Gastric carcinoma with lymphoid stroma (GCLS) is a rare histological variant, accounting for 1–4% of all gastric carcinomas [1–4]. Since Watanabe et al. first described “gastric carcinoma with lymphoid stroma” in 1976 [5], GCLS has also been reported as lymphoepithelioma-like carcinoma (LELC) or medullary carcinoma [6–9]. Histologically, GCLS is characterized by poorly developed tubular structures associated with prominent lymphoid infiltrate in a non-desmoplastic stroma [1]. However, standardized diagnostic criteria are lacking and the histological diagnosis for GCLS is obscure. As Crohn’s disease-like lymphocytic reaction (CLR), which shows frequent tubule or gland formation, demonstrates similar features to typical GCLS, it has been suggested as an expanded spectrum of GCLS [10].

GCLS consists of two subsets: Epstein–Barr virus (EBV)-positive and microsatellite instability (MSI)-high carcinomas [11]. Over 80% of GCLSs are associated with EBV infection as compared to 8.7% (range, 1.3–20.1%) of all gastric cancers [12, 13]. The prevalence of MSI-high carcinoma in GCLS ranges 7–39% [11].

GCLS has distinct clinicopathologic characteristics and confers a favorable prognosis with a low lymph node metastasis (LNM) rate [10–19]. Increased tumor-infiltrating lymphocytes, which reflect a host immune response to tumor cells, are associated with improved survival [10, 11, 20–22].

Although recent studies analyzed clinicopathologic features and survival outcomes of GCLS or EBV-associated gastric carcinoma (EBVaGC) [18, 19], current information on early GCLS (EGCLS) is not available [5, 23, 24]. As EGCLS has a low frequency of LNM despite deep submucosal invasion [5], some reports have described an expanded role for endoscopic submucosal dissection (ESD) in EGCLS [25–29]. We therefore aimed to investigate the clinicopathologic characteristics of EGCLS, and to define the feasibility of ESD for EGCLS.

Materials and methods

Study population

Between January 2007 and December 2014, 1702 patients with early gastric cancer (EGC) underwent gastrectomy with lymph node dissection at Pusan National University Hospital, Busan, Republic of Korea. After excluding 17 patients who had incomplete pathologic data, we retrospectively reviewed 59 EGCLS patients and 1626 patients with early non-GCLS (ENGCLS). Furthermore, data from 11 EGCLS patients who underwent ESD without surgical resection were analyzed. This study was approved by the Institutional Review Board of Pusan National University Hospital (2015137).

Treatment methods and follow-up strategy

In surgical patients, gastrectomy with D1+ or D2 lymph node dissection was routinely performed, and ESD was considered for patients with EGC according to the expanded criteria proposed by Gotoda et al. [30]. Patients who were beyond the expanded criteria after ESD were recommended to undergo additional surgery.

Follow-up endoscopy and abdominal computed tomography were performed every 3–6 months for 1–3 years based on the clinician’s decision, and then annually checked until 5 years after the treatment. After 5-years follow-up, annual endoscopic surveillance was done.

Pathologic definition

All histologic slides of patients with EGCLS were reviewed by two gastrointestinal pathologists (P.D.Y. and A.S.J.), including typical GCLS and CLR as subclassifications of GCLS [10]. Typical GCLS was determined by the following: (1) a well-defined tumor margin; (2) dense lymphocytic infiltration where there were more tumor-infiltrating lymphocytes than tumor cells; (3) indistinct cytoplasmic borders and a syncytial growth pattern with poorly formed glandular structures; and (4) no desmoplasia (Fig. 1A). CLR was defined by the following: (1) patchy lymphocytic infiltration with three or more lymphoid follicles with active germinal centers per tissue section at the advancing edge of the tumor; (2) fewer lymphocytes than tumor cells; (3) frequent tubule or gland formation; and (4) minimal or no desmoplasia (Fig. 1B). ENGCLS included two categories: differentiated type-predominant adenocarcinoma (D-ENGCLS) and undifferentiated-type predominant adenocarcinoma (U-ENGCLS). Well-differentiated and moderately differentiated tubular or papillary adenocarcinomas were classified as differentiated; poorly differentiated adenocarcinomas or signet ring cell carcinomas were classified as undifferentiated.

Fig. 1
figure 1

Histologic findings of typical gastric carcinoma with lymphoid stroma (GCLS) and Crohn’s disease-like lymphocytic reaction (CLR) showing positive Epstein–Barr virus in situ hybridization (EBV-ISH). A Typical GCLS shows dense lymphocytic infiltration, accompanied by a syncytial growth pattern with poorly formed glandular structures (hematoxylin and eosin [H&E], 40×). B In CLR, tumor cells show frequent tubule formation, and there is less lymphoid stroma than tumor cells (H&E, 40×). C, D EBV-ISH highlights tumor cells (40×)

Full size image

Cancers were classified by invasion depth as mucosal or submucosal invasive cancers. Submucosal invasive cancers were subdivided into three equal layers of the same thickness: SM1, SM2, and SM3. In patients who underwent ESD, SM1 was defined as an invasion depth of <500 µm, SM2 500–1000 µm, and SM3 ≥1000 µm.

Clinicopathologic data

We obtained clinicopathologic data from 70 EGCLS patients (59 treated with surgical resection and 11 treated with ESD) and 1626 patients with ENGCLS via medical record review as follows: age, sex, a multiplicity of gastric cancer, macroscopic type (elevated; type I or IIa, flat; type IIb, depressed; type IIc or III), tumor location (upper third, middle third, or lower third of the stomach, or remnant stomach), Lauren classification, maximum tumor diameter, invasion depth (mucosa and submucosa: SM1, SM2, and SM3), and the presence of lymphovascular or perineural invasion. Additional information was obtained from 70 EGCLSs including endoscopic and endosonographic features, EBV positivity, and clinical outcome. MSI status was only reviewed in routinely investigated surgical specimens from 59 EGCLSs due to an inability to perform MSI analysis on paraffin-embedded ESD specimens. LNM status in the 1685 patients who underwent surgical resection (59 with EGCLS and 1626 with ENGCLS) was further investigated and compared according to tumor histology and depth.

Endoscopic and endosonographic findings

Endoscopic images of 70 EGCLS patients were reviewed by two endoscopic specialists (K.G.H. and L.B.E). Discoloration (red or white), ulceration, and submucosal tumor (SMT)-like presentation were investigated. SMT-like presentation was defined as the longest diameter of the exposed mucosa being less than one-third of the hidden tumor mass [31]. Endosonographic images of 34 EGCLSs were reviewed to evaluate the presence of marked mucosal hypoechogenicity or hypoechoic submucosal nodules (Fig. 2A–H).

Fig. 2
figure 2

Endosonographic images of early gastric carcinoma with lymphoid stroma. A, E Marked hypoechogenicity in the mucosal layer at the center of the tumor; mucosal cancer (confined to muscularis mucosa) on the gastric fundus. B, F A hypoechoic round submucosal nodule; SM3 cancer (2000 µm invasion from the muscularis mucosa) on the upper body. C, G Hypoechoic submucosal nodules; SM3 cancer (3125 µm invasion from the muscularis mucosa) on the lower body. D, H A hypoechoic submucosal mass; SM3 cancer (6500 µm invasion from the muscularis mucosa) on the prepylorus

Full size image

EBV chromogenic in situ hybridization

EBV chromogenic in situ hybridization was performed on paraffin-embedded samples from 70 EGCLSs using an automatic staining device (BOND-MAX), with a Novocastra EBER probe (Leica Biosystems, Newcastle Upon Tyne, UK) according to the manufacturer instructions. Samples with strong staining within the tumor cell nuclei were considered positive (Fig. 1C, D).

MSI analysis

We screened microsatellite status according to protocols recommended by the National Cancer Institute (NCI) guidelines [32]. Briefly, extracted DNA from tumor and normal tissues was amplified by fluorescent multiplex PCR targeting five microsatellite loci: BAT25, BAT26, D5S346, D2S123, and D17S250. DNA was sequenced via fragment analysis using a temperature-controlled DNA sequencer (PRISM 377, Perkin-Elmer Corp., Foster City, CA, USA) and GeneScan software (Perkin-Elmer Corp.). In accordance with NCI criteria [32], MSI-high was defined as instability in at least two microsatellite loci; MSI-low as instability in one locus; and microsatellite stable when none of the loci were shifted.

Statistical analysis

Continuous variables are expressed as means ± standard deviations or ranges; categorical variables are presented as frequencies with percentages. Student’s t-tests and ANOVA were used to compare continuous variables; Chi-square and Fisher’s exact tests were performed for categorical variables. Statistical analyses were performed using SPSS version 21.0 statistical software package (SPSS Inc., Chicago, IL, USA). P values of less than 0.05 were considered significant.

Results

Clinicopathologic features of EGCLS as compared to ENGCLS

Table 1 summarize the clinicopathologic features of EGCLS as compared to ENGCLS. The mean age of EGCLS patients was 58 years (range, 36–77 years), and a male predominance (3.7:1) was observed as compared to ENGCLS patients (P = 0.023). The EGCLS group showed higher tumor multiplicity (P = 0.014), more macroscopically elevated tumors (P = 0.003), predominant upper locations (P < 0.001), higher incidence in the remnant stomach (P = 0.023), and more diffuse-type histology by Lauren classification (P < 0.001) as compared to the ENGCLS group. Despite a smaller tumor size (P < 0.001), submucosal tumor invasion was more frequent (P < 0.001) and submucosal infiltration was deeper (P = 0.026) in the EGCLS group than in the ENGCLS group. No significant differences were observed in the rates of lymphovascular or perineural tumor invasion between the EGCLS and ENGCLS groups.

Table 1 Clinicopathologic characteristics of EGCLS as compared to ENGCLS
Full size table

Further clinicopathologic analyses of EGCLS patients

Clinical progress and outcomes of EGCLS patients

Figure 3 shows a flow chart of the 70 EGCLS patients. As a primary treatment, 52 patients underwent surgical resection and 18 underwent ESD. Seven patients who were beyond the expanded criteria after ESD accepted additional surgery [30]. Of the 59 patients who underwent surgery, LNM was present in two (3.4%). Recurrence (a localized splenic hilar mass) was observed in one patient who had shown no LNM 38 months after surgery; this patient was cured with surgical resection. Of the 11 patients who underwent ESD without surgery, one was lost to follow-up and 10 had no recurrence with a mean follow-up of 37.2 months (Table 2). Of 69 patients excluding one who was lost to follow-up after ESD, two (2.9%) died of surgery-related acute respiratory distress syndrome and septic shock 39 months after surgery during a mean follow-up of 49.5 months (range 0–108 months). None of the patients died of cancer-related causes.

Fig. 3
figure 3

A flow chart of 70 patients with early gastric carcinoma with lymphoid stroma

Full size image
Table 2 Clinicopathologic characteristics of EGCLS patients who were treated by ESD alone
Full size table

Pathologic analyses of EGCLS

EGCLSs were histologically classified into 30 typical GCLSs (42.9%) and 40 CLRs (57.1%). Fifty-seven EGCLSs (81.4%) were positive for EBV by in situ hybridization. Of the 59 patients who underwent surgery, five (8.5%) had MSI-high tumors.

Meanwhile, only four cases (5.7%) were identified as a carcinoma with lymphoid infiltrate when we reviewed the pretreatment biopsy slides.

Endoscopic and endosonographic features of EGCLS

Sixty-two EGCLSs (88.6%) showed redness, while white discoloration was noted in four (5.7%). Ulceration was observed in nine patients (12.8%), and six (8.6%) had SMT-like presentation. Of the 34 patients who underwent endoscopic ultrasonography (EUS), 21 (61.8%) showed marked mucosal hypoechogenicity or hypoechoic submucosal nodules.

Prevalence of LNM in EGCLS as compared to ENGCLS

Table 3 shows LNM rates of EGCLS as compared to ENGCLS according to tumor histology and depth. The prevalence of LNM was 3.4% in EGCLS and 10.3% in ENGCLS (P = 0.084). Among mucosal cancers, LNM rates were 0.0% in EGCLS and 3.0% in ENGCLS (P = 1.000). Submucosal EGCLS had a lower LNM rate of 4.0% compared to 19.4% of submucosal ENGCLS (P = 0.007). According to the depth of submucosal invasion, SM2 EGCLS had a lower LNM rate than D-ENGCLS and U-ENGCLS (P = 0.026); SM3 EGCLS also showed a lower LNM rate than D-ENGCLS and U-ENGCLS (P = 0.026).

Table 3 Prevalence of LNM according to tumor histology and tumor depth in surgical patients
Full size table

Two cases of EGCLSs with LNM in our study were SM3 cancers with submucosal invasion depths of 1775 and 2625 µm from the muscularis mucosa.

Discussion

GCLS is a rare subtype of gastric carcinoma presenting unique clinicopathologic features from those of conventional adenocarcinoma [10–19]. Since Yanai et al. described the morphologic characteristics of seven patients with early EBVaGC with lymphoid stroma [23], only a few case reports or series have been presented [25–29]. To our knowledge, this is the largest study to investigate the clinicopathologic features of EGCLS.

GCLS is predominant in men (male: female ratio, 2–3.4:1), associated with multiple and remnant gastric cancers, often of the diffuse histologic subtype, and tends to be located proximally with a characteristic appearance of large-thickness-to-width ratio [5, 13, 18, 19, 33]. The present study also revealed consistent findings in EGCLS.

Similar to prior reports, EBV positivity was detected in 57 EGCLS patients (81.4%), and five (8.5%) had MSI-high tumors [11–13]. Although it is unclear whether EBV or MSI status is independent prognostic factors, tumor-infiltrating lymphocytes are thought to be associated with a low LNM rate and improved survival, representing a host immune response against tumor cells that prevents metastasis [10, 11, 20–22]. Park et al. described LELC as a less advanced disease than non-LELC in terms of invasion depth and lymphatic invasion at diagnosis [18]. However, LELC was not shown as an independent prognostic factor for gastric cancer because the 5-year survival rate of such patients was not different from that of non-LELC patients. Another study concluded that EBV positivity in GCLS indicates improved outcome as compared to conventional adenocarcinoma, but stage I and II gastric cancer show excellent prognoses regardless of EBV positivity or histopathologic type [19]. Song et al. suggested that the prognosis of EBVaGC depends on the patient inflammatory response and subclassified EBVaGC based on the pattern of host inflammatory responses as typical LELC, CLR, and conventional adenocarcinoma, concluding that the LELC definition should be expanded to include CLR because CLR showed a similar prognosis as typical LELC [10]. Although it is difficult to compare EGCLS patients’ clinical outcomes between those with typical GCLS and CLR owing to the relatively small numbers of cases and a very low LNM rate, we hypothesize that in EGCLS, CLR confers a similar prognosis as typical GCLS. Meanwhile, the majority of GCLS is associated with EBV infection; thus, clinicopathologic features of GCLS and EBVaGC significantly overlap, presenting challenges in discussing these diseases. As lymphoid infiltration itself is a strong indicator of LNM regardless of EBV infection, we suggest that GCLS is more appropriate for discussing clinical outcomes of gastric cancer than EBVaGC, and EBV positivity is a secondary matter.

Previous studies have reported that early EBVaGC predominantly appears superficially depressed or SMT-like and has hypoechoic submucosal nodules on EUS [5, 23, 24]. Our study analyzed a larger sample size, and we found that the macroscopically depressed type was predominant in EGCLS. However, as compared to ENGCLS, EGCLS was more frequently associated with elevated tumors. SMT-like presentation was observed in 8.6% of patients, which is higher than 0.1–1.0% for conventional adenocarcinoma [31]. Because prominent lymphoid stroma composed of tumor cells and infiltrating lymphocytes usually lead to marginal elevation or SMT-like protrusion, we assert that EGCLS is more closely associated with macroscopically elevated features. On EUS, marked mucosal hypoechogenicity or hypoechoic submucosal nodules corresponding to dense lymphocytic aggregations were observed in 61.8% of EGCLS patients. Although homogenous hypoechoic submucosal nodules suggest submucosal EGCLS, mucosal EGCLS could also show marked mucosal hypoechogenicity. It is difficult to recognize EGCLS via biopsy; we only identified four cases (5.7%) as a carcinoma with lymphoid infiltrate via biopsy review. Therefore, EUS findings could suggest EGCLS before treatment.

ESD is widely accepted as a minimally invasive curative treatment for EGC, and predicting lymph node status is the most critical factor in determining the treatment strategy for EGC. Consistent with previous reports, our study showed LNM rates of 3.0 and 19.4% in mucosal and submucosal ENGCLS, respectively. Regarding the risk of LNM using variable “tumor differentiation,” Gotoda et al. suggested that undifferentiated intramucosal cancer without ulceration smaller than 2 cm could be a candidate for curative ESD as an expanded criteria [30]. Conversely, recent studies have claimed that patients with poorly differentiated adenocarcinoma are not suitable for ESD owing to the increased risk of LNM and a high rate of non-curative resection. However, this standard seems inappropriate for EGCLS. EGCLS has unique features with a very low LNM rate despite generally exhibiting poorly differentiated histology and deep submucosal invasion. Despite a lack of detailed pathologic reports, Watanabe et al. reported 17 submucosal invasive EGCLSs without LNM,5 and several case studies have reported EGCLSs treated by endoscopic resection even with deep submucosal invasion [25–29]. Recently, four EBV-associated EGCLSs were successfully treated via ESD [29]. All cases showed submucosal invasion of >500 µm (range, 1800–2500 µm): one patient who underwent additional surgery had no LNM and the other three had no local recurrences or distant metastases during the 27–32 month follow-up after ESD. In our analysis of 59 EGCLS patients who underwent surgical resection, mucosal EGCLS showed no LNM and submucosal EGCLS showed a significantly low frequency of LNM (4%). Although there were few patients with SM1 or SM2 invasion, no LNM was seen, and only two patients with SM3 invasion had LNM. Despite such favorable outcomes, surgical resection is considered the primary treatment for EGCLS, and most patients with EGCLS undergo additional gastrectomy after complete resection by ESD because of undifferentiated-type histology or deep submucosal invasion. Patients with EGCLS might be good candidates for ESD, even those with EGCLS with deep submucosal invasion. We cautiously suggest that ESD is acceptable for SM1 or SM2 EGCLS and could be considered in selected SM3 cancers, especially in patients with severe comorbidities or high operative risk. Although it is difficult to suggest a cutoff value of permissible submucosal invasion for ESD in SM3 EGCLS, we consider 1500 µm appropriate based on previous studies and our data.

This is the first large study providing information on EGCLS. However, there are several limitations. First, the sample size remains insufficient to make definitive conclusions owing to the rarity and excellent prognosis of EGCLS. Further studies are necessary to determine more reliable criteria for ESD and define other prognostic factors affecting LNM. Second, we did not investigate clinical outcomes of control patients with ENGCLS. However, patients with EGC show very good outcomes, and the aims of our study were to analyze the clinicopathologic features of EGCLS and define the feasibility of ESD for EGCLS. Despite these limitations, this study is meaningful because we confirm EGCLS as an entirely different subtype of gastric cancer and provide a basis for considering ESD in EGCLS.

In conclusion, EGCLS is a specific type of gastric cancer with a very low LNM rate despite deep submucosal invasion. Clinicians need to be aware of EGCLS, as its early determination is important. EUS is a useful diagnostic tool, as EGCLS typically demonstrates marked mucosal hypoechogenicity or homogenous hypoechoic nodules in the submucosal layer. ESD is a feasible curative strategy for EGCLS despite SM1 or SM2 invasion. In SM3 EGCLS, ESD can be a potential curative treatment, particularly in patients with severe comorbidities, high operative risk, or poor performance. However, ESD should be approached cautiously, and careful considerations after the final pathologic report with intensive follow-up are required before a consensus is established.