Introduction

Papillary thyroid carcinoma (PTC) tall cell variant (PTC-TCV) is composed of cells that are ‘tall’, i.e. they are 2–3 times taller than wide. Additional features include complex papillary formation with trabecular architecture (‘tram track’ pattern), older patient age, and the common occurrence of BRAF V600E mutations. Starting with its first description by Hawk and Hazard in 1976 [1], numerous studies repeatedly found PTC-TCV to be associated with aggressive clinical features and reduced patient survival [2-9], and it has long been debated whether the tall cell features represent an independent prognostic factor for PTC [10-18]. Current American Thyroid Association (ATA) and European Society for Medical Oncology (ESMO) guidelines regard PTC-TCV as a variant with ‘aggressive histology’ [19, 20]. Accordingly, patients with PTC-TCV are considered at least intermediate-risk and are managed more aggressively in terms of both surgery and adjuvant radioactive iodine (RAI) administration. Thus, it is important to correctly recognize PTC-TCV but, unfortunately, its diagnosis is rather inconsistent among pathologists (even at the expert level) [21]. Causes of the discrepancies are the subjective nature of the interpretation of morphologic criteria, their frequently nonuniform expression within the tumor, the fact that tall cell clusters are frequently present in otherwise classic papillary carcinomas, and the lack of immunohistochemical or molecular markers for PTC-TCV [12, 18, 21]. The tall cells of PTC-TCV have abundant ‘oncocytoid’ eosinophilic cytoplasm. While it is generally recognized that they are rich in mitochondria [22, 23], very few electron microscopy studies have linked the abundance of mitochondria to PTC-TCV [24]. Homoplasmic/highly heteroplasmic somatic mitochondrial DNA (mtDNA) mutations that are pathogenic for the encoded molecules, and that often involve complex I subunits of the oxidative phosphorylation (OXPHOS) system, cause OXPHOS impairment leading to the accumulation of mitochondria, which defines oncocytic cells in the thyroid [25-27] as well as in other organs, including the kidney, pituitary, salivary, and parathyroid glands [28-32]. While these homoplasmic/highly heteroplasmic mutations occur in oncocytic tumors, they are also present in some papillary carcinomas [26], including those analyzed in the 2014 TCGA study [33, 34].

Given this context, the aim of this study is to clarify the relationship between PTC-TCV and oncocytic thyroid carcinoma. For this purpose, we collected a series of PTCs with and without tall cell features, reviewed their histologic and clinicopathologic features, performed electron microscopy, subjected all samples to next-generation sequencing (NGS) to analyze the entire mitochondrial genome of all cases and to identify the status of those nuclear genes that are frequently mutated in thyroid carcinoma, and assessed OXPHOS complex I integrity by immunohistochemistry (IHC).

Materials and methods Case selection

Thirty-three representative PTCs – with and without tall cell features – that belonged to 31 patients were randomly selected from the archival material of the Pathology Unit, Odrensklinikum Linz, Linz, Austria. Hematoxylin and eosin (H&E)-stained slides were reviewed by OT, GT, and MS-S and allocated to test (PTC-TCV, 17 cases) and control PTC cases (cPTC, 16 cases) following current histopathologic criteria [22] (see Supplementary materials and methods). In four cases, both primary and metastatic tissue was included; thus, a total of 37 samples were analyzed from 31 patients. The study complied with the ethics principles of the Declaration of Helsinki and followed Institutional Review Board approved protocols in Bologna, Italy, and Linz, Austria.

DNA extraction and mutation analysis of nuclear and mtDNA

Whole DNA was extracted from formalin-fixed paraffin-embedded tissue. NGS was performed with the MiSeq platform (Illumina, Inc., San Diego, CA, USA). To identify nuclear DNA mutational hot-spots for thyroid cancer genes, we used a custom-designed NGS multigene panel [35]. The entire mtDNA was sequenced using NGS and the MitoAll re-sequencing kit (Applied Biosystems, Foster City, CA, USA), as previously described [36] (see Supplementary materials and methods for additional information).

mtDNA variant annotation

Frameshift and truncating mutations in the coding sequence dramatically alter the encoded protein and are therefore classified as pathogenic. For all other variants (synonymous, non-synonymous, noncoding intergenic variants, and variants mapping within rRNAs and tRNAs), the pathogenic potential was evaluated in silico with MToolBox [37], HmtVar [38], and an in-house pipeline [39] (see Supplementary materials and methods for additional information).

Electron microscopy and IHC

Transmission electron microscopy (TEM) was performed according to standard procedures. For IHC, prohibitin was used as a pan-mitochondrial marker [40], and NDUFS4 and cytochrome c oxidase 1 (COX-I) as markers of complex I and IV integrity, respectively [41, 42]. BRAF V600E mutation-specific antibody was used to assess the distribution of the mutated protein (see Supplementary materials and methods and Table S1).

Statistical analysis

Statistical analysis was performed using the SPSS Statistics software package, version 23.0 (IBM Corp., Armonk, NY, USA). Contingency tables with exact tests (either chi-square test with Monte Carlo permutation technique or Fisher's exact test) were calculated for discrete variables. Nonparametric methods (median values and Mann–Whitney test) were used for continuous variables. For multiple comparisons, P values were obtained uncorrected and with family-wise error rate correction (Holm–Bonferroni method [43]). Differences between PTC-TCV and PTC controls and those between BRAF V600E mutated and BRAF wild-type tumors were tested separately.

Results Clinicopathologic features

The characteristics of the patient cohort are reported in Table 1. Representative cases of one PTC-TCV and one cPTC are illustrated in Figure 1A,B. In all 17 PTC-TCV, the majority (>60%) of the tumors showed tall cell features, i.e. cell height greater than 2–3 times the cell base and long non-branching papillae resulting in trabecular architecture (‘tram track’ pattern) on histology sections: tall cell features represented 90% or more of the tumor in 10 of 17 PTC-TCV (supplementary material, Table S2). Fourteen cPTC cases were of classic papillary morphology. Two cPTC were unencapsulated infiltrative follicular variant PTC (Table 1).

Table 1. Clinicopathologic features of the cases analyzed. Parameter All cases (n = 33*) PTC-TCV (n = 17) PTC control group (n = 16†) Statistical significance of the difference, p Patients' age at presentation, median (range), years 45.7 (16–81) 53.0 (28–81) 37.5 (16–66) 0.034 Patients' sex, female:male ratio 3.7:1 4.7:1 3.0:1 0.69 Median tumor size, maximal diameter, cm 2.5 (0.2–7.0) 2.9 (1.2–6.5) 1.8 (0.2–7.0) 0.14 Extrathyroidal tumor extension, n (%) No 16 (48%) 7 (41%) 9 (56%) Microscopic 15 (45%) 8 (47%) 7 (44%) Gross 2 (6%) 2 (12%) 0 0.49 Positive resection margins, n (%) Microscopic 6 (18%) 3 (18%) 3 (19%) Gross 1 (3%) 1 (6%) 0 0.62 LN metastases, n (%) 20/31 (65%) 10/16 (63%) 10/15 (67%) 1 Distant metastases at presentation, n (%) 1 (3%) 0 1 (6%) 0.49 Tumor stage (AJCC/UICC Eighth ed. 2017), n (%) Stage I 26 (79%) 12 (71%) 14 (88%) Stage II 7 (21%) 5 (29%) 2 (12%) 0.4 Thyroid remnant RAI ablation, n (%) 31 (100%) 17 (100%) 16 (100%) 1.0 Disease status at 1 year after surgery, treatment response (ATA guidelines) Excellent 26 (79%) 14 (82%) 12 (75%) Biochemical incomplete 4 (12%) 2 (12%) 2 (13%) Structural incomplete 1 (3%) 0 1 (6%) NA 2 (6%) 1 (6%) 1 (6%) 0.69 Median follow-up duration (range), months 20 (4–144) 14 (4–144) 28 (8–144) 0.37 Disease status at last follow-up, treatment response (ATA guidelines) Excellent 25 (76%) 14 (82%) 13 (81%) Biochemical incomplete 5 (15%) 1 (6%) 0 Structural incomplete 1 (3%) 1 (6%) 2 (13%) NA 2 (6%) 1 (6%) 1 (6%) 0.86 Bold font indicates P values that are statistically significant. AJCC, American Joint Committee on Cancer; ATA, American Thyroid Association [20]; LN, lymph node; NA: not available; UICC, Union for International Cancer Control. * Two patients had two different tumors: both had one PTC-TCV and one PTC, classic type in the same thyroid gland, and both tumors were analyzed separately. † The PTC control group included 14 classic PTCs and 2 infiltrative follicular variant PTCs.

Histologic appearance and immunohistochemical features of PTC-TCV and cPTC. PTC-TCV (A) and classic papillary carcinoma cPTC (B), H&E staining. IHC for the pan-mitochondrial marker prohibitin shows high expression levels with strong, homogenous granular staining in PTC-TCV (C), but low levels of granular staining in cPTC (D). Expression of complex I NDUFS4 subunit is lost in the tumor cells of PTC-TCV, whereas it is preserved in endothelial cells that act as internal positive control (E); NDUFS4 expression is preserved in cPTC (F). The BRAF V600E mutated protein is expressed in the majority of tumor cells (IHC; BRAF V600E specific antibody, clone VE1) (G), while NDUFS4 loss is restricted to the tall cell subpopulation of the tumor (H, arrows), consistent with the hypothesis that papillary carcinomas first acquire BRAF V600E and then the mtDNA alterations that cause the tall cell phenotype (case T14).

Electron microscopy shows an accumulation of mitochondria in papillary carcinoma TCV

TEM was performed on randomly selected PTC-TCV (n = 4) and cPTC (n = 4) for which fresh tumor tissue was available for TEM processing. PTC-TCV showed massive accumulation of mitochondria, many of which were abnormally large and swollen (Figure 2). In spite of some variability, all neoplastic cells showed large sections of the cytoplasm replaced by compact clusters of mitochondria (see also Figure 1C). The cytoplasm of cPTC was largely occupied by endoplasmic reticulum and vacuoles. Mitochondria were sparse and morphologically normal (Figure 3).

Cell organelle ultrastructure in PTC-TCV. Upper part: overview showing the massive accumulation of mitochondria that almost replace the entire cytoplasm of PTC-TCV cells (case T3). Lower left: same picture as above with mitochondrial mapping (yellow overlay) in two adjacent tumor cells to highlight quantity and distribution of mitochondria. Lower right: higher magnification of the inset in the upper picture demonstrates closely packed mitochondria that are enlarged and swollen.

Cell organelle ultrastructure in classic PTC. Upper part: overview showing endoplasmic reticulum and vacuoles occupying most of the cytoplasm (case C4). Lower left: same image as above with mitochondrial mapping (yellow overlay) in three adjacent tumor cells to highlight the quantity and distribution of mitochondria. Lower right: higher magnification of the inset showing only a few morphologically normal mitochondria scattered among dilated cisternae of the endoplasmic reticulum.

BRAF V600E is the nuclear DNA mutation signature of papillary carcinoma TCV

Nuclear DNA mutations are summarized in Figure 4 and listed in supplementary material, Table S2. BRAF V600E was identified in all but one of the PTC-TCV using both NGS and IHC with BRAF V600E specific antibodies. In four cases, both primary tumor and lymph node metastases were analyzed (two PTC-TCV and two cPTC), and in all of them the results were concordant (Figure 4 and supplementary material, Table S2). The high BRAF V600E mutated allele frequencies adjusted to the proportion of neoplastic cells in the samples (supplementary material, Table S2) were consistent with the immunohistochemical findings: BRAF V600E was expressed in the vast majority of neoplastic cells within the tumor. The correlation of BRAF V600E with PTC-TCV, cPTC, and mitochondrial alterations (mtDNA mutation and loss of complex I integrity) is reported in Tables 2 and 3. BRAF V600E was statistically associated with PTC-TCV (p = 0.002) (Table 2) and with the mitochondrial alterations of PTC-TCV (Tables 2 and 3 and the following paragraphs).

‘Heat map’ representing the distribution of the assessed parameters among test (PTC-TCV) and control (classic PTC) cases. C, control papillary carcinoma cases (classic PTC, n = 14; infiltrative variant PTC, n = 2); LN, lymph node metastasis; P, primary tumor; T, test cases. aIHC. Prohibitin expression: 0, low; 1, intermediate; 2, high; NDUFS4 expression: 0, lost; 1, partially lost; 2, preserved. bRelationship between mtDNA alteration type and PTC-TCV histotype: 0, no mtDNA alterations and no PTC-TCV; 1, mtDNA alterations in genes encoding complex I subunits and PTC-TCV; 2, mtDNA alterations in genes not encoding complex I subunits and PTC-TCV; 3, mtDNA alterations (regardless of the mtDNA gene affected), but no PTC-TCV. Mutations in mtDNA genes and in nuclear genes: 0, absent; 1, present.

Table 2. BRAF V600E mutation, mtDNA mutation, and complex I integrity in PTC-TCV and control papillary carcinomas. Parameter PTC-TCV (n = 17) PTC control group (n = 16) Statistical significance of the difference, p Uncorr. FWER-corr. BRAF V600E, n (%) 16 (94%) 7 (44%) 0.002 0.002 mtDNA mutation, n (%) 17 (100%) 3 (19%) <0.00001 0.00003 Mitochondrial quantity (IHC, prohibitin stain), n (%) Low 0 12 (75%) Intermediate 0 2 (12.5%) High 17 (100%) 2 (12.5%) <0.00001 0.00003 Complex I immunoreactivity (IHC, NDUFS4 stain), n (%) Completely lost 13 (76%) 0 Partially lost 4 (24%) 3 (19%) Preserved 0 13 (81%) <0.00001 0.00003 Bold font indicates P values that are statistically significant. FWER-corr., family-wise error rate correction; Uncorr.: uncorrected. Table 3. mtDNA mutation and complex I integrity in PTC with and without BRAF V600E mutation. Parameter BRAF status Statistical significance of the difference, p WT (n = 10) V600E (n = 23) Uncorr. FWER-Corr. mtDNA mutation, n (%) 3 (30%) 17 (74%) 0.026 0.026 Mitochondrial quantity (IHC, prohibitin stain), n (%) Low 8 (80%) 4 (17%) Intermediate 1 (10%) 1 (4%) High 1 (10%) 18 (78%) 0.001 0.002 Complex I immunoreactivity (IHC, NDUFS4 stain), n (%) Completely lost 0 13 (57%) Partially lost 1 (10%) 6 (26%) Preserved 9 (90%) 4 (17%) 0.0004 0.0012 Bold font indicates P values that are statistically significant. FWER-corr., family-wise error rate correction; Uncorr., uncorrected; WT, wild type. Homoplasmic/highly heteroplasmic mtDNA mutations are a defining feature of papillary carcinoma TCV

The results of mtDNA analysis are summarized in Figure 4 and listed in supplementary material, Table S2. The entire mtDNA sequence was obtained in all 37 samples, of which 33 were primary tumors (17 PTC-TCV and 16 cPTC) and 4 lymph node metastases (2 from PTC-TCV and 2 from cPTC) (supplementary material, Table S2). To determine the levels of heteroplasmy, mutated allele frequencies were adjusted to the proportion of tall cells in the neoplastic area marked for analysis (supplementary material, Table S2). The results of mtDNA analysis in the four cases where both the primary tumor and lymph node metastases were analyzed (two from PTC-TCV and two from cPTC) were concordant for each sample pair (Figure 4 and supplementary material, Table S2). All 17 PTC-TCV harbored a total of 21 pathogenic mtDNA mutations, with single tumors carrying a maximum of 2 different mutations. All were at homoplasmic/highly heteroplasmic levels (supplementary material, Table S2). Sixteen of the 21 mutations (76%) identified in PTC-TCV mapped within mtDNA-encoded complex I subunits (Figure 5). Nineteen of the 21 mutations were severely pathogenic for the encoded molecule. The remaining two scored ‘likely polymorphic’ in silico by the bioinformatics tools used for mtDNA variant annotation but were associated with loss of complex I integrity (see mtDNA alterations correlate with loss of OXPHOS complex I integrity in papillary carcinoma TCV, below) (Figure 4 and supplementary material, Table S2). No mtDNA alterations were found in 13 of 16 (81%) cPTC. In each of the three remaining cPTCs, there was one mtDNA mutation per sample, pathogenic in silico for the encoded molecules. In two cases, there were homoplasmic mtDNA-encoded complex I subunit mutations (m.3389T>C/MT-ND1, case C2; m.10371G>A/MT-ND3, case C10); both tumors harbored small tall cell subpopulations. One mutation with very low heteroplasmy (m.9654A>G) affecting the MT-CO3 gene encoding a complex IV subunit was found in the third PTC (case C11); this last case did not have any tall cell subpopulation.

Distribution of mtDNA mutations in papillary carcinoma TCV. Distribution according to the mtDNA-encoded gene products (left) and according to OXPHOS mitochondrial complexes (right). Number of mtDNA mutations and percentage of the total number of mtDNA mutations identified (%). C, OXPHOS complex; COX, complex IV subunits encoded by MT-CO genes; ND, complex I subunits encoded by MT-ND genes; tRNALeu, mitochondrial transfer RNA for leucine encoded by the MT-TL1 gene.

Papillary carcinoma TCV shows loss of OXPHOS complex I integrity by IHC

The structural stability of the respiratory complex I was evaluated by comparing the relative immunohistochemical expression of the prohibitin pan-mitochondrial marker with the nuclear DNA-encoded NDUFS4 complex I subunit. Results are illustrated in Figure 1C–F and reported in Table 2, Figure 4, and supplementary material, Table S2. PTC-TCV showed extensive granular prohibitin expression consistent with large numbers of mitochondria in neoplastic cells (Table 2 and Figure 1C; see also TEM, Figure 2). PTC control cases showed low-level granular prohibitin expression consistent with a limited number of mitochondria in neoplastic cells (Table 2 and Figure 1D; see also TEM, Figure 3). PTC-TCV showed immunohistochemical loss of NDUFS4 complex I subunit consistent with lack of complex I integrity (Table 2 and Figure 1E). Preserved NDUFS4 complex I subunit expression is consistent with proper assembly of complex I in cPTC (Table 2 and Figure 1F). Immunohistochemical results of the four cases where both the primary tumor and lymph node metastases were analyzed were concordant for each sample pair: in two PTC-TCV cases, IHC was consistent with loss of complex I integrity, in two cPTC it was consistent with proper assembly of complex I (Figure 4 and supplementary material, Table S2). Partial NDUFS4 complex I subunit loss was observed in 4 of 17 (24%) PTC-TCV and in 3 of 16 (19%) cPTC (Table 2, Figure 4, and supplementary material, Table S2). Loss (partial or complete) of complex I integrity evaluated immunohistochemically was strongly associated with PTC-TCV: it occurred in all 17 PTC-TCVs, while only partial NDUFS4 complex I subunit loss was found in 3 of 16 cPTCs (p < 0.0001) (Table 2, Figure 4, and supplementary material, Table S2).

mtDNA alterations correlate with loss of OXPHOS complex I integrity in papillary carcinoma TCV

Comparison of mtDNA mutations with loss of complex I assembly evaluated immunohistochemically in PTC-TCV and cPTC is reported in Table 2, Figure 4, and supplementary material, Table S2. mtDNA alterations were statistically associated with lack of complex I assembly: they were identified in 18 of 20 cases with loss (partial or complete) of complex I integrity; conversely, lack of mtDNA alterations and proper assembly of complex I occurred in 11 of 13 cases (p < 0.0001) (Figure 4 and supplementary material, Table S2). Among the 33 PTC cases analyzed, 9 tumors carrying nearly homoplasmic mtDNA mutations pathogenic for the mtDNA-encoded gene product in the MT-ND genes encoding complex I subunits showed complete loss of complex I integrity: all tumors were PTC-TCV. This is fully consistent with in silico data showing the disassembling potential of mtDNA mutations in PTC-TCV (Figure 4 and supplementary material, Table S2). Of the remaining eight PTC-TCV, homoplasmic/highly heteroplasmic mtDNA mutations in the MT-ND genes, pathogenic for the mtDNA-encoded complex I subunits, were associated with partial loss of complex I integrity restricted by IHC to a variable proportion of the population of tall cells within the tumor in four cases (Figure 4 and supplementary material, Table S2). In one additional PTC-TCV with MT-ND1 mutation scored ‘likely polymorphic’ in silico by the bioinformatics tools (ND1 mutation-‘likely polymorphic’, case T2), complete NDUFS4 immunohistochemical loss in the tumor cells was consistent with in vivo lack of complex I integrity (Figure 4 and supplementary material, Table S2). Three further PTC-TCVs showed mtDNA mutation in genes other than the MT-ND genes encoding for complex I subunits as single alterations (cases T1, T8, and T10) (Figure 4 and supplementary material, Table S2): one case (T10) carried the pathogenic m.14864G>A affecting MT-CYB which encodes for cytochrome b (Cyt b), a subunit of complex III (CIII); the second case (T1) carried the pathogenic m.3244G>A in MT-TL1 (the mitochondrially encoded gene for the tRNALeu transfer RNA); and the third case (T8) also carried the m.3239G>A in MT-TL1 scored in silico as ‘likely polymorphic’. The apparent discordance between mtDNA mutation in genes other than the MT-ND genes and the complete loss of complex I integrity (immunohistochemical absence of NDUFS4 in tumor cells) in the PTC-TCVs with altered MT-CYB can be explained by a combined CI/CIII deficiency induced by the mutation, especially considering how critical Cyt b, encoded by MT-CYB, is for CIII assembly [44, 45]. In the two PTC-TCVs with altered MT-TL1, there was reduced ability of the tumor to synthesize mitochondrial proteins due to mutation of tRNALeu, as confirmed by the lack of expression of the mtDNA-encoded complex IV COX-I subunit demonstrated by IHC (data not shown). Alterations of mtDNA were found in three cPTCs (Figure 4 and supplementary material, Table S2). In one case (C11), there was a mutation (m.9654A>G) affecting the MT-CO3 gene encoding a complex IV subunit which is not expected to affect complex I integrity, as confirmed by preserved immunohistochemical NDUFS4 reactivity. In two additional cPTCs, there were homoplasmic mtDNA-encoded complex I subunit mutations (m.3389T>C/MT-ND1, case C2; m.10371G>A/MT-ND3, case C10) pathogenic in silico for the encoded molecule. In the case with m.3389T>C/MT-ND1 (case C2), there was partial loss of NDUFS4, but the preservation of NDUFS4 by IHC in most of the tumor cells of both cases was more in line with proper assembly of complex I in vivo. Interestingly, a small proportion of neoplastic cells with tall cell morphology (5–10% of the sample analyzed) was present in both tumors.