Currently, definitive pathologic diagnosis of autophagic vacuolar myopathy requires electron microscopic identification of moderate to frequent well-developed autophagic vacuoles. In this study, we demonstrate that immunohistochemistry for LC3 and/or p62 can be used to detect autophagosome accumulation by light microscopy, thus providing a valuable diagnostic tool for this group of disorders.
In both autophagic myopathy and drug-treated control muscle, LC3 and p62 staining was largely punctate in nature. The punctate pattern of LC3 staining reflects the association of LC3-II with the membranes of early autophagosomes, whereas p62 puncta correspond to the accumulation of protein aggregates within early autophagic (LC3-positive) vesicles; hence, the increased punctate staining seen with these markers corresponds to autophagosome buildup 
. In drug-treated control specimens, the puncta were largely small and distributed evenly throughout the sarcoplasm (). In autophagic myopathy specimens, on the other hand, the puncta were larger and primarily (although not exclusively) located in the center of the fiber, creating a linear structure aligned with the fiber's long axis. This linear pattern of staining may be related to the linear intermyofibrillar organization of microtubules (along which autophagosomes propagate) 
, and likely corresponds to the centrally-located zone of autophagic buildup that was detected in autophagy-deficient (ATG5−/−) murine skeletal muscle 
. In addition to these qualitative differences in the pattern of staining between autophagic myopathy and drug-treated control samples, there was a significant difference in the percentage of LC3 and p62-positive fibers between the two drug-treated groups (no staining was observed in the normal control group). In autophagic myopathy subjects, staining was present in a large fraction of muscle fibers (12.5% to 95% for LC3; 12.5% to 93% for p62). In contrast, the majority of drug-treated control samples showed staining in less than 5% of muscle fibers, with only 1 of 7 specimens showing staining in up to 13% of fibers.
Based on the ROC analysis of our data, 100% specificity can be achieved by setting a diagnostic threshold at 15% LC3-positive muscle fibers. With this threshold, sensitivity is 80% (meaning that 20% of autophagic myopathies would be missed). In contrast, threshold of 8% LC3-positive fibers would achieve 100% sensitivity and 86% specificity (i.e., 14% of non-specific cases would be falsely diagnosed as autophagic vacuolar myopathy). Thus, using LC3 immunohistochemistry alone, a subset of cases with 8% to 15% of LC3-positive fibers would fall in the diagnostic “gray zone". For cases in this range, diagnostic accuracy can be improved by using additional diagnostic modalities, such as p62 immunohistochemistry and electron microscopy, and integrating the information gleaned from the standard histology and histochemistry stains (for example, the presence or absence of basophilic cores). In this context, it is worth noting that a single specimen from the drug-treated control group (#12, ) showed a degree of staining (13% for LC3, 10.5% for p62) that overlapped with the range generally seen in the autophagic myopathy group. This specimen was designated a drug-treated control because only rare well-developed autophagic vacuoles were identified on ultrastructural examination. However, light microscopy showed scattered basophilic cores in the context of a mild necrotizing myopathy. Thus, it is possible that this case represents an instance of early and/or mild autophagic vacuolar myopathy that was missed on electron microscopy. (Ultrastructural examination has high specificity but low sensitivity, resulting in sampling bias and significant possibility of false negative results.) When this case is excluded from the ROC analysis, the optimal (100% sensitivity and 100% specificity) threshold value for diagnosis of autophagic vacuolar myopathy is 8% positive fibers on either LC3 or p62 immunohistochemistry.
In rodent models of chloroquine toxicity, concurrent denervation greatly contributes to the development of vacuolar pathology 
. Similarly, basophilic core-like structures were not present in the autophagy deficient (ATG7−/−) murine muscle at baseline, developing only after muscle denervation 
. These findings raise the possibility that a concurrent neurogenic process contributes to the development of autophagic vacuolar myopathy in patients treated with chloroquine, hydroxychloroquine, or colchicine. Our data do not support this possibility: the drug-treated control group included two specimens (#15 and #16) with well-developed neurogenic changes, both of which showed a very low degree of LC3 and p62 labeling. However, a separate study will be required to definitively answer this question.
Rimmed vacuoles were detected in a subset of autophagic myopathy specimens both by standard histology () and on LC3 or p62 immunohistochemistry (). However, rimmed vacuoles are not specific for autophagic vacuolar myopathies and can be seen in other disorders of skeletal muscle such as inclusion body myositis and several subtypes of muscular dystrophy. Indeed, positivity for either LC3 or p62 has already been noted in inclusion body myositis 
, although careful quantification and determination of proper diagnostic thresholds still needs to be done. Similarly, inherited autophagic vacuolar myopathies (such as Danon disease or XMEA) would be expected to be highly LC3 and p62 positive, but were not included in the current study because our archive does not include a sufficient number of well documented cases. While diagnostically helpful, neither LC3 nor p62 positivity is therefore pathognomonic for drug-induced autophagic vacuolar myopathies. To establish the correct diagnosis, positive LC3 and/or p62 staining needs to be correlated with the remainder of histologic findings and with full clinical history (including age at presentation, family history, and medication history).
The current study was not designed to detect differences in age, sex, or drug treatment distribution between the two drug-treated study groups. Nonetheless, we found that male, colchicine-treated subjects were more common in the autophagic myopathy group, while female, hydroxychloroquine-treated subjects were more common in the drug-treated control group. The differences we observed were not statistically significant; however, a study with 95% power to detect the effects of magnitude we observed (with a significance level of 0.05) would have needed to have 40 subjects in each group. Therefore, a larger study – ideally with a prospective design – will need to be performed in the future to establish whether certain types of patients are more vulnerable to development of drug-induced autophagic vacuolar myopathy.
In summary, we used specimens from human subjects treated with either hydroxychloroquine or colchicine to establish that immunohistochemical stains for LC3 and/or p62 are useful ancillary tools in pathologic diagnosis of drug-induced autophagic vacuolar myopathies. By limiting the need for electron microscopy, use of LC3 and/or p62 immunohistochemistry will decrease both the time required to establish the diagnosis and the false negative rate resulting from sampling bias, thus resulting in significantly improved clinical care.
The major strengths of the current study are (1) the use of pathologically well characterized subjects, (2) the inclusion of two control groups (normal controls and drug-treated controls), (3) the inclusion of drug-treated control subjects with abnormal muscle, mimicking clinically relevant scenarios for diagnostic test use, (4) the concordance of results across several experimental modalities (immunostaining of FFPE tissue vs. immunostaining of frozen tissue vs. immunoblotting of frozen tissue), and (5) the quantitative study design, which enabled calculation of sensitivity and specificity values for different diagnostic thresholds. The major limitations are (1) the under-sampling of the drug-treated control group, as patients with no symptoms are unlikely to undergo a muscle biopsy, and (2) a non-negligible probability that one (or more) drug-treated subjects were miss-assigned to the control group given the significant false negative rate of electron microscopy (the current “gold standard" method for diagnosis of autophagic vacuolar myopathies). As discussed above, these limitations would be expected to result in an elevated diagnostic threshold and artificial widening of the diagnostic “gray zone". Given that alternative treatments exist for both rheumatologic diseases and gout, it would thus be reasonable to use a fairly low diagnostic threshold (5% of LC3- or p62-positive fibers) for recommendation to discontinue hydroxychloroquine or colchicine therapy in a symptomatic patient with otherwise equivocal pathologic findings.