A stepwise data interpretation process for renal amyloidosis typing by LMD-MS

Systemic amyloidosis is a group of heterogeneous diseases caused by protein structural abnormality. The pathological character of amyloidosis is the formation of extracellular deposition of beta-sheet fibril through aggregation of insoluble proteins or peptides. The cytotoxicity of the deposited proteins cause destruction of tissue and cellular structure, which then induce functional injury of organs, such as kidney [1‐3]. Systemic amyloidosis is classified according to the type of deposited protein, these mainly include immunoglobulin light/heavy chain, and AFPs other than immunoglobulin [3‐8]. The major affected organs, prognosis and effective treatment strategies of each amyloidosis type are different [9‐12], therefore, precise typing is of paramount importance for providing patients with the most appropriate care.

For diagnosis of amyloidosis, histopathologic examination methods are mostly used in routine practice, such as histochemical stain by Congo red (CR) [13]. Further typing methods are usually based on immunohistochemistry and immunofluorescence, which rely on antibody recognition that might be interfered by background and epitope loss, and could not identify new AFPs [14]. In 2008, Mayo Clinic proposed the LMD-MS technique for amyloidosis typing [15], and it gradually became the gold standard in recent years [3]. MS-based assay possesses cost advantage for it is a test once for all proteins, whereas antibody-based immunoassay is a test once for only one protein [16, 17].

However, there is still a gap between the LMD-MS technology itself and its transition to clinical application, that is largely due to the uncertainty of how to interpret the MS data. As proposed by Mayo Clinic, the type of amyloidosis is called by considering the most abundant amyloid protein that has the maximal MS/MS spectral count (SC) (Mayo’s rule) [18‐20]. Nevertheless, in routine practical detection, multiple AFPs are often identified at the same time. The pathogenic protein may not have an absolute preponderance of SC when compared with other AFPs in background for: 1) the deposition degree of the pathogenic amyloid protein is not much significantly superior to others, or 2) blood contamination happens, or 3) little amount of material sampled from microdissection. This makes it challenging to make a precise typing by simply extracting the protein with the highest SC. For instance, IgG is often identified with very high SC in immunoglobulin light chain amyloidosis (AL) [21]. In Leukocyte chemotactic factor-2 amyloidosis (ALECT2), empirical data showed that the SC of LECT2 is generally low, and is almost impossible to be the highest [22]. Furthermore, this relative quantification of protein SCs within a specimen cannot be extended to comparing different samples. Thus, it may need a global normalization method for establishing a relatively unified quantitative analysis system.

The prevalence of different amyloidosis types varies greatly in different regions, however, AL is always the most frequent one (59%-68%); serum amyloid A amyloidosis (AA, 4–12%), ALECT2 (~ 3%) and transthyretin amyloidosis (ATTR, 3%-33%) could be considered as less frequent types; whereas the other types are rare [1, 23, 24]. In AL, κ-type (ALκ) and λ-type (ALλ) interfere with each other, mainly because of the high background of κ chain. In terms of typing fibrinogen A α chain amyloidosis (AFib), the SC of FIBA should be greater than the sum of fibrinogen B β chain (FIBB) and fibrinogen G γ chain (FIBG) in addition to having the highest SC among all AFPs [25]. Besides, special considerations are required for making definite diagnosis of immunoglobulin heavy chain amyloidosis (AH) and its involvement with AL, that is immunoglobulin heavy-light chain amyloidosis (AHL) [26]. Thus, it is reasonable to propose a stepwise process to include or exclude specific amyloidosis types. In this study, we aimed to establish an LMD-MS based amyloidosis typing procedure, especially for the rules of MS data interpretation.

Another important aspect of LMD-MS based amyloidosis typing is quantifying accompanying proteins (AAPs) that are co-deposited with amyloid fibrils. As far as we know, three proteins named apolipoprotein E (APOE), apolipoprotein A-IV (APOA4), and serum amyloid P component (SAP) are usually accompanied with amyloid deposition in amyloidosis. Although the LMD-MS analysis is usually executed based on histopathologic diagnosis, that is usually a positive staining of CR, the three AAPs are often regarded as the evidence of amyloid deposition. Even, the criterion of detection of at least two of the three AAPs has been surrogate for CR staining [27]. However, whether these three AAPs’ quantitative information is valuable for improving amyloidosis diagnosis needs further investigation.

Currently, the LMD-MS based amyloidosis typing is increasingly accepted as a clinical test. However, how to reading the MS data properly to make the right diagnosis is still a challenging for clinicians. In this study, we adopted a stepwise data interpretation procedure that includes or excludes the types group by group. 1) Other types besides immunoglobulin related amyloidosis were firstly confirmed or excluded by analyzing the S-scores; 2) followed by considering whether and which immunoglobulin light chain might be involved by analyzing λ-κ; and 3) whether and which immunoglobulin heavy chain participated in amyloid deposition was judged by analyzing the H-scores at last. This procedure was verified to have a total accuracy rate of 88% in validation phase. ALκ and ALλ are the main types, achieving accuracy of 92% and 88% respectively. The 100% accuracy in the less common types of ALECT2, AGel, AA and ATTR was achieved. But for immunoglobulin heavy chain related types, it varied from 67%-80%, that could be the most difficult systemic amyloidosis to type.

Of the 7 ALECT2 cases in this study, the SC of LECT2 was relatively low (absolute SC ranged from 3–16 and normalized SC ranged from 1.01–4.22), and were always lower than Igκ and IgG. This was similar to the results reported by Li et al. [22] and Mereuta et al. [29]. These results implied that the abundance of LECT2 may be underestimated by shotgun proteomics quantification. In addition, LECT2 was not identified in all other amyloidosis other than ALECT2 type and non-amyloid nephropathies. Therefore, ALECT2 can be determined by the identification of LECT2.

The major type of amyloidosis is AL, which is further subdivided into ALκ and ALλ. The background of Igκ was relatively high, especially in some ALλ cases, it was even slightly higher than Igλ. Here, we innovatively proposed the variable of λ-κ to distinguish the interference between Igκ and Igλ, and found that non-κ/λ related categories can be effectively identified by this variable (except for AFib and AApoAI, which could be recognized before λ-κ analysis). In validation dataset, there was only one AL (Igκ) case that had the λ-κ value (-4.06) a little higher than cutt-off (-5.45) and was not recognized immediately. Yet, none of the other types were assigned to this case. Other than that, the λ-κ analysis is precise for distinguishing κ, λ and non-κ/λ related cases.

Amyloid deposition of immunoglobulin heavy chain is usually accompanied with light chain. This brings a major challenge to distinguish AH and AHL from each other. Moreover, in some AL cases, the SC of IgG may have a higher value over light chains, which causes the confusion of AH, AHL and AL. Therefore, we introduced another score (H-score) to recognize the involvement of immunoglobulin heavy chain, especially for IgG. Of the total 51 AL specimens, 5 (9.8%) had higher IgG SC than light chains, which may be easily misclassified to IgG related types. This phenomenon was not observed for IgA and IgM. This may be due to the high background of IgG in the blood that causes interference, and thus the analysis for IgG should be different from IgA & IgM. Consequently, we found that the H-score cut-off for IgG was higher than IgA & IgM. Despite the utilization of H-score, several mistakes were generated (Fig. 3 and Table 3). Among which, the SC of IgG was overrepresented in an AL (Igλ) case and was underrepresented in an AHL (IgGλ) case, as well as in an AHL (IgAλ) case, IgG and IgA both exceeded their cut-off, and in an AH (IgG) case, contradiction happened between IgG and IgM. These mistakes were all associated with IgG, requiring the typing method to be further improved in this respect. Also, these cut-off values may need to be optimized by more clinical specimens in future study. For now, it might be better to take a look at histopathological or hematological test results and rely on experienced clinicopathologists to determine whether and which heavy chain is involved. Nevertheless, these mistakes have little influence on the formulation of clinical treatment plan.

Qualitative analysis of AAPs is not an appropriate method to diagnose amyloidosis, since it could not exclude other renal diseases. This may be mainly because of the development of MS technology, which becomes more sensitive and can identify more proteins at lower abundance. Here, we proposed the quantitative analysis way that use the average abundance of the three AAPs, and indicated it as a fine indicator to diagnose amyloidosis.

For AL amyloidosis, it is desirable to assess the clonal disorder in serum and urine, and for hereditary amyloidosis, genotyping analysis is pertinent to confirm the typing result. Nevertheless, we failed to collect enough data in these respects, which are limitations of the present study, however, relevant analyses should be addressed when possible. There were still several rare types such as ACys, Aβ2M, AApoAII, AApoCII and AApoCIII that were not included in this study. This was mainly because of the low incidence of systemic amyloidosis (10 per million) and even lower for rare types. The data interpretation process will be further verified and optimized in future clinical tests and follow-up studies, as well as the MS data characteristics of these not included types are to be determined more detailly in the future.

This study demonstrates the value of LMD-MS technique for precise diagnosis of amyloidosis. The principle and details of the stepwise process of MS data analysis that can improve typing accuracy was proposed for the first time, where some cut-off values are defined and set. This scheme includes not only the diagnosis of AL, but also relatively rare subtypes, which has a high clinical practical value. The development of this method will be helpful for clinicians to accurately typing amyloidosis. In addition, we discovered that the quantitative analysis of the AAPs could be a fine indicator to identify amyloidosis from other nephropathies, which is valuable to be confirmed and developed as diagnostic indicator.