diff --git a/jupytext/cell_to_text.py b/jupytext/cell_to_text.py
index c5ae835f5..d72ca25ba 100644
--- a/jupytext/cell_to_text.py
+++ b/jupytext/cell_to_text.py
@@ -36,8 +36,8 @@ def __init__(self, cell, default_language, ext, comment_magics=None, cell_metada
         self.ext = ext
         self.cell_type = cell.cell_type
         self.source = cell_source(cell)
-        self.metadata = copy(cell.metadata)
-        filter_metadata(self.metadata, cell_metadata_filter, _IGNORE_CELL_METADATA)
+        self.unfiltered_metadata = cell.metadata
+        self.metadata = filter_metadata(copy(cell.metadata), cell_metadata_filter, _IGNORE_CELL_METADATA)
         self.language = cell_language(self.source) or default_language
         self.default_language = default_language
         self.comment = _SCRIPT_EXTENSIONS.get(ext, {}).get('comment', '#')
@@ -159,6 +159,12 @@ class LightScriptCellExporter(BaseCellExporter):
     """A class that represent a notebook cell as a Python or Julia script"""
     default_comment_magics = True
 
+    def __init__(self, *args, **kwargs):
+        BaseCellExporter.__init__(self, *args, **kwargs)
+        for key in ['endofcell']:
+            if key in self.unfiltered_metadata:
+                self.metadata[key] = self.unfiltered_metadata[key]
+
     def is_code(self):
         # Treat markdown cells with metadata as code cells (#66)
         if self.cell_type == 'markdown' and self.metadata:
@@ -308,6 +314,10 @@ def __init__(self, *args, **kwargs):
         BaseCellExporter.__init__(self, *args, **kwargs)
         self.comment = '#'
 
+        for key in ['cell_marker']:
+            if key in self.unfiltered_metadata:
+                self.metadata[key] = self.unfiltered_metadata[key]
+
     def code_to_text(self):
         """Not used"""
         pass