Dura mater grafts used to repair the dural defects at neurosurgery can cause fatal disease years to decades later. The tragedy of dura mater graft-associated Creutzfeldt-Jakob disease (dCJD) was considered to be nearly over. However, recent progress in research of the pathogenesis of dCJD has revealed previously unrecognized problems. In this review, we summarize the past and future problems surrounding dCJD.
Creutzfeldt-Jakob disease (CJD) is a lethal transmissible neurodegenerative disease. The central event in the pathogenesis of CJD is a conformational change of the normal cellular isoform of prion protein (PrPC) into an abnormal infectious isoform of prion protein (PrPSc) . The conformational change of PrPC can occur due to either one of three causes: spontaneous conversion in sporadic CJD (sCJD), mutations in the PRNP gene in genetic CJD, or infection with PrPSc in iatrogenic CJD and variant CJD.
One of the most frequent sources of iatrogenic PrPSc infection is dura mater grafts obtained from human cadavers undiagnosed as CJD. The sum of dCJD (228 cases) and growth hormone-associated CJD (226 cases) accounts for 97% of total iatrogenic CJD cases . A single brand of dura mater graft, Lyodura®, was used for all the dCJD cases in whom the brand name was identified. Although the causative dura mater grafts were manufactured by a German company, 62% (142 cases) of total dCJD cases have been found in Japan [2, 3]. Persistent efforts of a Japanese CJD surveillance team have clarified the outline of dCJD outbreaks. The onset of Japanese dCJD patients peaked in the late 1990s, and most of the patients had received the grafts during 1983–1987, while as many as 100,000 persons received the Lyodura® grafts during this period [4, 5]. In the process of conducting this elaborate survey, a puzzling mystery about dCJD emerged.
A mystery about dCJD
There is growing evidence that dCJD can be divided into two subgroups that exhibit distinct clinical and neuropathological phenotypes, with the majority (68%) represented by a non-plaque-type of dCJD (np-dCJD) and the minority (32%) by a plaque-type of dCJD (p-dCJD) (Figure 1) [6–12]. The clinicopathological features of np-dCJD are identical to those of classical sCJD, whereas p-dCJD is characterized by (i) ataxic gait as an initial symptom, (ii) slow progression of neurological symptoms, (iii) absence or late occurrence of periodic sharp-wave complexes (PSWC) on electroencephalogram (EEG), and (iv) widespread PrPSc amyloid plaques in the brain [11–16]. There is no significant difference in gender composition, site of graft, age at grafting, and year of grafting between the two subgroups [11, 12].
sCJD also shows wide phenotypic heterogeneity, and its clinicopathological phenotypes are determined by the genotype at polymorphic codon 129 of the PRNP gene and type of PrPSc in the brain (Figure 2) [17, 18]. The codon 129 of the PRNP gene shows methionine (M)/valine (V) polymorphism. Two types of PrPSc (type 1 and type 2) are distinguishable according to the size of the proteinase K-resistant core of unglycosylated PrPSc (21 and 19 kDa, respectively), reflecting differences in the proteinase K-cleavage site (at residues 82 and 97, respectively) . On the other hand, the two distinct phenotypes of dCJD had been considered to be unrelated to their PRNP genotype or type of PrPSc in the brain . In Japan, almost all dCJD patients had the same genotype, i.e., homozygous for methionine at codon 129 (129 M/M), except two heterozygotes , and the type of PrPSc in their brains had been reported as type 1 [6, 10, 11]. The reason for the existence of two distinct subgroups in dCJD had remained elusive.
Solving the mystery
In 2003, an unusual p-dCJD case was reported . This patient showed the accumulation of unusual PrPSc with intermediate electrophoretic mobility between types 1 and 2 PrPSc. Then, we reevaluated the biochemical properties of PrPSc in the two subgroups of dCJD and found that the size of PrPSc from p-dCJD was invariably smaller than that of type 1 PrPSc from np-dCJD (Figure 3) . This intermediate-sized PrPSc was designated as type i PrPSc.
To resolve the mystery of the existence of two distinct subgroups in dCJD, we hypothesized that they might be caused by infection with different PrPSc strains from distinct sCJD subgroups. According to the PRNP genotype and type of PrPSc in the brain, sCJD is classified into six subgroups (MM1, MV1, VV1, MM2, MV2, or VV2) . MM1 and MV1, which are the predominant subgroups in sCJD, show the same clinicopathological features. Meanwhile, MM2 can be divided into three subgroups based on histopathological criteria (MM2T, thalamic form showing characteristic atrophy of thalamic and inferior olivary nuclei; MM2C, cortical form showing a predominant cortical pathology; or MM2T + C, mixed form) [17, 21]. In addition, MV2 is also divided into three subgroups based on histopathological criteria (MV2K showing kuru type PrPSc amyloid plaques, MV2C showing a predominant cortical pathology, or MV2K + C showing mixed histopathology) [17, 18]. The clinicopathological features of np-dCJD, such as short duration of illness, PSWC on EEG, or diffuse synaptic-type PrPSc deposition in the brain, are identical to those of sCJD-MM1/MV1. In contrast, the clinicopathological features of p-dCJD, such as ataxic gait as an initial symptom, slow progression of neurological symptoms, absence or late occurrence of PSWC on EEG, or formation of PrPSc plaques in the brain, are similar to those of sCJD-VV2, −MV2K, or -MV2K + C. These similarities raised the possibility that np-dCJD might be caused by infection with sCJD-MM1/MV1, whereas p-dCJD might be caused by infection with sCJD-VV2, −MV2K, or -MV2K + C.
To test this possibility, we examined the transmission properties of the dCJD and sCJD subgroups using humanized mice carrying human PrP with either the 129 M/M or V/V genotype [20, 22, 23]. In these transmission experiments, p-dCJD and sCJD-VV2, −MV2K, or -MV2K + C were identical in the transmissibility to the PrP-humanized mice (Table 1, Figure 4a) and in the neuropathological and biochemical features in the inoculated mice (Figure 4b, c). By contrast, np-dCJD showed the same transmission properties as sCJD-MM1. In particular, the 129 M/M mice inoculated with sCJD-VV2, −MV2K, or -MV2K + C material showed widespread PrPSc plaques and type i PrPSc accumulation similar to the p-dCJD patients, whereas the 129 M/M mice inoculated with sCJD-MM1 material showed diffuse synaptic-type PrPSc deposition and type 1 PrPSc accumulation similar to the np-dCJD patients. Thus, these animal models support the hypothesis that the origin of np-dCJD is sCJD-MM1/MV1 and that of p-dCJD is sCJD-VV2, −MV2K, or -MV2K + C. Indeed, the incidence rate of p-dCJD (32%) among total dCJD is close to the sum total of the incidence of sCJD-VV2 (15%), −MV2K (8%), and -MV2K + C (3%) .
Molecular basis of the generation of two distinct subgroups in dCJD
At the molecular level, np-dCJD contains type 1 PrPSc with the codon 129 M genotype (denoted as M1 PrPSc), whereas p-dCJD contains type i PrPSc with the codon 129 M genotype (Mi PrPSc) (Table 2). Meanwhile, sCJD-MM1/MV1 contains M1 PrPSc, and sCJD-VV2 contains type 2 PrPSc with the codon 129 V genotype (V2 PrPSc). Recently, we found that sCJD-MV2K contains Mi PrPSc and V2 PrPSc, whereas sCJD-MV2K + C also contains type 2 PrPSc with the codon 129 M genotype and cortical pathology (M2C PrPSc) in addition to Mi PrPSc and V2 PrPSc (Table 2) . M2 PrPSc can be divided into two subgroups based on histopathological phenotypes. M2C PrPSc causes a predominant cortical pathology in sCJD-MM2C, −MV2C, or -MV2K + C, whereas M2T PrPSc causes atrophy of thalamic and inferior olivery nuclei in sCJD-MM2T.
The generation of M1 PrPSc in np-dCJD is simply due to the infection with M1 PrPSc from sCJD-MM1/MV1. On the other hand, the generation of Mi PrPSc in p-dCJD is rather complicated. Transmission of V2 PrPSc, i.e., sCJD-VV2, to the 129 M/M mice generated Mi PrPSc (Figure 4c) . Similarly, transmission of Mi PrPSc, i.e., p-dCJD, to the 129 M/M mice also generated Mi PrPSc. Therefore, both V2 PrPSc and Mi PrPSc can generate Mi PrPSc if transmitted to individuals with the 129 M/M genotype. Indeed, transmission of sCJD-MV2K containing Mi PrPSc and V2 PrPSc to the 129 M/M mice also generated Mi PrPSc (Figure 4c). Meanwhile, sCJD-MV2K + C contains M2C PrPSc besides Mi PrPSc and V2 PrPSc (Table 2). However, M2C PrPSc lacks or has very low infectivity and does not affect the transmission properties of the coexisting PrPSc. Therefore, the transmission of sCJD-MV2K + C to the 129 M/M mice can also result in the generation of Mi PrPSc (Figure 4c). Taken together, Mi PrPSc in p-dCJD is generated by infection with Mi PrPSc and/or V2 PrPSc from sCJD-VV2, −MV2K, or -MV2K + C. It is noteworthy that Mi PrPSc can be observed in the 129 M/M mice inoculated with V2 PrPSc but not in sCJD patients with the 129 M/M genotype, suggesting that Mi PrPSc in sCJD-MV2K or -MV2K + C is also generated by V2 PrPSc seed-dependent conversion but not by spontaneous conversion of the 129 M PrPC. Therefore, the primary origin of Mi PrPSc is V2 PrPSc. This can account for the similarities in transmission properties between Mi PrPSc and V2 PrPSc. Thus, M1 PrPSc in np-dCJD and Mi PrPSc in p-dCJD are completely different with regard to the neuropathological phenotypes, biochemical features, and transmission properties, reflecting their distinct PrPSc origins. In contrast to M1 PrPSc, which is the most common PrPSc observed in sCJD patients with the 129 M/M genotype, Mi PrPSc has never been observed in sCJD patients with the 129 M/M genotype. Therefore, the detection of Mi PrPSc can be sound evidence of iatrogenic infection in individuals with the 129 M/M genotype and would contribute to reliable surveillance of iatrogenic cases such as p-dCJD.
To verify experimentally that Mi PrPSc originates from V2 PrPSc and its transmission properties are identical to those of the parental V2 PrPSc, we performed a modeling study using PrP-humanized mice (Figure 5a) . As described above, the 129 M/M mice inoculated with V2 PrPSc showed widespread PrPSc plaques and Mi PrPSc accumulation in the brain as an experimental model of p-dCJD. We then inoculated the Mi PrPSc from these mice into other PrP-humanized mice with either the 129 M/M or V/V genotype. This secondary passage revealed that the transmission properties of the Mi PrPSc, i.e., 129 M/M mouse-passaged sCJD-VV2, are identical to those of the parental V2 PrPSc. In particular, although the incompatibility of the codon 129 genotypes between host and inoculum usually results in a prolonged incubation period , the 129 V/V mice inoculated with the Mi PrPSc showed a shorter incubation period compared with the 129 M/M mice (Table 1). Moreover, the altered neuropathological phenotype and biochemical properties at the primary passage in the 129 M/M mice reverted to the original ones in the secondary passage in the 129 V/V mice (Figure 5b, c). Thus, this modeling study shows that (i) V2 PrPSc infection in a host with the incompatible codon 129 M/M genotype generates an unusual PrPSc with altered conformational properties, i.e., Mi PrPSc, (ii) the emerging Mi PrPSc retains the memory of the parental V2 PrPSc within its conformational properties, and (iii) the parental V2 PrPSc re-emerges and proliferates rapidly if the Mi PrPSc is transmitted to the original host with the codon 129 V/V genotype. This phenomenon, designated as traceback, can be a useful tool to identify the origin of PrPSc infection if atypical PrPSc emerges in the future [20, 22, 26].
Our transmission studies resolved the complicated pathogenesis of dCJD. However, they have also revealed several issues surrounding dCJD that need to be addressed in the future.
First, the numbers of p-dCJD patients may increase in the future. The experimental p-dCJD model, i.e., the 129 M/M mice inoculated with Mi PrPSc and/or V2 PrPSc from sCJD-VV2, −MV2K, or -MV2K + C, showed a longer incubation period compared with the np-dCJD model, i.e., the 129 M/M mice inoculated with M1 PrPSc from sCJD-MM1 (Table 1). This raises the concern that additional p-dCJD patients, who are presenting still at the subclinical stage, may emerge after a longer incubation period in the future. Although the numbers of patients with newly developed dCJD have dropped off, continuous surveillance will be required to find remaining p-dCJD cases.
Second, the potential risks of secondary infection from dCJD, particularly from p-dCJD, may be considerable. As described above, the transmission studies raise a concern about the existence of subclinical p-dCJD patients. dCJD patients may undergo more than one neurosurgical operation due to their underlying diseases (the primary disease for which the neurosurgery was performed) . In addition, p-dCJD patients may be more frequently autopsied because the clinical features of p-dCJD are atypical compared with those of classical sCJD . These facts suggest that there may be considerable risk of secondary infection from p-dCJD patients. Individuals with the 129 V/V genotype may be more vulnerable to the infection with Mi PrPSc from p-dCJD, as suggested by the fact that the 129 V/V mice were highly susceptible to Mi PrPSc in the transmission study (Table 1). Additionally, 129 M/M individuals may be also affected after a prolonged incubation period, as suggested by the high attack rate (100%) of the 129 M/M mice inoculated with Mi PrPSc. Therefore, secondary infection from p-dCJD can occur regardless of the codon 129 genotype. Comprehensive analysis of the distribution of PrPSc in the peripheral tissues of p-dCJD patients will be also required to assess the potential risks of secondary infection.
Finally, the efficacy of the current PrPSc decontamination procedures against Mi PrPSc needs to be tested in the future. Mi PrPSc in p-dCJD and M1 PrPSc in np-dCJD differ in the sizes of the proteinase K-resistant core, suggesting their conformational differences. Moreover, their parental PrPSc strains are also different. Different PrPSc strains can show different thermostability [27, 28] and different susceptibility to the decontamination procedures . To prevent the spread of secondary infection from dCJD patients to medical workers or other patients, adequate decontamination and disinfection of the instruments used for neurosurgery or autopsy are essential. However, the current PrPSc decontamination procedures were developed using scrapie isolates and tested using CJD isolates other than p-dCJD [30–32]. Therefore, further studies using Mi PrPSc will be needed to assess the effectiveness of the current procedures. For this purpose, sensitive detection systems for Mi PrPSc are also prerequisite to evaluating quantitatively the reduction of infectivity after the decontamination procedures. Real-time quaking-induced conversion [33, 34], protein misfolding cyclic amplification [35–39], or transgenic mice overexpressing human PrP with the 129 V genotype  might be useful to detect the reduced infectivity of Mi PrPSc at high sensitivity. Using such sensitive detection systems, effective decontamination procedures for Mi PrPSc can be established in the future.
Recent progress in the study of the pathogenesis of dCJD has revealed that the two distinct subgroups of dCJD are caused by infection with different PrPSc strains of sCJD, i.e., np-dCJD caused by M1 PrPSc from sCJD-MM1/MV1 and p-dCJD caused by Mi PrPSc and/or V2 PrPSc from sCJD-VV2, −MV2K, or -MV2K + C. Studies have also revealed previously unrecognized problems such as the considerable risks of secondary infection from dCJD, particularly from p-dCJD. To prevent secondary infection from p-dCJD, the effectiveness of the current decontamination procedures should be tested urgently using sensitive Mi PrPSc detection systems.