions of time.
Results
Dose response model for Rickettsia typhi(Murine Typhus)
Doseresponse model of adult rat exposed intradermally to R. typhi, seroconversion as end point of response
Model Fit Comparison for Seroconversion in Intradermally inoculated adult rats
Data set  Number of Doses  Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters  Difference in deviances  χ ^{2} Value at 1 degree of freedom 

[8]  8  Exponential*  0.88  7  14.06  k = 0.756  0  3.84 
8  Beta Poisson  0.88  6  12.59  α = 01.16e8  
N_{50} = 0.91 
Doseresponse model of newborn rats exposed subcutaneously to R. typhi, seroconversion as the end point of response
Model Fit Comparison for Seroconversion in Subcutaneously Exposed Newborn rats
Data set  Number of Doses  Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters  Difference in deviances  χ ^{2} Value at 1 degree of freedom 

[8]  7  Exponential*  1.12  6  12.59  k = 0.831  0  3.84 
7  Beta Poisson  1.12  5  11.07  α = 4.2e7  
N_{50} = 0.83 
Doseresponse model of rats of different age groups inoculated orally to R. typhi, seroconversion as end point of response
AringoJaramillo et al. (1988) conducted experimental inoculation of R. typhi in young rats of different age groups to study the influence of R. typhi in maternal rats has on the offspring. Five different doses were inoculated orally in 3day, 7day and 30day old rats and seroconversion was recorded as the end point of response.
Doseresponse model of young rat (3days old) exposed orally to R. typhi, seroconversion as the end point of response
Model Fit Comparison for Seroconversion in Oral Exposed Young rat (3 days old)
Data set  Number of Doses  Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters  Difference in deviances  χ ^{2} Value at 1 degree of freedom 

[9]  5  Exponential  34.63  4  11.07  k = 0.0006  31.60  3.84 
5  Beta Poisson*  3.023  3  9.48  α = 0.286  
N_{50} = 25.7 
Doseresponse model of young rat (7 day old) exposed orally to R. typhi, seroconversion as the end point of response
Model Fit Comparison for Seroconversion in Oral Exposed Young rat (7 day old)
Data set  Number of Doses  Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters  Difference in deviances  χ ^{2} Value at 1 degree of freedom 

[9]  5  Exponential  66.172  4  11.07  k = 0.0149  63.83  3.84 
5  Beta Poisson*  2.34  3  9.48  α = 0.241  
N_{50} = 28.92 
Doseresponse model of young rat (30 days old) exposed orally to R. typhi, seroconversion as the end point of response
Model Fit Comparison for Seroconversion in Oral Exposed Young rat (30 day old)
Data set  Number of Doses  Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters  Difference in deviances  χ ^{2} Value at 1 degree of freedom 

[9]  5  Exponential  20.55  4  11.07  K = 4.2e5  19.55  3.84 
5  Beta Poisson*  1.0  3  9.48  α = 0.20  
N_{50} = 16.17 
Doseresponse model with post time inoculation analysis of BALB/c mice exposed subcutaneously to R. typhi, seroconversion as the end point of response
Best fit model for seroconversion on different days after inoculation in subcutaneously exposed BALB/c mice
Data  Number of Doses  Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters 

Seroconversion after day 12  7  Beta Poisson*  4.19  5  11.07  α = 0.79 
N_{50} = 94.80  
Seroconversion after day 15  7  Beta Poisson*  1.40  5  11.07  α = 0.57 
N_{50} = 10.46  
Seroconversion after day 21  7  Beta Poisson*  2.77  5  11.07  α = 0.60 
N_{50} = 1.67  
Seroconversion after day 28  7  Beta Poisson*  1.59  5  11.07  α = 0.72 
N_{50} = 0.28 
Timedoseresponse model for Murine Typhus in BALB/c mice
Huang et al. (Huang, Bartrand et al. 2009; Huang and Haas 2009) proposed a class of timedoseresponse models by incorporating the time post inoculation into the classical doseresponse models for microbial infection. The parameter k in the exponential doseresponse model and the parameter N _{50} in the BetaPoisson model were set equal to functions of time, which presumably model the in vivo bacterial kinetics for a single microorganism.
The proposed exponential and BetaPoisson timedoseresponse models from these prior studies can be given as:
where parameter dependency is given by $g\left(TPI\right)={e}^{{}_{{}^{\left({j}_{{}_{0}}/{\left(TPI\right)}^{{j}_{2}}+{j}_{{}_{1}}\right)}}}$
Pooling analysis
Pooling analysis was performed to ascertain whether different data sets could be described by a single doseresponse relationship. Different combinations of species, strains, routes of infection and hosts were pooled together and a likelihood ratio test was used to determine whether the data could be pooled or not.
Adult rat and Newborn rat Pooled Data
Data  Number of Doses  Best fit Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters 

Adult rat exposed intradermally  8  Exponential  0.88  7  14.06  K = 0.756 
Newborn rat exposed subcutaneously  7  Exponential  1.12  6  12.59  K = 0.831 
Pooled data  15  Exponential  2.0  14  23.68  K = 0.801 
Young rat of different ages Pooled Data
Data  Number of Doses  Best fit Model  Minimized Deviance  Degrees of Freedom  χ_{α,nk} ^{2}  Parameters 

Young rat 3day old  5  BetaPoisson  3.023  3  7.81  α = 0.286 
N_{50} = 25.7  
Young rat 7day old  5  BetaPoisson  2.34  3  7.81  α = 0.241 
N_{50} = 28.92  
Young rat 30day old  5  BetaPoisson  1.0  3  7.81  α = 0.20 
N_{50} = 16.17  
Pooled data  15  BetaPoisson  13.53  13  22.36  α = 0.213 
N_{50} = 106.14 
Discussion
Comparative values of ID50, ID10 and ID01
Pathogen  Host  Route  ID_{01}  ID_{10}  ID_{50} 

R. mooseri  Newborn rat  sc  0.012  0.12  0.83 
Adult rat  id  0.013  0.14  0.91  
Pooled data  0.012  0.13  0.86 
Similarly, AringoJaramillo et al. (1988) studied the influence of maternal R. typhi in young rats of different age groups inoculating R. typhi orally. All data for every age of rat (3 dayold, 7 dayold and 30 dayold) could be pooled and represented by a best fit BetaPoisson model. Scattered points at right of the graphs (Figure 6, 7 and 8) are high alpha values and those points tend to fit exponential model while majority 99.9% were at dense cluster. The result indicates that the rats inoculated orally show homogeneity in response and a single doseresponse relationship could describe all the age groups. There are no significant differences in ID_{50}s of individual age groups but the median infective value is significantly higher than rats inoculated intradermally or subcutaneously. The reason behind the higher ID_{50} may be the route of inoculation. The number of pathogens that reach endothelial cells via the oral route might significantly be less than the initial inoculation.
Crist et al. (1984) recorded post time inoculation responses while inoculating R. typhi (R. mooseri) to BALB/c mice. In days 06, there was no response. There was only one response at the highest dose on day 9. From day 12 to 28, there were systemic responses to corresponding doses. The data of seroconversion of BALB/c mice after days 12, 15, 21 and 28 were best fit to a BetaPoisson model. The post inoculation effect was analyzed with the data set using a time post inoculation model as described in prior work [13, 14]. Figure 10 shows the different doseresponse curves for different time (days) after inoculation.
Conclusion
Human murine typhus caused by Rickettsia typhi (R. mooseri) is an infectious disease that requires the fewest number of pathogens to initiate disease. The doseresponse models developed in this study support this effect. Intradermally or subcutaneously inoculated rats (adult and newborn) dose response models suggest that less than 1 PFU (ranging between 0.38 and 1.33 PFU as estimated within the models' 95% confidence limits) of the pathogen is enough to seroconvert 50% of the exposed population o average. The BALB/c mouse time post inoculation model also indicates that an average dose of 0.28 PFU (0.75 to 0.11 PFU within a 95% confidence interval) will cause seroconversion in of 50% of the exposed population with a mean time to effect of 28 days. The difference in median infectious dose (seroconversion) in adult rat and BALB/c mice is not a significant one and it may be because of the different strains of pathogen used to infect the test subjects. The model suggests that the higher the number of pathogens, the sooner the seroconversion. Pooling analysis of adult and newborn rats shows that there is no significant effect due to different routes of inoculation and can be described by the same doseresponse relationship. Similarly, pooling of orally inoculated young rats of different age groups indicates that there is no significant effect of age in serocenversion. However, there is an observed variation in response related to age groups seen in human case studies. According to AlAwadi et al., the most susceptible age group was the 1525 years, followed by the 26 to 44 year age group. But there is no definite pattern of variation [17, 18]. This is also the first study to incorporate time in a doseresponse model for murine typhus. The outcome may improve current understanding of in vivo bacterial dynamics, postexposure decisionmaking or as a component to assist epidemiological investigations.
Declarations
Acknowledgements
The advice of Sondra Teske (postdoctoral associate) is gratefully acknowledged. This work was performed as part of the Center for Advancing Microbial Risk Assessment (CAMRA). CAMRA is a US EPA/Department of Homeland Security Cooperative Center of Excellence funded under USEPA STAR grant R83236201. This work does not express official policy of either USEPA or the Department of Homeland Security.
Authors’ Affiliations
References
 Azad AF, Traub R: Transmission of Murine Typhus Rickettsiae by Xenopsylla cheopis, with Notes on Experimental Infection and Effects of Temperature. The American Journal of Tropical Medicine and Hygiene. 1985, 34: 555563.Google Scholar
 Azad AF: Epidemiology of Murine Typhus. Annu Rev Entomol. 1990, 35: 553570. 10.1146/annurev.en.35.010190.003005.View ArticlePubMedGoogle Scholar
 Houhamdi Fournier PE, Fang R, Raoult D: An experimental model of human body louse infection with Rickettsia typhi. Annal of the New York Academy of Sciences. 2003, 990: 617627. 10.1111/j.17496632.2003.tb07436.x.View ArticleGoogle Scholar
 Walker DH: Rickettsiae. Medical Microbiology. Edited by: Samuel Baron M. 1996, Irving, TX: University of Texas Medical Branch at Galveston, 4Google Scholar
 Azad AF, Radulovic S, Higgins JA, Noden BH, Troyer JM: Fleaborne Rickettsioses: ecologic considerations. Emerg Infect Dis. 1997, 3: 319327. 10.3201/eid0303.970308.View ArticlePubMedPubMed CentralGoogle Scholar
 McLeod MP, Qin X, Karpathy SE, Gioia J, Highlander SK, Fox GE, McNeill TZ, Jiang H, Muzny D, Jacob LS, et al: Complete genome sequence of Rickettsia typhi and comparison with sequences of other Rickettsiae. J Bacteriol. 2004, 186: 58425855. 10.1128/JB.186.17.58425855.2004.View ArticlePubMedPubMed CentralGoogle Scholar
 Fergie JE, Purcell K, Wanat D: Murine typhus in South Texas children. Pediatr Infect Dis J. 2000, 19: 535538. 10.1097/0000645420000600000009.View ArticlePubMedGoogle Scholar
 ArangoJaramillo S, Azad AF, Wisseman CL: Experimental infection with Rickettsia mooseri and antibody response of adult and newborn laboratory rats. AmJTrop Med Hyg. 1984, 33: 10171025.Google Scholar
 ArangoJaramillo S, Wisseman CL, Azad AF: Newborn rats in the murine typhus enzootic infection cycle: studies on transplacental infection and passively acquired maternal antirickettsial antibodies. AmJTrop Med Hyg. 1988, 39: 391397.Google Scholar
 Crist EA, Wisseman CL, Murphy JR: Characteristics of Rickettsia mooseri infection of normal and immune mice. Infect Immun. 1984, 43: 3842.PubMedPubMed CentralGoogle Scholar
 Haas CN, Rose JB, Gerba CP: Quantitative Microbial Risk Assessment. 1999, John Wiley and Sons, Inc (New York)Google Scholar
 Teunis PF, Havelaar AH: The beta poisson dose response model is not a singlehit model. Risk Anal. 2000, 20: 513520. 10.1111/02724332.204048.View ArticlePubMedGoogle Scholar
 Huang Y, Haas CN: Timedoseresponse models for microbial risk assessment. Risk Anal. 2009, 29: 648661. 10.1111/j.15396924.2008.01195.x.View ArticlePubMedGoogle Scholar
 Huang Y, Bartrand TA, Haas CN, Weir MH: Incorporating time postinoculation into a doseresponse model of Yersinia pestis in mice. J Appl Microbiol. 2009, 107: 727735. 10.1111/j.13652672.2009.04248.x.View ArticlePubMedGoogle Scholar
 Huang Y, Haas CN: Timedoseresponse models for microbial risk assessment. Risk Anal. 2009, 29: 648661. 10.1111/j.15396924.2008.01195.x.View ArticlePubMedGoogle Scholar
 Walker DH: Rocky Mountain spotted fever: a disease in need of microbiological concern. Clin Microbiol Rev. 1989, 2: 227240.View ArticlePubMedPubMed CentralGoogle Scholar
 Brown AE, Meek SR, Maneechai N, Lewis GE: Murine typhus among Khmers living at an evacuation site on the ThaiKampuchean border. AmJTrop Med Hyg. 1988, 38: 168171.Google Scholar
 AlAwadi AR, AiKazemi N, Ezzat G, Saah AL, Shepard C, Zaghloul T, Gherdian B: Murine typhus in Kuwait in 1978. Bull World Health Organ. 1982, 60: 283289.PubMedPubMed CentralGoogle Scholar
 The prepublication history for this paper can be accessed here:http://www.biomedcentral.com/14712334/12/77/prepub
Prepublication history
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.